Java’s popularity – according to IDC, more than $11 billion, or 20% of worldwide server spend, was devoted to Java servers in 2005 and analysts estimate that number is growing 15-20% annually – is sweeping the Street, at least according to anecdotal evidence. “There are four areas in which we’ve succeeded in the last 18 months: derivatives trading, risk analysis, hedge funds and foreign exchange,” reports Ram Appalaraju, vice president of marketing at Azul Systems, maker of a Java acceleration appliance used by many large Wall Street firms. “The reason they prefer Java is the cost of development in Java is a lot cheaper than any other programming environment.”

The Achilles' heel of Java applications is that they can be subject to performance issues such as garbage collection (a periodic cleaning out of memory that can take several seconds to complete). Programs that deal with large data sets -- such as trading and risk analysis applications -- are particularly prone to Java performance bottlenecks. “The applications pause, they need to reboot the machine, all those things cut down on either the volume of trades, the closing or the risk analysis that cannot be computed in real-time,” Appalaraju says. “So they end up providing data to the portfolio managers that’s two hours if not one week old.”

Azul’s Vega appliance, version 3 of which is being announced today, is a high-performance server that takes on the role of a shared Java Virtual Machine. “The entire Java application gets pushed out to the Azul appliance,” Appalaraju says. “We have multicore technology [864 processing cores per appliance] and lots of memory [three-quarters of a terabtye] in the box, that takes away any barriers to the application to scale.”

According to Appalaraju, the company has documented customer before-and-after benchmarks that show performance improvements of five times up to twenty times using the Vega. A side benefit is server consolidation – a firm might replace ten servers with two Azul appliances, for instance. The Vega is often deployed to complement grid solutions. “Grid is good up to a point,” Appalaraju says. “If you have too many nodes, it becomes unwieldy and overly complex.”

Azul appears to be part of the growing Wall Street data center trend toward greater virtualization and commoditization of hardware. Where once upon a time, a server would be dedicated to an application, today firms are creating shared pools of hardware that are deployed to whatever applications or workloads need them.

Appalaraju also says the Vega is being used to support Java-based complex event processing. When numerous Java applets need to communicate with each other, latency occurs. Appalaraju says hosting all the components on the appliance reduces latency across events by almost 100 times, dropping latency to 40 microseconds or less, he says.

What’s new about the Vega 3 is its Azul-designed multicore chip has 54 cores where the previous generation had 48 cores. Each appliance contains 16 chips, which adds up to 864 processing cores with 768 gigabytes of memory. A new DirectPath feature improves communication between virtual machines. The new appliance also has a real-time performance monitor that lets customers watch and react to performance needs in real time. Pricing starts at $100,000.

Appalaraju has observed a trend toward more cautious IT spending on Wall Street. “It’s no longer the way it used to be two years ago,” he says. “Capital markets tend to buy the best toys, but now we’re seeing a higher degree of value-based assessment.” He says Azul passes the test because its product can be deployed within two to three months.