Kurdistan refineries selling gasoline
The first of three new refineries in Iraq's northern Kurdistan have begun to sell gasoline. The Khabat refinery, west of the city of Erbil, currently produces at a capacity of 40000 barrels per day which is set to rise to 80000 barrels a day within 18 months. Iraq's central government has a $20 billion-plan to build four more large refineries to almost double capacity and reduce dependence on imports. Iraq, holder of the world's fifth-largest oil reserves, generates most of its revenue from crude exports. The country imports about a quarter of all the refined products it consumes, according to data from the U.S. Energy Information Administration.
Oligopoly says German Cartel Office
Germany's Federal Cartel Office said recently that it believes the country's gasoline market is dominated by few companies, and said mergers and takeovers will only be allowed under strict conditions. The Federal Cartel Office still thinks that oil companies Royal Dutch Shell plc, BP plc, ExxonMobil Corp., ConocoPhillips and Total SA form a market dominant oligopoly on regional filling station markets, the antitrust authority said. Shell's German unit earlier this month acquired 41 filling stations from German retailer Edeka, fewer than previously intended after the cartel
Copyright The views expressed in print are those of the author and do not necessarily represent those of the publisher, McLean Events Europe Ltd. All rights reserved. No part of this publication may be reproduced, stored in a retrieval system or transmitted in any form or by means electronic, mechanical, photocopying, recorded or otherwise without the prior permission of the copyright holder.
office had opposed the original plan on competition grounds. The cartel office will present the findings of a comprehensive sector probe of the gasoline market, which was launched in 2008, at the end of January 2011.
Shell to close German Harburg refinery
Shell plans to shut its 110000 barrels per day (bpd) German Harburg refinery and convert the facility into a storage site after failing to find a buyer. Shell has been trying to sell the unit for nearly two years and was previously in non-exclusive negotiations with India's Essar Oil. Harburg will continue normal operations until the second quarter of 2012, the company said. One unit of the refinery used for base oil manufacturing is still for sale and the company said it is now in negotiations with potential buyers.

The Serbian government have removed price caps on gas, diesel and heating oil to coincide with a liberalization of oil imports said Deputy Energy Minister Dusan Mrakic. This will end Naftna Industrija Srbije AD's monopoly on oil imports, allowing fuel retailers to buy products abroad without having to rely on NIS refineries, and determination of retail prices may be left entirely to the market, according to a report.
Estonia demands VAT deposits from retailers
The state plans to start asking fuel companies for a 100000-euro deposit in an effort to bring massive VAT fraud under control A draft amendment endorsed by the finance committee of Parliament is designed to address a situation in which 60 percent of the motor fuel in Estonia could be termed contraband, as no value-added tax has been paid on it. The revenue forgone by the state each year amounts to over 40 million euros. This is a very large and systemic problem for one sector, said Egon Veerme, Deputy Director General of the Tax and Customs Board.
CEO of Slovenian oil company Petrol resigns
Svetelsek with Tomaz Berlocnik, chief executive of Slovenia's energy and tourism group Istrabenz. Svetelsek's resignation followed local media reports of poor management. The new CEO Berlocnik had been a member of a supervisory board of Petrol in the past. Petrol is also the largest owner of Istrabenz, with a stake of 32.63 percent. Petrol, which has market capitalization of about 549 million euros, operates some 440 filling stations in Slovenia, Croatia, Bosnia, Serbia, Kosovo and Montenegro.
The chief executive of Petrol, Slovenia's largest fuel retailer, has resigned. Petrol's supervisory board will replace Aleksander
New importation rules announced in Serbia
Serbia has announced new rules on crude oil derivative imports and refining as the market prepares for the end of the monopoly held by Naftna Industrija Srbije AD. The new rules for the 3 billion euro ($3.96 billion) retail market will let fuel retailers import oil products independently from NIS from Jan. 1st. The companies are also free to continue buying the products from local refineries run by NIS, Serbia's biggest oil company and majority owned by OAO Gazprom Neft. The government has issued 226 licenses so far for oil products trading. The market will now have the same conditions for all importers and dealers, Snezana Lukic, a Ministry of Trade adviser, said at a meeting in Belgrade to representatives from NIS, local units of Mol Nyrt, OMV AG and OAO Lukoil.

Romania to sell off Petrom shares
T he Romanian government plans to reduce its interest in Petrom, the largest subsidiary of Austrian oil and gas firm OMV, it has been reported. Romanian newspapers claim that the countr y's economy ministry is considering halving its current 20percent interest. OMV has said it has no plans to increase its stake in Petrom since it already holds 51percent and one share in the company. The Romanian government could rake in around 425 million euros by selling half of its Petrom stocks and finance a planned 600 million euros capital increase of Petrom. The company achieved a turnover of 3.02 billion euros last year. It is the biggest oil and gas group in Romania. Petrom also manages more than 260 petrol stations in Moldova, Bulgaria and Serbia.
Alliance Oil company expands in Siberia
Stockholm-listed Alliance Oil Company Ltd and diesel per year. The takeover is in line has acquired seven filling stations in the with Alliance Oil strategy to optimise and Republic of Buryatia in Eastern Siberia, expand its retail network, managing director Russia. The stations, located mainly in the Arsen Idrisov said. Following the acquisition, city of Ulan-Ude and along the Irkutsk-Chita the company operates a network of 268 petrol highway, will be rebranded and operated stations and 16 oil product terminals in the under the Alliance Oil brand. The sold as- Russian Far East and Siberia. In January sets currently market about 23000 tonnes, June 2010, Alliance Oil sold 1.7 million corresponding to 185000 barrels, of petrol barrels of oil products to its retail customers.
Gazprom Neft buys 18 percent of Sibir Energy from Serbia
Gazprom Neft, the oil arm of Gazprom, has bought 17.97 percent of shares in Sibir Energy, which develops oil and gas fields in Siberia and operates petrol stations. Gazprom Neft now holds 51.69percent of Sibir Energy. Last July, Gazprom Neft sold a 3.02percent stake in Sibir Energy to the Moscow government. Sibir Energy's core assets are the Salym oil fields in the Siberian Nefteyugansk area, operated together with Dutch Shell via a joint venture. The company also holds slightly less than 50percent of shares in the Moscow Oil Refinery and operates 138 filling stations in the city of Moscow.
BP shares rise on possible Rosneft deal
BP shares rose 1.5 percent as investors reacted to a share-swap deal with Russia's state-controlled Rosneft, which gives the British company access to areas of the Arctic previously reserved for Russian oil groups. The deal covers huge areas of the South Kara Sea in the Arctic that BP said could contain billions of barrels of oil and gas previously off limits to foreign companies. Billionaire Russian investors in TNK-BP are examining whether the London-based oil major's exploration deal with Rosneft violates the terms of their partnership. AAR, the consortium through which the oligarchs hold their stake, believes BP has agreed to only pursue new ventures in Russia via TNK-BP.Oil company Yukos has also fired off a warning to BP shareholders, saying their deal with the Kremlin is founded on illegal auctions, bogus bankruptcy fire sales and expropriations. Yukos has urged BP investors to question the real ownership of 80pc of Rosneft's assets. There has also been disquiet in the United States over the deal. US Congressman Edward Markey, who is the top Democrat on the House Natural Resources Committee, immediately called for a review of the deal by US regulators to see whether it affects the national and economic security of the United States. The venture underscores Europe's dependence on Russia for a rising share of its energy needs particularly for clean-burning natural gas. Russia holds one-fifth of the world's reserves of natural gas. Chris Huhne, British Secretary of State for Energy and Climate Change, has welcomed the groundbreaking deal and called it good news for Europe, for the UK's energy security and worldwide.

NNPC to reactivate floating fuel stations
State-owned Nigerian National Petroleum Corporation informed it will reactivate the 12 floating fuel outlets in the oil-producing Niger Delta abandoned due to violence by militants. The 12 fuel outlets, each holding 300000 litres (around 2516 barrels) of petroleum products, were built in 2006 in five southern oil producing states of Bayelsa, Rivers, Ondo, Akwa Ibom and Delta, to ease access to products by villagers in the oil creeks. The reactivation program is seen as a major boost to President Goodluck Jonathan who recently clinched the ruling party's ticket
in the primaries. Jonathan has trumpeted the relative stability in fuel supply across Africa's most populous country, as one of his key achievements since he took power in May 2010.
China independents import fuel to make diesel
Independent Chinese oil dealers have been importing an unusual fuel in recent months that can be easily turned into diesel to plug a domestic shortage, in transactions that could be legally contested. Up to 700000 tonnes (5.5 million barrels) of power kerosene, amounting to 60 percent of China's total light diesel imports, a fuel that can be turned to diesel after simple skimming or blending, have been imported since last August through southern and eastern Chinese ports. Mostly sourced from Asia oil hub Singapore, oil major BP and one or two Western trading houses were among the suppliers of the product that is exempted from a hefty consumption tax of about 830 yuan ($125) per tonne, which would otherwise be levied on diesel or aviation fuel.
Vietnam: HCM City to close old and unsafe petrol stations
The municipal administration plans to shut down petrol stations that fail to meet safety standards and other criteria set by the HCM City authority. Relevant agencies will check the operation of all petrol stations and report the results in the first quarter of the year. 61 petrol stations were asked to close in 2006 because they failed to meet the standards including 54 that did not have a construction licence. An additional 108 petrol stations were also found to lack some safety and planning requirements. HCM City has nearly 500 petrol stations, according to the city's Department of Industry and Trade.
BP enhances ability to provide differentiated fuels
Oil giant BP says that its R40-million investment in a fuel technology centre in Johannesburg demonstrates the company's commitment to South Africa. The launch of the centre is part of the oil major's long-term investment strategy into the country, which it views as a key growth market. Over the past year alone, BP has spent more than R500-million on growing its forecourt infrastructure, expanding fuel delivery capacity and optimising refining in the country, said BP South Africa CEO Sipho Maseko. The company employs about people in the country and has a strong forecourt network of over 600 service stations.

Vietnamese traders want lower import tariffs
The Ministry of Finance has said some petrol importers filed a petition asking for import tax cuts because they find it difficult to keep prices stable, as ordered by the government. The tax rates are expected to decrease by 45percent. Earlier, the Ministry had approved to reduce petrol import tax from 17percent to 12 percent, while petroleum and diesel is to be reduced from 10 to five percent. The Ministry has assessed that the increased oil price would make it inevitable for petrol and oil businesses to suffer losses. The import tax reduction is aimed at sharing difficulties with businesses and implementing the Prime Minister's instruction on slowing down price increase throughout the first quarter of 2011.
Fuel queues resurface in Zimbabwe
Fuel queues resurfaced at filling stations in and around Bulawayo as prices went up, resulting in the re-emergence of the black market due to low supplies in the city. After a respite of several months, long-winding queues resurfaced a week ago prompting a quick response from illegal dealers who have hiked prices. The price of petrol has risen from US$1,20 per litre to as high as US$1,50 per litre on the black market, but at a few filling stations with stock, a litre is offered at US$1,30.
Expanding retail network in Philippines
Independent oil player Flying V will expand its retail network next year by adding 152 filling stations bringing to 387 the number of Flying V stations nationwide. Flying V plans to raise the P1 billion through additional equity investments from existing and new stockholders, and by converting assets to cash. Some 65 percent of the capital expenditure will be spent on expanding its retail network, building more retail stations in the Visayas and Mindanao as well as in northeastern Luzon, the oil firm noted. The
rest of the capex will be used to bolster its logistics infrastructure and increase the capacity of its five existing depots, Flying V said.
Oman Oil launches Fuel, Shop and Win
Continuing its signature rewards and value add- reached more than 38000000 customers annuing offerings, Oman Oil Marketing Com-pany ally. Covering all Governorates and regions of (Oman Oil) has launched a new four-month the Sultanate, Oman Oil's growing network of Grand Retail Promotion under the theme 118 filling stations and 67 Ahlain convenience Fuel, Shop and Win in association with Suhail stores continue to witness widespread demand Bahwan Automotives and Samsung Electronics. and popularity, as they continues to set new The Grand Retail Promotion comes in light benchmarks and operational standards in the of their strategic rewards platform which has local retail market.
Sri Lanka stateowned stores expand with fuel stations

The Cooperative Wholesale Establishment also known as Lak Sathosa network of department stores will be expanded. To include fuel stations, pharmacies, cafeteria, banks, and bill payment centers, all under one roof to provide customers the convenience of one-stop shopping. The first of this kind is to be opened soon in Kilinochchi in the Northern Province.
Shell stations in New Zealand to get new payment systems
Greenstone Energy, the joint venture between Infratil and the NZ Superannuation Fund, which bought Shell's network of New Zealand petrol stations last year, has announced it will replace all point of sale systems across its Shell-branded service stations and truck stops. Greenstone signed a contract with local technology company Fusion Transactive to replace all such systems at company operated service stations nationwide. In a statement announcing the move, Greenstone's general manager of retail, Mark Forsyth, says the US$10 million investment in new technology will increase the speed of transactions and improve customer service.
Qatar Petroleum and Shell sign energy deal
The agreement was signed in Doha by His Excellency Abdulla bin Hamad Al-Attiyah, Deputy Prime Minister and Minister of Energy and Industry of the State of Qatar, and Peter Voser, Chief Executive Officer of Shell. The scope under consideration would include a mono-ethylene glycol plant of up to 1.5 million tonnes per annum using Shell's proprietary OMEGA (Only MEG Advantaged) technology and other olefin derivatives to yield over 2million tonnes of finished products. Minister Al-Attiyah said: This agreement represents an important step towards implementing the optimal utilisation of the country's natural gas resources and to expand the downstream industries in Qatar.
Kenya introduces price controls on fuel
Kenya has introduced controls on petroleum prices after the cost of fuel rose by at least 12 percent last year. The new regulations will set limits on wholesale and retail fuel prices Energy Minister Kiraitu Murungi said. Controls will also be imposed on importers of petroleum products, refineries, storage facilities, transporters and sellers. Wholesale profit margins will be capped at 6 shillings (7 US cents) per litre (0.26 gallons) and retail margins at 3 shillings, Murungi said. Fuel prices rose to 96.2 shillings a litre in September from 86.2 shillings in January, according to the Kenya National Bureau of Statistics. Oil futures in New York increased by 0.8 percent over the same period.
NNPC to construct 3greenfield refineries
The Nigerian National Petroleum Corporation (NNPC) disclosed that three new Greenfield refineries are set for construction. More are expected to come on stream to upgrade its production effort from its current level of 80000 barrels a day to 150000 barrels in the next few years. Group Managing Director Austen Oniwon, said: We have already begun operations in about 304 affiliated filling stations nationwide to ensure regular availability of products which will provide business opportunities and create employment.

Government to unbundle Noczim of Zimbabwe
The Zimbabwe government plans to unbundle the National Oil Company of Zimbabwe (Noczim) into two separate entities saying it will not place restrictions on investors. The government has said it would be willing to sell stakes way above the statutory ceiling of 49percent. One entity would be responsible for national fuel depots and infrastructure, and the task of managing the importation of petroleum products. The second entity would control fuel retailing. Several players in the oil sector have been calling for an overhaul of Noczim to transform it into a regulatory board overseeing the petroleum industry instead of maintaining its dual role of importing and distribution fuel while at the same time controlling competition. This gave the company an inordinate advantage over private players who would also pay it for using the Harare-Feruka-Beira pipeline.
UK service station crisis
Retail Motor Industry Federation chairman Brian Madderson has said that England's refuelling network is under grave threat. Independent fuel service stations in rural areas will be a thing of the past by the end of the decade because soaring fuel prices will have forced them out of business. He predicted that 500 fuel retailers the bulk of whom were in rural areas would go out of business in the next 12 months with the loss of about 5000 jobs. This will continue as long as fuel prices continued to soar, he said. He added I fear the whole refuelling network in England is now under threat. Already some of our members are reporting a 10percent 20percent reduction in business.
Afghanistan awards first oil development deal
The Afghan government has awarded its first oil development contract,regarded as the first phase in a process that could potentially create enormous revenue streams for the region. Afghan and US officials expect the deal to place one of the world's least economically attractive regions on the industry map. As part of a sixmonth crude oil production deal, the Afghan Mines Ministry will operate several wells that are to be uncapped in the Angot field, which straddles the Afghanistan / Turkmenistan bor-

der in the Sar-I Pol province of Afghanistan's north. Ghazanfar Neft Gas, a local family-run firm that owns petrol stations in Afghanistan, will collect and market the crude.
Petron to expand market reach
oil industry, the company is continuing its service station network expansion, Petron told the Philippine Stock Exchange. Petron said it will continue to establish micro-filling stations. Last year, Petron said it is planning to put up 5000 micro-filling stations as part of its aggressive marketing strategy. We intend to pursue this expansion program to further enhance customer convenience and bring the Petron brand closer to Filipinos, Petron Chairman and CEO Ramon Ang said. Petron, controlled by conglomerate San Miguel Corp., owns a 180000-barrels-a-day refinery in Bataan, supplying nearly 40 percent of the country's total fuel requirements.
Petron Corp., the country's largest oil refiner and retailer, is spending about P1billion a year until 2016 to widen its market reach and maintain its leadership in the oil industry. At present, Petron has some 1500 retail stations nationwide, up from around 1300 stations last year. To protect its leadership in the domestic
eMaintenance and Asset Management
eMaintenance is a comprehensive software application that automates maintenance call handling, contractor, safety compliance and management processes right through from call placement to contractor payment. eMaintenance has been adopted globally by BP retail because of its compelling business case which is all about cost saving and efficiency. eMaintenance is now sponsored by the American Petroleum Institute in the USA and more recently has been adopted by Murco Petroleum in the UK. The Modules and benefits are shown below. Helpdesks and call handling drastically reduced (Typically 80percent removed). Field based managers real world examples show a 3050percent reduction in involvement with no loss in service. Invoicing process highly efficient, hence lower process cost Customer example 2005 cost US$32 to process each invoice, 2010 using eMaintenance and the automated SAP interface US$1. Lower charges , eMaintenance identifies invoices to dispute based on cost, labour & parts agreements, Approx 8percent of invoices are disputed and of those 90percent have a reduced request for payment which equals a saving. Customer example = over US$1million per year saved in this area alone. SLA adherence & EFH eMaintenance increases SLA performance giving higher asset uptime. EFH (Equipment Failure Hours) measures key money generating asset SLA performance. Customer example starting SLA around 50percent, increased to around 85percent. This is worth millions. Incident Reporting tracking and handling incidents in a real time manner saves money and reduces risk to the business. eMaintenance Risk module reduces exposure by managing all risks in the most efficient way using defined logic to get the data to the right people quickly (Insurance, legal, police, regulatory bodies, internal risk teams etc). Document Control All approvals, licences, certificates and drawings, easily uploaded and stored against each site by category. Faxed hard copies uploaded to the web! Expiry dates logged with auto reminder email and SMS. A compliance dream! Spontaneous audit and control real time geomapper shows every contractor on every site real time. Mouse hover shows detail of every task being carried out. On line audit function allows audit form to be selected from drop down menus and spontaneous audits to be carried out. Revolutionised the true audit world for BP. On line surveys selectable web form allows contractors and sites to tasked with specific survey tasks and response dates. Reports on any non response. Upload of response data into asset database. Asset Management complete history of every asset, with MTBF, calls, performance, total costs at the click of a button. Makes purchasing and contractor decisions easy. Real Time every item in the database is time and date stamped and available real time forever. Enquiries, free demonstrations, webinars and pilots available. www.urgtech.com

Smartflex develops new technology
Nupigeco developed a new technology for its electro-fusion fittings. The welding wire of all reducers (mounted on the fittings to join them to the secondary pipe of the double wall Smartflex fittings from 63) will be replaced by a new polyethylene coated wire. This innovation dramatically reduces the welding time by almost three times without compromising the reliability of the junction and without affecting the production efficiency. This new technology will be applied to the whole Smartflex double wall fitting range 63, i.e. Elbow 90, Elbow 45, Tee, Coupling, etc. The same change will be made soon to the reducers of double wall fittings 50.
11 feature On the road 11 11
So ma ny peo ple to see and so
Three months out and the list of atten dees is growing every day. Suppliers have woken up after the Christmas break and realised that May is not that far away and the oil companies are finalising their decisions on who will represent them in Barcelona. As we move into February, trips around Europe to Spain, Portugal, Finland, Germany, Holland, Austria, Slovenia, Poland, Turkey and Russia meeting new retailers, becomes the main priority, especially if they have never attended erpec before.
Lisbon atmospheric view of the city centre near Rossio
Last month we went to visit Cepsa in Madrid and followed that with a trip to see Petrogal in Lisbon. Petrogal, very often referred to simply as Galp, have attended every erpec event in the past, but changes in the Procurement department headed up by Joo Bastos, has meant some new people to meet and following a positive one hour meeting, a decision will be made shortly concerning their participation. At Cepsa we met with Head of Planning and Technology, Jos Molina Hita, who was
joined in the meeting by Chief Engineer Operating 1534 sites, Cepsa is the second largVicente Roures Soler and Directora General est petroleum retailer in the Spanish market, de Promimer, Beln Mateo Ercilla. Attend- which in total accounts for just over 9000 retail ing erpec was very much on the agenda, sites. Spain's largest retailer is Repsol, with throughout an extremely positive meeting, 3600 outlets. Included in general discussions which was concluded with Sr Molina saying was Cepsa's convenience retailing initiative, that serious consideration would be given to branded Depaso which is featured strongly on having Cepsa represented for the first time a huge Cepsa site, just a short walk from their in Barcelona. For erpec should Cepsa attend, offices. With an extensive shopping area and it would be a significant step towards having a significantly sized stand alone coffee shop every major European oil company in attend- and restaurant, branded separately as Como en casa, it showed itself to be an extremely ance for the 2011 event.

Cepsa Head of Planning Jos Molina Hita, Chief Engineer Vicente Roures Soler and Directora General de Promimer Beln Mateo Ercilla (from left to right)

feature On the road

mu ch to do!

By Nick Needs

Galp conveniece store branded Tangerina
impressive retail operation and demonstrated clearly Cepsa's future c-store retailing commitment. On technology, the most interesting point to emerge from the meeting was to learn that Cepsa manufactures approximately 50percent of their own fuel pumps, through their associate company Cedipsa and apparently, are the only major oil company in Europe to do so. For reference, the remaining 50percent are supplied by Dresser Wayne.
During the 2011 campaign we have 11 discovered many changes since the last event in 2009. What follows is a brief digest of what we have learned so far.
Nick Needs interviewing Purchasing Director of Galp Energia, Mr. Joo Bastos.
Henry den Toom is a name many will remember from two past events, as he headed up the Chevron Texaco operation in Brussels, which was then bought by Israel oil company Delek. Henry is now the man in charge of a rapidly growing retailer in Holland and Belgium called Argos. As for Delek Europe, they have re-located to Breda in Holland having recently bought the BP network in France. Construction, design and procurement for Delek Europe is headed up by Mr Alon Hasdai and he is supported by Team Leader Procurement, Anita Gunters. At Coteba Michel Clement is now the Multisites Petrol Projects specialist and he will be attending erpec for the first time. At Deutsche BP, Senior Procurement Manager Hermann Ketelaers has moved on within the company and has been replaced by Tim Erichen and Kamila Lipa who is also responsible on a European level for BP's car wash business. From Deutsche Tamoil,
Cluster Filling stations & Retail Related the two delegates attending erpec 2009, Services. Frank Sternberg and Wilfried Thiede are no longer in their positions, but Carsten At Opet in Turkey, following the departure Pohl who heads up this division has said of Gurdal Orak, Hakan Koka is now the that there are new representatives to take person responsible for Retail Engineering. their place, which he will inform us about Rik Nieuwstraten, former Global Design shortly. and Construction Manager at Shell, who has attended many erpec events in the past, has Changes at ExxonMobil, has meant that Michal Tesar Procurement Team Leader moved on to pastures new and unfortunately Asset Management based in Prague, has we will miss him in Barcelona this year. now moved on and has been replaced by At Total, several changes have meant Nargiz Azimova. Heading up this departFlorence Varescon, former Global Purchasment is Jiri Urban. ing Category Manager and Hans Van Lierde, former Marketing Europe/DDM Depart At MOL in Hungary, Bla Csorba is still ment Engineering Optimisation Resource, very much Director of Group Retail Nethave changed jobs. Purchasing Direction works& Asset Management, but Lszl Category Managers now include; Philippe Bartha, formerly Director of Group Retail Petitjean, Alain Gauthier, Pierre Gourlay, Sales & Business Development, has now Cline Fouriez-Arnaud and Regis Benoitchanged jobs. du-Rey. Engineering personnel include In Scandinavia, former erpec delegates Kari Georg Okuka and Olivier Sebileau. Jaanila and Ralph Hckerstedt are no longer in their former jobs and Risto Rsnen is Mark Todd has taken over from Phil Maud now the person to talk to on Resources, at UK supermarket chain Morrisons. Engineering & Purchasing. At the OMV Group in Austria, Duane A revamped website for erpec will be launched Runciman, formerly Head of Procurement during the first week of February. Look out at Petrom in Romania, has taken over the for that, which will contain lists of all those role of Head of Category Management for suppliers and retail organizations who have OMV Group wide procurement and Slobo- registered or been invited to attend. More dan Milic has become Head of Category details www.erpec.com

U.S. technology firm Honeywell will implement projects for the production of equipment for motor vehicles operating on gas in Uzbekistan, according to the Uzbekneftegaz national holding company. Honeywell and Uzbekneftegaz have determined a number of joint projects over the next three years. Firstly the American company will organize the production of engine converting equipment worth US$7 million to switch vehicles to compressed natural gas. The total production capacity will be 100000 units per year
CNG at DTC bus depots in Delhi, India
Delhi Transport Corporation (DTC) has agreed body signed a memorandum of understandto install hybrid gas dispensing machines at ing with Indraprastha Gas Limited the sole its bus depots across the Capital. These dis- distributor of CNG in Delhi in this regard. pensing machines would fill tanks of private At present, IGL has its gas dispensing facilcars, as well as autorickshaws, running on ity at 30 DTC depots; the figure is likely to compressed natural gas (CNG). The transport increase to 39 in next three months.
Portugal and Ireland are charging ahead
Two countries in Europe developing an effective electric vehicle infrastructure are Portugal and Ireland. Both have pitched themselves head first into getting ready for the new technology. Portugal has been storming ahead in EV infrastructure, with 1300 expected to be in place this year. But the infrastructure in Portugal and Ireland for that matter is surely only the tip of the continental iceberg. A public initially entranced by the idea of paying a nominal 2 Euro for 100kilometres for their shiny new EV, might quickly become disillusioned by the thought of only having access to an initial 25 cities in Portugal and what about everywhere else?
Silicon Valley of fuel cells
Japan, wants to become the Silicon Valley of fuel-cell technology as it seeks to ease restrictions on the use of hydrogen fuel for vehicles as part of the central government's special economic zones scheme to be launched this year. According to the plan, fuel-cell vehicles could be charged on public highways and expressway services, which is currently banned under the High Pressure Gas Safety Law and the Road Traffic Law. It would install further hydrogen filling stations and conduct longdistance tests of fuel-cell vehicles on highways.

Trinidad and Tobago focusing on CNG
Trinidad and Tobago's plans for increasing bill and costs of transportation and vehicle natural gas vehicles are translating to ac- maintenance through increased uptake of tion with the establishment of a government natural gas. Since the first pilot project in CNG Task Force launched in October and 1984, there have been insufficient resources the first of at least five new CNG refuel- and commitment and education to sustain ling stations commissioned last week. The the natural gas vehicle industry there but CNG Task Force is aimed at diversifying the Government is now actively seeking to the country's transportation fuel mix as accelerate the development of CNG as a well as reducing the petroleum subsidy major transportation fuel.
EC approves Royal Dutch Shell & Cosan
The European Commission approved the joint venture with Royal Dutch Shell PLC unit, Shell Brazil Holding B.V., and Brazilian sugar and ethanol firm, Cosan S.A. Industria e Comercio of Brazil. The joint venture will sell, trade and produce sugar and ethanol worldwide, as well as develop and license ethanol technologies and sell transportation fuel products within Brazil. European Union officials said in a statement: After examining the operation, the Commission concluded that the transaction would not significantly impede effective competition in the European Economic Area (EEA) or any substantial part of it.
CNG fuel stations at Los Angeles airport
Hertz has partnered with Clean Energy Fuels Corp. to build, operate, and supply compressed natural gas (CNG) fueling stations at Hertz's Los Angeles Airport facility. The Hertz/Clean Energy CNG fueling station will include public access fueling islands for local fleets of CNG taxis and shared-ride vans, plus hotel, parking, and rental car buses. A private Hertz island will be capable of fueling its internal fleet, simultaneously. Station construction is expected to begin in March and be completed
by the third quarter of 2011. By 2015, Hertz expects to have transitioned its entire LAX transit bus fleet to clean, green, domestic CNG gas power.
latest Alternative Fuel news www.PetrolPlaza.com

Websites and Logos supporting erpecnews

www.air-serv.eu

www.franklinfueling.com

www.opw-fce.com

www.gilbarco.eu www.alucobond.com www.christ-ag.com
www.hectronic.com www.aspentech.com

www.petrotec.eu

www.iisltd.com www.atosworldline.com www.planova.com
www.istobal.com www.belugacorporation.com www.psdcodax.com www.kaercher.com
www.bennettpump.com www.ruudlighting.net & www.ruudled.net
www.kpsystem.com www.bennett-sauser.ch www.scheidt-bachmann.com
www.kubald.com www.beverinnovations.com www.secu-tech.at
www.brugg.de www.leightonobrien.com www.sloanled.com
www.ceccato.it www.mepsan.com.tr

www.stc-norway.com

www.dresserwayne.com

www.tammerneon.com

www.mueller-offenburg.de www.plxpipe.com www.tanknology.com

www.eurotank.eu.com

www.nupigeco.com

www.tokheim.com

www.fafnir.com

www.ono-oil.com

www.washtec.de
If you have not yet sent us your logo and website address, please do so for the next issue by mailing editor@erpecnews.com
latest news, Events, jobs online www.PetrolPlaza.com 23

Ll 2 p

exp i (kz pl + l) r2
Fi (t) Fj (t ) = 2Dij (t t )

ik 1 + 2 2R w

where i, j = P, N, , the angle brackets denote the ensemble average, and Dij is the diffusion coefcient in the Markovian approximation. The main contribution to the laser noise is caused by the diffusion coefcients DP P and D given by
where p 0, l 0 are the radial and angular mode numbers, w0 is the spot size at the waist, k = 2/ is the free space wavenumber, Ll are the generalized Laguerre polynomials, p and pl (z, w0 , k) = (2p + l + 1) arctan 2z 2 kw0
vg a0 (N N0 ) vg a0 (N N0 ) (6) P ; D = VP V Single mode rate equations (1)-(3) have been found to be a very good approximation to the large signal behavior for MMF [5]. For the analogous applications, the relation between the SNR and the relative intensity noise (RIN) is given by

DP P =

w (z, w0 , k) = w+

2z 2 kw0

2 kw0 2z
m2 (7) SN R = 2RIN where m = I/ (I Ith ) is the electrical modulation depth, I, Ith are the input and the threshold currents, respectively. RIN is given by

R (z, w0 , k) = z 1 +

2 Popt (t)

Popt (t)

2 where Popt (t) is the mean square optical power uctuation and Popt (t) is the average optical power. MMF links performance is affected by degradation due to nite rise and fall times at the transmitter and the receiver, the intermodal and intramodal dispersion, and noises specic
For the few-moded VCSEL the Gaussian beam model is a reasonable approximation [5]. A VCSEL upl mode at the airber interface is transformed into a different Gaussian beam mode which then excites the various modes lmp (r, ) of MMF corresponding to the transverse components Ex,y of the electric eld in the ber. The modes lmp (r, ) are given by [5] (13) lmp (r, ) = flm (r) (l) p where l 0 and m > 0 are the eigen-values of the radial and angular parts of (13), the index denotes angular dependence
sin l or cos l, and polarization p = x, y. The modes satisfy the normalization condition

rdrdflm (r) 2 (l) = 0

by Xi where i refers to the ith realization. The channel coefcients kl are given by [15] kl = pkl l kl (21)
l The coupling amplitudes aplm of the incident Gaussian beam mode with the ber mode lmp (r, ) are given by [5] l aplm = A
where l is the fading associated with the lth cluster, and kl is the fading associated with the kth ray of the lth cluster. The cluster arrival time distribution p (Tl |Tl1 ) and the ray arrival time distribution p kl | (k1)l are given by, respectively [15] p (Tl |Tl1 ) = exp [ (Tl Tl1 )] , l > 0 and p kl | (k1)l = exp kl (k1)l , k > 0 (23) (22)

Response function

45mW 30mW 10mW

Power [dBm]

Fig. 4. Schematic view of an optically controlled microstrip convertor (OCMC)

400 frequency[Mhz]

by UWB RF signal, is illuminated on the substrate near the open end of the MS line. The down conversion from the optical domain to the MW domain can be modeled by an optically controlled load connected at the open end of the MS line. The variations of the photocurrent at the optically controlled load of the MS produce an electromagnetic (EM) waves that propagate along the MS line towards the output port of OCMC from which they are probed by a coaxial line of the same characteristic impedance, Z0. The efciency of the optical-microwave frequency down conversion depends on the ability to collect the photocarriers at the bottom contact. In the case of silicon technology, the thickness conventional substrates is in the range of 350 500m which is quite large compared to the diffusion length of the photocarriers Ln,p = Dn,p (10 30) m. In the case of surface absorption characterized by large values of absorption coefcient and consequently a very small absorption length 1 the effective depth the photocarriers can reach is determined by the diffusion and drift properties of the photocarriers. The feasibility of the proposed OCMC device was experimentally veried by an open-ended MS line with Z0 = 50 implemented on a high resistivity > 3000cm slightly p-type Si substrate are shown in Fig. 4. The optical source was a tunable laser diode with wavelengths from = 680 up to = 980nm. The results for the OCMC response function at the different levels of the optical power are shown in Fig. 5. These results do not satisfy the requirements of the UWB RF signal detection. An alternative approach has been proposed recently in a number of works [27]-[30]. It has been demonstrated experimentally that thin Ge-on-Si, SiGe/Si, or Si layers of a thickness about one up to several micrometers can operate successfully as UWB RF signal detectors providing a bandwidth of about (10 20) GHz [27]-[30]. A resonant cavity-enhanced Si photodetector permits to overcome the comparatively low absorption in Si by using the substrate with a distributed Bragg reector (DBR) that provides 90% reection of an optical power back into the detector layer [27]. Silicon photodetectors monolithically integrated with preamplier circuits have achieved error-free detection at up

strongly inhomogeneous behavior in such a case in the vicinity of 10GHz is caused by the VCSEL bandwidth limitations. The p-i-n PD used in these measurements has the bandwidth of about 25GHz. The bandpass lter behavior of the MMF caused by the multimode dispersion is strongly manifested for longer MMFs. The magnitude and the phase of the 650m MMF are shown in Fig. 11. In the TFC7 frequency range the transfer function magnitude curve is at. On the contrary, in the TFC6 frequency range the transfer function magnitude has a notch. The transmission of the multiplexed MB OFDM signals is limited by the MMF length of about 100m. VI. T HE E XPERIMENTAL R ESULTS A. MB OFDM Signal Transmission MB OFDM UWB signal was directly applied to the VCSEL and after the propagation through the MMF was detected by the p-i-n PD. The objective of the measurements was to study the performance of the proposed link by means of the packet error rate (PER). The measurements have been carried out for the TFC5, TFC6 and TFC7 band of MB OFDM UWB signals. Fig. 12 presents the PER versus MMF length for the optical link only. The PER dependence versus the MMF length for the different MB OFDM UWB signals shows a peculiar behavior. The PER of the TFC7 band located at higher carrier frequency (4.488GHz) stays at and has values of an order of magnitude of 106 for MMF length up to1km. However, the PER in the case of TFC5 band located at the 3.5GHz carrier frequency and the TFC6 band located at the 4.0GHz carrier frequency increases dramatically for the MMF lengths longer than 300m. In order to understand this behavior of the PER versus MMF length we have measured the MMF transfer function for different MMF lengths. The short MMF with a length of 10m has an almost at frequency response up to the frequency of 10GHz as it is shown in Fig. 13. The strongly inhomogeneous behavior in such a case in the vicinity of 10GHz is caused by the VCSEL bandwidth limitations. The p-i-n PD used in these measurements has the bandwidth of about 25GHz. The

Fig. 13.

The measurement results for the optical link frequency response

Figure 6. Experimental setup for: a) 10 Gb/s NRZ and b) 40 Gb/s RZ contention resolution
Figure 8. Oscilloscope traces for optical contention resolution of 40Gb/s data using a quad wavelength converter
Figure 9. BER results for optical contention resolution of 10Gb/s NRZ and 40Gb/s RZ
wavelength converted by the PEDs and forwarded through the third port of the coupler avoiding any collision. The Optical Spectrum Analyzer (OSA) trace indicates the presence of two wavelengths at the output of the system, namely 1558nm for packets 1, 5, 2 and 3 and 1555nm for packets 4 and 6. Fig. 8 depicts the results recorded for 40Gb/s RZ data packets. Fig. 9 shows the BER curves obtained for the deflected, time multiplexed and wavelength converted packets for both 10 Gb/s NRZ and 40 Gb/s RZ operation. The power penalties are less than 1 dB in the case of 10 Gb/s operation and less than 2 dB at 40 Gb/s. C. High-speed, all-optical wavelength conversion, label recovery and data regeneration Apart from a scalable solution for multi-channel wavelength conversion, the hybrid integrated wavelength converter arrays present a unique opportunity for multifunctionality, i.e. simultaneous performance of functionalities using multi-signal processing on a single chip. Here we demonstrate the processing power and multi-functionality of the quadruple array device by performing wavelength conversion, clock recovery, data regeneration and label/payload separation using a single photonic integrated device. Figure 10 shows the basic setup and the time-domain results. Each SOA-MZI of the array is assigned a different role operating as: wavelength converter (wavelength converter 1), amplitude equalizer (wavelength converter 2) in the clock recovery unit, regenerator (wavelength converter 3) and AND gate (wavelength converter 4) for label/payload separation. After the wavelength conversion process, the signal is fed into the clock recovery circuit which employs a low-Q FPF filter and SOA-MZI1 operating as an equalizer. The recovered clock has a double role: 1) it is used as the input signal to the 3R decision gate, where retiming and reshaping is achieved by triggering the incoming data with the retimed optical recovered clock pulses and 2) it
slot in stream A. Stream A is launched in the PED circuit which extracts the envelopes of packets 1, 2 and 3. When Stream B reaches the 1x2 switch, it is triggered by the PED signal so that packets 4 and 6 appear at the switched port, whereas packet 5 is spatially separated. Packet 5 is directly forwarded to the output together with stream A and fills the empty slot after combination of the two signals in a passive coupler. As a result, a time multiplexed packet stream consisting of packets 1, 5, 2 and 3 appears at the output, whereas packets 4 and 6 are

Figure 13: Experimental results of the 4-BMR circuit. Trace and eye diagrams of (a) incoming 6 dB power fluctuated data packets, (b) power equalized packets, (c) packet clock recovery, (d) regenerated data packets. Trace and eye-diagram time scales are 1 ns/div and 10 ps/div.
due to the timing jitter of the signal generator used to drive the EAM. Push-pull control configuration was also adopted in the HMZIs of the third array. Finally BER measurements were performed for all four data channels. Fig. 13 illustrates the evolution of the Burst Mode Regeneration process through temporal oscilloscope traces and eye diagrams obtained at each stage of the 4BMR circuit. Fig. 13(a) shows two incoming data packets at 10 Gb/s having a length of 119 and 60 bits, respectively, and exhibiting power fluctuation of 6 dB. Fig. 13(b) shows the respective power equalized packet stream obtained at the outputs of the SOA-MZIs of the first quad array. The 6 dB power fluctuation between the incoming packets has been reduced into roughly 1 dB amplitude modulation between the pulses within the power equalized packets. Fig. 13(c) depicts the recovered clock packets obtained at the output of the CR stage. They persist for time duration that equals that of the corresponding input data packet, extended on its leading edge by a 4-bit rising time and on its trailing edge by a 16-bit decay time. The former value indicates the time required by the CR to lock to the line-rate of the incoming data packet, while the latter value determines the time required by the CR signal to decay to 1/e after each packet. The guardband bits depend on the finesse of the FPF, which is dictated by the order of the PRBS [8]. Fig. 13(d) illustrates the received equalized data packets
at the output of the 4-BMR circuit, indicating timing jitter and amplitude modulation reduction, with respect to the corresponding signal at the output of the power equalization unit (Quad 1). Jitter reduction was confirmed after conducting measurements on the incoming data signal, the recovered packet clocks and the regenerated output. In particular, the root- mean-square (rms) values were 1 ps for the input, 600 fs for the packet clock and 730 fs for the regenerated signal. Fig. 14 shows the BER curves obtained for the input packets with 6 dB power fluctuation, the power equalized data packets and the regenerated data packets. Error-free operation was obtained for all wavelengths with similar results. The recovered data channels exhibit a negative power penalty of up to 2.5 dB and 1.3 dB with regard to the corresponding input data channels with 6 dB power fluctuation and the power equalizer output respectively. Error-rate measurements were also taken for the input packets with 0 dB power fluctuation that serves as the baseline for the complete BER measurement procedure. This BER curve lies between the BMR output and the power equalizer output curve and has a 1 dB positive power penalty offset with respect to the BMR output revealing the 3R capability of the 4-BMR circuit. IV. DISCUSSION The expansion of traffic growth in broadband core networks is driving the development and deployment of new transmission and switching equipment: the hardware for the new information age. The driver for implementing the photonic equipment for next generation broadband networks is the integration of photonic processing systems-on-chip with the development of multi-element, functional integration. In this context EU project ISTMUFINS has managed a major step towards the establishment of a photonic platform capable to generate such multi-element photonic chips with the demonstration of the first arrays of all-optical wavelength converters. Here we have presented how the enhanced processing power and capacity of these devices can be exploited to implement key functionalities required in next generation photonic routers. All-optical processing has been realized either by interconnecting externally the SOA-MZI wavelength converters of a single quad chip or by using multiple quads as arrays of photonic gates for

2010 ACADEMY PUBLISHER doi:10.4304/jnw.5.2.197-211
a few. In this paper, we will focus on throughput maximization in an OW OFDM-based system in a typical WLAN application. As often implemented in practical systems to improve power efciency, we will adopt the simplest approach and allow a certain controllable amount of clipping at the transmitter, assuming that the additional errors can be tolerated. A special OFDM scheme, which foresees clipping, was suggested, e.g. in [28], for a narrowband OW channel. This scheme, however, exploits only half of the channel bandwidth since it assumes modulation of odd subcarriers only. We will show how a loading algorithm, which performs best in power-limited systems, needs to be adjusted under the specic constraints that the OW channel poses on the transmit signal waveform. We will also show that a dynamically adaptive system can hugely enhance transmission rates compared to a statically designed one, even if we put a conservative constraint on the electrical signal waveform by avoiding any clipping. In addition, we will investigate the inuence of controlled symmetric clipping on system performance in terms of both error and transmission rate. The results show that the system capacity can be exploited even more by tolerating a minor increase in error rate as a result of clipping. The paper is organized as follows. A communication scenario and channel model are introduced in Chapter 2. Realistic system parameters, which are necessary for our investigations, are derived in Chapter 3, while the properties of the proposed rate-adaptive system are discussed in the subsequent section. Chapters 5 and 6 provide the performance analysis in the absence and presence of clipping, respectively, while conclusions are drawn in the nal chapter. II. C OMMUNICATION S CENARIO AND C HANNEL M ODEL A. Indoor Communication Scenario Figure 1 shows a typical example of a scenario where a non-directed LOS link could be of use. It represents communication between a xed access point (Tx) and a mobile terminal unit, e.g. a laptop (Rx) in a moderatesize room. Numerical simulations performed throughout

Tx (fixed access point)

(distance, r to the Tx) and which has a free orientation (parameter ) but the work is applicable to WLAN scenarios in general. Tx is assumed to be positioned at the center of the ceiling and pointing downwards. Tx and Rx are assumed to have wide-beam radiation characteristics and a wide Field-Of-View (FOV), respectively, so that both LOS and diffuse signals are simultaneously present at the receiver. It should also be noted that even though the access point at the ceiling is regarded in this paper as Tx and the mobile end-terminal as Rx, the communication scenario foreseen is bidirectional and the conclusions are valid for both directions. B. Noiseless Channel Model An analytical model for the OW channel of a nondirected LOS link was developed by Pohl et al. [29]. This model provides a good rule-of-thumb prediction of essential channel properties, such as path loss and bandwidth. It is simple and yet takes into account an innite number of reections unlike some the other models proposed. Unless the LOS is blocked, the impulse response of the non-directed LOS channel consists of two distinctive components - a discrete Dirac-like pulse (LOS contribution) and a continuous signal, arriving some time later (contribution of diffuse reections) at the receiver. It was, in fact, recognized in [12] that the response of the diffuse channel is similar to that of an integrating sphere and that it can be well modeled with an exponentially decaying function. Accordingly, the channel frequency response is modeled as a superposition of a at and a rst-order low-pass transfer function H(f ) = = LOS + HDIFF (f )ej2f T ej2f T , LOS + DIFF f 1 + j f0

I. INTRODUCTION When the first single mode fibre links were installed the most important fibre parameter was the attenuation measured in dB/km. Data rates up to 2.5 Gbit/s had to be transmitted over distances of up to 80 km. Attenuation and chromatic dispersion were the limiting effects determining data rates and link lengths. However, with
the advent of optical amplifiers and techniques for compensation of the chromatic dispersion optical network links could be lengthened significantly and/or the transmitted data rates markedly increased. For optical single channel transmission at 40 Gbit/s, and even higher bitrates, PMD is a huge problem for network carriers. In particular, legacy fibres, installed when PMD was not of concern for the prevalent bitrates, often exhibit very high PMD values that are not suitable for high-speed transmission in todays networks. The occurrence of link PMD can be divided into two major categories: intrinsic and extrinsic reasons. The intrinsic reason mainly arises in the production process and is based on fibre birefringence due to non-ideal roundness of the fibre and other imperfections, causing such a fibre to possess inherently a high PMD value. The extrinsic reason is caused by asymmetric pressure on the fibres and cables during or after the cable burying process. Such external influences can increase the PMD value of the fibre links. During the last few years Deutsche Telekom has reviewed its network and characterised the PMD value of large number of fibres (Fig. 1). The fibre links where the PMD value is too high for 40 Gbit/s transmission were identified. One high-PMD section can be sufficient to render the entire 400 km link bad. In order to avoid limitations of data rate and transmission length, PMDcompensators [1] can be used for each channel, an approach that can become quite costly and unwieldy. Alternatively, or in addition, advanced modulation scheme (e.g., 40G RZ-DQPSK) [2] can be applied to partially mitigate the sensitivity to PMD [3]. The third possibility for reducing the link PMD is to identify and replace the bad fibre section [4]. In fact, the latter approach may be the most economically viable solution, because replacement of the fibres along an entire link is extremely costly and time consuming. Fortunately, it is generally the case that the PMD of a
2010 ACADEMY PUBLISHER doi:10.4304/jnw.5.2.212-218
fibre link is not uniformly distributed but tends to be localized in a few distinct pieces or sections accounting for the large majority of the total link PMD [5]. A possible distribution of the PMD value over a fibre link is shown in Fig. 2. There are two fibre pieces having high PMD value - one bad fibre section between L1 and L2 and another very short piece fibre located at L3 that, for example, may be a short but high birefringence fibre which is the primary contributor to the link PMD. The other portions of the link exhibit acceptable PMD behaviour, suitable for high bitrate transmission. As a result, such replacement is much more cost effective and could be economically competitive to the deployment of PMD tolerant transmission systems.

This study further provides a performance evaluation of such a local CAC strategy in terms of capacity planning for the network control plane and bandwidth efciency. On the other hand, we also assess the responsiveness of such local CAC to the input trafc demands. It follows immediately that the more frequent the resource allocation updates occur, the more closely the CAC controller follows the input demand. However, this may lead to undesired oscillations of network resources, together with a signicant signalling overhead. In order to tackle these issues we rst evaluate the dynamic behaviour of a single CAC controller. Our aim is to nd the rate of the resource allocation updates such that the input demand is satised with a certain probability. A. Application scenarios Local CAC is originally thought to be used at the border of multi-service networks that must deal with a huge number of service requests that require QoS guarantees. As a centralized approach for handling these requests, it may show important scalability problems. The MUSE IST project (see [1]) proposes the usage of local CAC in the (aggregation network) access nodes, such as the Digital Subscriber Line Access Multiplexers (DSLAMs) or the Optical Line Terminals (OLTs) for a subset of services that have stringent QoS requirements such as IPTV trafc. Note that the implementation of a local CAC requires being locally able of handling service signalling in order to be able to make the user aware of the CAC decision result (acceptance or rejection). In this way, local CAC can easily handle multicast trafc (e.g. IP Television), as it is possible to be locally aware of Internet Group Management Protocol (IGMP) messages in most of current IP DSLAMs. The proposed technique is also suitable to be employed in the delivery of VoD services. In this scenario, there are unicast connections established between the video server and each one of the end users. So, with the application of the local CAC in the video server, it is possible to make a local decision about the network resource availability for the desired route according the assigned quota for the corresponding destination, stored in the CAC module of the video server.
2010 ACADEMY PUBLISHER doi:10.4304/jnw.5.2.225-229
For other services such as VoIP, local CAC may be used provided that the access nodes were able to participate in the signalling process. In this way, the proposal of distributing Session Border Control (SBC) capabilities [2] into the access nodes would allow to locally take the CAC decisions. RUBENS proposes to take into account several parameters to accept the request of a new application (it can be a game session, VoD service, etc). One of these inputs to make such decision is the availability of bandwidth for a session of such application, which may be performed in a scalable way by means of the local CAC. A long term application scenario consists of local CAC at the edge of an Optical Burst Switching (OBS) Network. In this light, edge nodes would have pre-established quotas per destination and class of service, assuring that the load inside the network is controlled, avoiding harmful congestion. II. A NALYSIS We provide capacity planning rules for the local CAC controller, based on the input demand (trafc matrix). Second, we analyze the timescale for updates. This is the timescale at which the capacity allocation (quota) per CAC controller should be updated. Third, a reliability analysis is presented. Finally, we perform an experiment using real trafc traces from the Spanish National Research and Education Network, in order to show the suitability of the results in a real network scenario. A. Capacity planning for the local CAC controller (packet case) Let us rst assume a packet network case in which trafc is expressed as a Gaussian uid over the graph (V, E) where V is the set of nodes and E the set of links. The input demand is expressed as a matrix T , with V rows and V columns and the uplink and downlink trafc to a certain node i is expressed as the sum of the entries of row i and column i respectively. The goal is to derive the capacity planning rule (quota assignment) for local CAC i assuming that the average trafc intensity is I in bits per second and its standard deviation is given by. The capacity planning problem is stated as follow, nd C such that P (X > C) < (1)

0 N 0 0

0 0

Unavailability ratio

0.005 0.001

0.007 0.006 0.005 0.004 Times (x average session duration)
2 Number of CAC controllers
Figure 4. Time to saturation, loaded system (90%)
Figure 5. Unavailability ratio
is much higher. In that case, we have a resource allocation request per session which implies a much higher rate. C. Reliability analysis

Processing system

Flow Collector
In what follows, let us assume that the failure probability of a CAC, either local or centralized is p, which can be obtained as the ratio between the MTBF and MTTR. Let us now compare the reliability ratios obtained by the local and centralized CAC. We consider the probability that resources are unavailable, i.e, no bandwidth allocation request can be satised in the network. This implies failure of all CACs in the local case and failure of the centralized CAC in the centralized case, which can be calculated using a Binomial distribution. The following graph shows the ratio between such probabilities (resource unavailability) versus the number of local CAC units, for a very large failure probability of 10%. The results show that the unavailabilyty ratio is as high as 10000 times as much in the centralized case, for a small network of 3 CAC controllers and very high failure rate. For smaller CAC failure rates, the unavailability ratio is even higher. III. A CASE STUDY WITH REAL DATA In this section, we consider a case study with real data from the Spanish academic network. We collect the busy hour trafc in the access links of four universities, two of them being small and the other two large. The following gure shows the topology of the Spanish academic network, along with the measurement collection infrastructure.

Data repository

Monitoring system
Figure 6. Measurement testbed in the Spanish academic network
We consider that each of the access links is controlled by a local CAC and proceed with the Gaussian capacity planning rule. The results are shown in the following table.
Mean (Mbps) Std. deviation (Mbps) 5.6 2.6 7.6 35.6 Quota ( = 0.99) 45.02 14.04 223.68 288.81

University U1 U2 U3 U4

TABLE I. R ESULTS IN A CASE STUDY WITH REAL DATA
It turns out that the variability of the busy hour is relatively small in a real case, thus making the local CAC scheme amenable for use, as the resulting quotas are not too large in comparison with the real demand.
IV. C ONCLUSIONS In this paper we have analyzed a case of local CAC. The analysis has been performed both from an analytical and from an experimental point of view, including real trafc traces from the Spanish NREN. Our results show the advantages of the local CAC mechanism, with resulting quotas that are close to the average trafc demand per CAC controller. ACKNOWLEDGEMENTS The authors would like to acknowledge the BONE Network of Excellence, partially funded by the European Union Seventh Framework Programme, the MUSE project, partially funded by the EU Sixth FP, and the RUBENS project as part of the EUREKA CELTIC initiative. R EFERENCES

The results indicated that this restriction reduces the bandwidth utilization efficiency of the ring while introducing little change in the throughput difference between nodes. Therefore this restriction is not used here. MAC_Protocol_Fixed_Size_Slot () Begin Global_Time:= time since simulation started for (each node) do { for (each logical ring ) do { if (Node_Buffer is not empty AND Current_Slot is empty ) { Move packet to the Current_Slot; Log Queuing_Delay; Transmitted_Packets ++; }}} Global_Time= +Fixed_Slot_Time; End
Fig. 3 Algorithm of the MAC protocol of the fixed size slot scheme
MAC_Protocol_Variable_Size_Slot () Begin Global_Time:= time since simulation started Current_Acting_Node:= node chosen to transmit Node_Next_Time:= Global_Time + time increase to make the node point to the next slot Min_Time:= minimum increase to the Global_Time Min_Time = random number > Node_Next_Time; for (each logical ring) do { for (each node) do { if (Node_Next_Time < Min_Time) { Min_Time = Node_Next_Time; Current_Acting_Node = Node_Id; }} Node_Next_Time =+ Slot_Time; Global_Time = Min_Time; /*Set Global_Time to the least Node_Next_Time)*/ if(Current_Acting_Node_Current_Slot is empty ){ Check the size of Current_Acting_Node_Current_Slot; Determine the Node_Buffer according to the Current_ Slot_Size; if( Node_Buffer is not empty){ Move packet to the slot; Log Queuing_Delay; Transmitted_Packets ++;} else Wasted_Slots ++; }} End
Fig. 4 Algorithm of the MAC protocol of the variable size slot scheme
MAC_Protocol_Super_Size_Slot () Begin Global_Time:= time since simulation started Slot_Table:= points to slots in each super size slot for (each node) do { for (each logical ring) do { while (Node_Buffer is not empty AND Slot_Capacity- Slot_Load >= Packet_Size) { if (Super_Slot_Table is empty ) { Move packet to the Current_Super_Slot; Log Queuing_Delay; Slot_Load =+ Packet_Size Transmitted_Packets ++; } else SlotTable ++; (go to the next slot in the super slot)}}} Global_Time= +Super_Slot_Time; End
Fig. 5 Algorithm of the MAC protocol of the super size slot scheme
IV. MIRRORING TECHNIQUE FOR THE SAN NODES A simple technique is used to mirror each primary SAN node to its corresponding secondary SAN node. Under this mirroring technique, the secondary SAN nodes do not send any traffic to the primary SAN nodes and ordinary nodes do not send any traffic to the secondary SAN nodes. However, the secondary SAN nodes ultimately receive all the traffic addressed to the primary SAN nodes as the primary SAN node remove a packet from a slot upon reception only if its corresponding secondary SAN node has already received this packet. Otherwise it will let the packet remain in the ring to go to the secondary SAN node. Therefore those packets passing by the primary SAN node first will travel further in the network to be mirrored in the corresponding secondary SAN node which means extra bandwidth is used. However, on average this proposed mirroring scheme saves bandwidth and introduces efficiency in that separate transmissions are not needed to synchronise the SAN and its mirror. The Two remain synchronised at all time subject to the ring propagation delay. The receiving algorithm under the mirroring technique is shown in Fig. 6. To indicate whither the packet is received by the other corresponding SAN node, a flag, initially set to zero, is attached to each packet destined to one of the SAN nodes. As it receives a packet, each SAN node checks this flag. If it has been changed to 1 (indicating that the packet has been received by the other corresponding SAN) the packet will be removed from the ring. Otherwise it remains in the ring to continue its way to the other corresponding SAN. According to the original MAC protocol, for the single section architecture, nodes in the upper part of the network can use either ring B or ring A to send to the primary SAN node. Although both rings result in the same distance to the primary SAN node, ring A results in a longer distance to the secondary SAN node. This extra distance increases mirroring time and leads to inefficient bandwidth usage. To reduce bandwidth usage and the mirroring time for the upper nodes, a modification can be introduced to the MAC protocol. In this modified version of the protocol, the upper nodes have to use ring B to send to the primary SAN node. To overcome the extra load introduced to ring B, a wavelength can be taken from ring A and assigned to rings B and C. Ring A can accommodate its traffic in a single wavelength as less traffic travels through it (40% to 60% is assumed to be destined to the SAN node in the asymmetric scenario). The mirroring technique with the modified MAC protocol is applied to the two section ring. Fig. 7 illustrates the modified mirroring technique with the modified MAC protocol for both architectures.

1-1-5 1-1-250

Intensity [a.u] Intensity [a.u]

Intensity [a.u]

250 1-1-2 250

1-1-1-1-3 250

300 1-1-1-1-250
Intensity [a.u][a.u] Intensity

100 100

200 200

1-1-6 1-1-250

150 150

1-1-7 1-1-7 250

1-1-8 1-1-8 250

300 1-1-1-1-9

-100 -100

0[ps] 0[ps]

100 time 100 [ps] range

time range [ps]

100 --100
100 time100 [ps] range time range [ps]
100 time range [ps] 100 time range [ps]

150 100

50 200
0[ps] 0[ps] 100 --100 0[ps] 0[ps]
100 time range [ps] 100 time range [ps] 150

250 250

100 time 100 [ps] range time range [ps]

#10 #10

#11 #11
time range [ps] time range [ps]

#12 #12

#13 #13

#14 #14

#15 #15
100 time range [ps] time range [ps]

1-1-1-1-250 250

1-1-12 1-1-12

1-1-13 1-1-13

1-1-15 1-1-250

1-01-16 2501-01-16 250

#16 #16
(a) Waveforms of encoder output (a) Waveforms of encoder output

SG SG @10GHz @10GHz

EDFA EDFA

PC POL PC POL

BPF EDFABPF EDFA 5nm 5nm #1 #1

#2 #2 #16 #16

BPF EDFA BPF 5nm EDFA 5nm #1 #1
BPF @1550nm BPF @1550nm 5nm 5nm

MLLD MLLD

Multiple Multiple Encoder Encoder
Input #1 Input #1 Output #16 Output #16

400 400

Multiple Multiple Decoder Decoder

Input #16 Input #16

(b) Waveforms of decoder output (b) Waveforms of decoder output

300 300

time [ps] time [ps]

200 00

00 5050
time [ps] time [ps] 150 150
100100 time [ps] time [ps]
100100 time time [ps][ps] 150150 200200

0[ps] 100 --100 0[ps]

0[ps] 100 --100 0[ps] 0[ps] 0[ps] 100 --100 0[ps] 0[ps]
0[ps] 0[ps] 100 --100 0[ps] 0[ps] 100 --100 0[ps] 0[ps] 100 --100 0[ps] 0[ps]

200200

Figure 7. Experimental demonstration of multiple optical code processing.
The full E/D processes all the 16 OC labels simultaneously, in parallel: if a codeword is forwarded to one of the device inputs, at the device outputs we measure the correlation signals between the incoming label and all the labels the device is able to generate. The autocorrelation peak (ACP), detected at one of the device outputs unequivocally identifies the incoming label. We measure the autocorrelation waveform at the output port #1 of the decoder (second device), whereas crosscorrelation signals of much lower intensity are detected at all the other output ports (see Figs. 7(b)). The intensities at outputs #3 to #15 are practically zero, whereas at outputs #2 and #16, we can discern a crosscorrelation peak (CCP). The code detection parameter, evaluated as the ratio between the ACP and the maximum CCP, is about 5. These results show this novel planar lightwave device, that is able to generate or recognize a large number of optical codes simultaneously. We have also developed 50-chip, 500Gchip/s multiple OC encoder/decoders with arrayed waveguide configuration [20]. It can generate and recognize simultaneously 50 different optical phase shift keying (PSK) codes with low latency. III. SYSTEM APPLICATION A. All optical Hierarchical Code Processin or OPS Network We have also proposed optical code/code-band routing and photonic UNI for optical label switched networks as an application of all-optical hierarchical code processing. The proof-of-principle experiments of these concepts are shown. Figure 8 (a) represents conventional wavelength path and wave-band routing. As an analogy with it, we have proposed a novel concept of OC-path and OC-band routing (Fig.8 (b)) [13]. OC/OC-band routing can be realized by using hierarchical label processing.

. Application o i e o ain ultiple Optical Code Generation and Reco nition The impact of time domain multiple optical code generation and recognition method to the OPS and OCDM system research is immense. In last 3 years, we have done many demonstrations on OPS and OCDM with this multiple optical code processing technologies. For example, asynchronous OCDM transmission capacity increased from 1 Gbps x 10 user to 10 Gbps x 25 user. In 1 Gbps x 10 user demonstration, 10 conventional encoder devices are used to generate 10 different optical codes signal. On the other hand, we have used one device to generate 25 optical codes in the demonstration of 10 Gbps x 25 user. Addition, in latter demonstration, we have also show the 10 Gbps x 25 OCDM-user x 5 WDM transmission by using one encoding device and another one decoding device only. Total transmission capacity is 1.24Tbps [20]. This multiple optical code processing technology can accelerate the performance of OCDM and OPS systems
IV. CONCLUSIONS All-optical code processing technologies have been explained as examples of recent progress. Ultra-high speed time domain, spectral domain, hybrid domain, and multiple optical code processing deices and systems have been shown. As application of these technologies, OCDMA-PON, OPS network, and ultra high-speed optical clock generation have been also demonstrated. ACKNOWLEDGMENT The author aknowledge the contribution of his collegues and visiting students of NICT to these works. The author would like to extend his thank. Wang of Heriot-Watt University, G. Cincotti of University Roma Tre, K. Kitayama of Osaka University, and K. Kodate of Japan Womens University for their valuable discussion and collaboration in all-optical label processing. REFERENCES
[1]. Wang, N. Wada, T. Miyazaki, G. Cincotti, K. Kitayama: Asynchronous Coherent OCDMA System with Code-Shift-Keying and Balanced Detection, IEEE Journal o Selected opics in Quantu Electronics, vol. 13, no. 5, pp. 1463-1470, September 2007. N. Wada, H. Furukawa, T. Miyazaki: Prototype 160Gbit/s/port optical packet switch based on optical code label processing, IEEE Journal o Selected opics in Quantu Electronics, vol. 13, no. 5, pp. 1551-1559, September 2007. N.Wada, K.Kitayama: 10Gbit/s, 8-chip coherent optical code division multiplexing using all-optical dncoder/decoder, he Paci ic Ri Con erence on Lasers and Electro Optics (CLEO/Pci icRi '97), PD2-14, July 1997. N.Wada, K.Kitayama: 10Gbit/s optical code division multiplexing using 8-chip optical bipolar code and coherent detection, IEEE/OSA J. Li htwave echnol., vol. 17, no. 10, pp. 1758-1765, October 1999. N.Wada, H.Sotobayashi, K.Kitayama: 2.5 Gbit/s timespread/wavelength-hop optical code division multiplexing using fibre Bragg grating with supercontimuum light source, IEE Electron. Lett., vol.36, no.9, pp.815-817, Apr. 2000. N. Wada, H.Harai, W. Chujo, F. Kubota: Photonic packet routing based on multi-wavelength label switch using fiber bragg gratings, 26th European Con erence on Optical Co unication (ECOC2000), vol. 4 no.10.4.6, pp.71-72, September 2001. A. M. Weiner, J. P. Heritage, J. A. Salehi: Encoding and decoding of femto-second pulses, IEE Opt. Lett., vol. 13, pp. 300-302, 1988. J. Cao, R. G. Broeke, N. K. Fontaine, C. Ji, Y. Du, N. Chubun, K. Aihara, A. V. Pham, F. Olsson, S. Lourdudoss, S. J. B. Yoo: Demonstration of spectral phase O-CDMA encoding and decoding in monolithically integrated arrayed-waveguide-gratingbased encoder, IEEE Photon. echnol. Lett., vol. 18, no. 24, pp. 26022604, Dec. 2006. A. Agarwal, P. Toliver, R. Menendez, S. Etemad, J. Jackel, J. Young, T. Banwell, B. E. Little, S. T. Chu, C. Wei, C. Wenlu, J. Hryniewicz, F. Johnson, D. Gill, O. King, R. Davidson, K. Donovan, and P. J. Delfyett: Fully programmable ring-resonator-based integrated photonic [10]

Feature

JAVOBS
OBS Protocols Scheduling Algorithms OBS Architectures Model Building Routing Traffic characterization Programming Language
JET, Horizon LAUC, LAUC-VF FFUC, FFUC-VF C-OBS, E-OBS
JET, JIT, Horizon, E-JIT, JIT+ LAUC, LAUC-VF FFUC, FFUC-VF C-OBS, E-OBS
Graphically edited, Predefined input file input file, or created dynamically Specified Fixed (Poisson) C++ Configurable Configurable Java
Figure 2. JAVOBS / ADOBS Analytical Validation.
Figure 3. JAVOBS / ADOBS Runtime Test.
Figure 4. Burst Loss Probability vs. Load in C-OBS.
Figure 5. Burst Loss Probability vs. Load in E-OBS.
A test measuring the running time of both simulators has also been performed. Simulations were run according to the number of bursts generated and prompted more than one hundred hours of simulation (on an Intel Core 2 Quad 2.4 GHz desktop computer). In this case, we consider two network topologies: (1) NSFNET [27] (US network); (2) EON [28] (a pan-European network defined in European COST 266 action) with 15 and 28 nodes, and 23 and 39 links respectively. JET signaling and LAUCVF scheduling are used. The results obtained are shown in Figure 3. ADOBS performs better at low values of generated bursts, probably taking advantage of the C++ performance. However, tendency changes at about 1 million bursts. From this point on, the ADOBS curves exhibit an exponential increase which finally creates gaps of up to 96 hours between both simulators. This gap is apparently due to unoptimized memory utilization in the ADOBS simulator which obliges the operating system to use the hard disk as RAM extension, and thus, drastically reduces the simulator throughput. Although JAVOBS is outperformed in short simulations, we observe a constant growth of the running times for all time scales which exhibits its robustness. Thus, in this case, the benefits of
using a garbage-collected language, which dispenses the user to take care of many memory management related operations, become apparent. B. Evaluation of the E-OBS and C-OBS Architectures. Considering that fiber delay line (FDL) buffers are not used, it has been proved in [29] that the best worst-case performance of an online best-effort scheduling algorithm is achieved when all bursts have the same offset time and the same length. One of the benefits of E-OBS comes from the fact that offset times are introduced at each core node by means of additional fiber delay coils inserted in the data path at the input port of the node. As a result, E-OBS does not experience offset variation inside the network. In such scenario, scheduling algorithms do not need to implement any void filling technique. Therefore, in an E-OBS network, JIT and Horizon reservation mechanisms seem to be the most appropriate ones due to its low complexity compared to JET. Indeed, the overprovisioning of resources that characterizes JIT is substantially reduced using E-OBS due to smaller offset times. Figures 4 and 5 present the results obtained in both control architectures under the different signaling protocols supported by JAVOBS.