Wombat Addresses Data Deluge With RDMA Technology
January 7, 2008
Market-feed distribution specialist Wombat Financial Software has introduced a product that employs a shared memory software architecture that the company calls an innovative way to address the flood of market data that active trading firms--particularly frequent users of algorithmic trading strategies--must receive and process on a daily basis.
New York-based Wombat is touting the product, Wombat Data Fabric, as a cost-effective alternative to new hardware-based solutions that employ field-programmable gate array technology to improve messaging speeds. "Bottlenecks are in the IO [input-output] more than anything and a lot people have been reaching out for a hardware solution in desperation," says Ken Barnes, co-principal of Wombat's middleware division.
"We are the first firm to apply RDMA [remote direct memory access] technology for data management purposes in the financial services market," claims Barnes, "and the beauty of it is that it's a strictly" software-based solution.
According to Barnes, such an approach has been taken by medical firms that use three-dimensional imaging technology and government agencies, but it is now finding its way into the financial services sector.
RDMA allows data to move directly from the memory of one computer to another without involving the operating system, allowing for high-throughput, low-latency networking. When a network moves to Wombat Data Fabric, says Barnes, it frees up computers' central processing units (CPUs) to work on other demanding tasks.
Barnes notes that the software was developed over an 18-month period with Billerica, Mass.-based Voltaire, an InfiniBand switch provider. "We did a white-board session with ... Voltaire about two years ago at one of our customer conferences," explains Barnes. "We started to talk about the benefits and features of the InfiniBand protocol--and the properties that middleware software working in such an environment would require--and had an "a-ha" moment about how RDMA technology could revolutionize the nature of middleware for market data management."
"Most middleware uses 50 percent of the CPU," says Barnes. "When we clean that up we see a four-times message-per-second improvement." He adds: "We are getting latencies of 50 microseconds or lower where we publish close to 2 million messages per second across a single connection."
Brad Bailey, senior analyst at Boston-based Aite Group, says that market data management is a complicated, highly competitive space with many different solutions being offered to address escalating messaging volumes, including hardware approaches from companies such as Exegy and Tervela.
"Many firms are scrambling to figure out what is the best approach, given the legacy systems they have in place and the overall goals they have," says Bailey, who adds that Wombat's off-the-shelf product can be used with familiar server systems and could make sense for firms seeking to keep down data management costs. According to Barnes, Wombat Data Fabric can slash data center usage by 50 percent to 75 percent, depending on a firm's server use, growth rate and buy-back agreements.
"The algorithmic trading community has made it clear--escalating volumes demand a new approach to radically lower the data center investments current technologies impose," says Wombat COO Danny Moore. "The challenge has been how to marry that innovation with the low risk and cost of commodity hardware and proven software solutions."
Moore continues: "We've integrated our Wombat Data Fabric middleware under the hood of the same Wombat APIs [application programming interfaces] in use today to deliver breakthrough performance with battle-tested software on the standard, affordable servers customers know and trust."
