Country missing? Please select your nearest region...
Bloomberg CIO Vipul Nagrath gives us an inside look at the company’s data center strategy
2 May 2012 by Yevgeniy Sverdlik - DatacenterDynamics
Bloomberg wants to be the most influential news organization in the world. That’s what Bloomberg Media Group's CEO Andrew Lack told the New York Times in 2009. And judging by its commercial success and the size of its news team, in comparison to the rest of the general-news industry (Vanity Fair estimated in 2008 that it was bigger than the editorial muscle of the New York Times and The Washington Post combined), the company may well be heading towards its goal.
Bloomberg’s professional service package includes real-time market data, news feeds and alerts, financial-information analytics, news-traffic analysis, order-management systems and more.
This is all served around the clock by Bloomberg’s global data center infrastructure. The man in charge of it is Vipul Nagrath, Bloomberg CIO and head of the company’s research and development team.
Bloomberg data centers support a cloud-like infrastructure, which Nagrath says was implemented before the expression ‘cloud computing’ entered the common vocabulary.
Bloomberg sold the first batch of 20 data terminals it is famous for to Merrill Lynch about 30 years ago. The terminal set-up could be described as a cloud: many user end points are served data from a centralized processing system.
Today, the set-up is still a client-server-based architecture, although Bloomberg no longer sells the physical terminal. That is as much detail as Nagrath will share about the firm’s secret IT sauce. At the end of the day, Bloomberg is as much a technology company as it is a media company and its technology strategy is its lifeblood as much as its newsrooms are.
Mostly owned, some leased
Bloomberg owns and operates a lot of its own data centers but also takes retail colocation space all over the world, Nagrath says. Primary server work takes place on the US East Coast, on thousands of servers. Bloomberg owns and operates most of that capacity.
While many companies in the financial vertical still build and operate most of their data center capacity, there is a general trend in this industry of moving toward a more outsourced model.
Jason Schafer, research manager at Tier1, says many financial-services players want to focus on their core business “and leave the data center management to a company specializing in that. If you look at the growth of financial services business for Digital Realty [Trust] and Equinix, it becomes pretty obvious.”
Reliability above all
Financial-services companies will often say reducing latency is their top priority but latency is typically second to reliability. This is true for Bloomberg. “Reliability is always the number-one agenda item for me,” Nagrath says. The company has customers around the world and its infrastructure must be up around the clock.
Bob Landstrom, senior manager of data center services at HP-owned Mphasis Infrastructure Services, says the latency-versus-reliability balance is different depending on what segment of the financial-services industry you are in.
There are businesses within the vertical that are not directly involved in execution and while high on the list, latency is not their top concern. “For firms like Bloomberg, availability of data and delivery of content is the bread and butter,” Landstrom says. “For that reason, it would make sense that reliability is a fundamental critical-to-quality factor rather than latency.”
CADE, the underdog metric
Bloomberg goes to great lengths to optimize energy efficiency of its data centers, according to Nagrath. From applying waterside economization at the chiller level to evaluating energy efficiency of software applications, his team looks at saving energy from as many angles as it can find. One of these angles is using the controversial CADE (Corporate Average Datacenter Efficiency) metric for data center efficiency. CADE was developed by the management consultancy McKinsey and Company together with the Uptime Institute (now owned by the 451 Group). Consisting of a set of four separate metrics, it is intended to measure efficiency of both IT and
Bloomberg is one of few companies that have put CADE to use. Schafer says he has not seen a lot of companies run with CADE since it was first introduced in 2008. “It’s an interesting and novel idea on paper but, practically speaking, it just hasn’t seen much adoption.”
Taking into account both sides of the house is CADE’s main appeal, Landstrom says. It has been criticized, however, for not taking into account efficiencies gained by removing “zombie servers”. The calculations also get murky when some of an organization’s equipment runs in a colocation company’s facility or in a hosting-provider’s data center, Landstrom explains.
As Bloomberg continues its quest for world domination, its infrastructure demands continue to grow. The company keeps adding services and its user base gets bigger and bigger.
As of March, its full service suite had more than 310,000 clients. And the infrastructure has to keep up, scaling quickly, seamlessly and in an energy efficient way.
Corrected: a previous version incorrectly said Andrew Lack was Bloomberg's CEO. Lack is actually CEO of the Bloomberg Media Group. DatacenterDynamics regrets the error.