To say that network connectivity is mission-critical for Thomson Reuters would be an understatement.
Thomson Reuters calls itself “The Answer Company.” Its business is expeditiously delivering information over direct connections to 40,000 customers across numerous industries worldwide. Financial services is by far the largest of these verticals, representing more than half of the company’s business.
“Delivering products reliably and making sure [customers] have the best possible services” are the most important requirements for the Thomson Reuters WAN, explained Dimitrios Voutsinas, head of network design and development for Thomson Reuters, at WAN Summit London 2017.
After an evaluation process, Thomson Reuters determined that moving to an Ethernet WAN would be the best way to address those requirements. In his session, Voutsinas discussed the process that the company went through in making that decision.
WAN Landscape Analysis
Voutsinas offered some data points to illustrate the magnitude of Thomson Reuters’ business:
- The company has 400,000 end users in 150 countries
- It supports $250 billion in bond trading and $200 billion in foreign exchange trading daily
- Its data centers hold 60,000 terabytes of data
- The company’s systems distribute up to 5 million price updates per second to the financial markets
Around 2010, Thomson Reuters undertook what Voutsinas called a “WAN landscape analysis,” which revealed several important concerns.
The company’s network and server infrastructure was “very old,” according to Voutsinas. It was expensive and, in some cases, impossible to scale. End-to-end latency no longer met the increasing demands of the financial markets. And in some markets, the company was not fully recovering the cost of its communications networks that it had employed to be able to assure performance for its global customers.
As Voutsinas’ team thought about their future requirements, “at the top of the list was how to achieve an order of magnitude in throughput,” Voutsinas said.
In exploring how the Thomson Reuters network could meet these future requirements, Voutsinas said, “we realized very quickly that we had to exploit new technology and the internet; which at the time was becoming a viable platform to distribute out products.”
Voutsinas and his team also determined that the network would be much improved if it were re-architected. The architecture in place at that time was very centralized. “We brought all the exchange feeds and data into super data centers,” he explained. “From there, we processed [the data, did our] proprietary value add [and sent the data to] secondary data centers from which it was distributed.”
After rethinking that approach, the network design team decided that “we wanted to move to a distributed architecture” so that customers could “get as close as possible to the source of the content” – a move that would improve latency. The plan was to establish 34 new points of presence and to use multicasting to distribute content.
After rethinking that approach, the network design team decided that “we wanted to move to a distributed architecture” so that customers could “get as close as possible to the source” – a move that would improve latency.
Implementing this plan, however, would involve breaking new ground.
“We quickly realized that there wasn’t anyone who was doing what we wanted to do on a global scale,” Voutsinas recalled.
Implementing the Plan
In preparing to implement their plan, IT personnel at Thomson Reuters mapped all their customers, many of whom were in financial hub cities such as Chicago, Frankfurt, London, New York, Hong Kong, Tokyo, and others. The good news, according to Voutsinas, was that most were already clustered within 20 miles of a Thomson Reuters exchange or point of presence.
In exploring WAN connectivity options for customers, Voutsinas and his team relied heavily on data provided by TeleGeography about bandwidth rates and pricing options on a global scale. The analysis revealed that the company could save substantially by shifting to Ethernet. The move to Ethernet also would enable the company to put customer premises equipment on the same network as the Thomson Reuters distributed POP.
A key concern about Ethernet was achieving consistency network-wide. But standards developed by the Metro Ethernet Forum helped address that concern.
Ultimately, Thomson Reuters established two network POPs per city. In planning those POPs, the company took care to establish commercial and physical diversity.
IT personnel also migrated 9,000 customers to a new provider. Many customers were previously connected via a single global provider but now have multiple connections from best-in-region service providers. One of the final customers to be migrated had its migration completed in September, Voutsinas told WAN Summit attendees.
As a result of the move to Ethernet, Thomson Reuters has been able to “delight” its customers, Voutsinas said, by offering more bandwidth at lower cost, reducing latency, and improving reliability, resiliency, scalability, and time to market.
As a result of the move to Ethernet, Thomson Reuters has been able to “delight” its customers, Voutsinas said, by offering more bandwidth at lower cost, reducing latency, and improving reliability, resiliency, scalability, and time to market. At the same time, Voutsinas and his team saved Thomson Reuters hundreds of millions of dollars per year.
Lessons Learned
Voutsinas shared several lessons learned from the Thomson Reuters network upgrade. Three of the most important were:
1. Determine what you want to achieve and then try to map what’s available to meet those requirements. Network architects might want to find out what is available first, but that should come second. Recognize that you may need to influence standards bodies, service providers, and vendors to shape your future WAN.
2. Only you understand your business’ unique differentiation points and what the future of your WAN should look like. Following vendor suggestions or trends may not always help you.
3. Always keep the needs of your customers, both internal and external, in mind. Don’t lose sight of their requirements as you plan for WAN reliability, scalability, and cost effectiveness.
Brianna Boudreau
Senior Research Manager Brianna Boudreau joined TeleGeography in 2008. She specializes in pricing and market analysis for wholesale and enterprise network services with a regional focus on Asia and Oceania. While at TeleGeography, Brianna has helped develop and launch several new lines of research, including our Cloud and WAN Research Service.