During its “Get the Facts” campaign Microsoft produced a case study documenting the LSE’s adoption of a C#, SQL Server 2000 application to facilitate real-time stock information prices. About 40% of the LSE’s revenue came from sale of such information and consequently the application had the potential to deliver real value, provided it was up to the task.
Previously, the LSE disseminated data through the London Market Information Link (LMIL), a COBOL application running atop a Hewlett Packard Tandem environment.
This architecture was noted as being extremely reliable and with a full 100% availability record. Despite this, it was not known for its agility. Any development or scaling took time to design and implement.
The LSE thus sought a new system which would retain reliability but with greater capacity and be cheaper to manage.
Several proposals were reviewed with Microsoft and Accenture having the winning tender in the form of Infolect, designed to handle more than 15 million real-time messages per day through a network of 120 HP ProLiant servers across multiple data centres. These messages were distributed to more than 107,000 trading screens in over 100 countries.
Although the volume of messages is massive, Infolect is, according to Microsoft, “100% reliable on high-volume trading days.” At a cost of 40 million pounds or $USD 65 million you would certainly hope so.
The LSE Chief Technology Officer (CTO) at the time said the decision to implement Infolect on a Microsoft architecture “was a high-visibility to customers of the importance placed on reliability and support.”
Infolect went live on September 19, 2005 with big boots to fill. In the preceding six years there had been no production outages at the LSE. It was critical the new Microsoft-based platform prove to be just as reliable.
Although Infolect was not replacing a Linux system Microsoft still touted this case study as an example of Windows Server being chosen over Linux for reliability reason. Microsoft used this claim as the headline for the first issue of “The highly reliable times.”
Of course, nothing in the case study makes any reference whatsoever to Linux, unless it is assumed that one or more of the potential architectures reviewed was Linux-based. (It’s not uncommon, it seems, for the “Get the facts” case studies to have sensational titles which are not supported by the story.)
Nevertheless, fast forward to today and the London Stock Exchange is writing off its huge $USD 65m development, looking instead to Linux. What happened?