Virtualization Saves Natixis Money, Space—and Downtime
May 18, 2009
Virtualization projects tend to have effects that extend far beyond cost savings. Take, for example, Natixis Capital Markets, whose efforts have enabled it to perform maintenance on any of its trading operation's 300 virtual servers in the middle of the day.
When securities trading systems have downtime, it's a source of friction among a firm's executives, traders and the technologists that support them. "If they're down, we don't want to hear the screaming," says Fazil Habibulla, VP and systems engineer at Natixis Capital Markets, the North American investment banking unit of Paris-based Natixis.
Maintenance used to take place at night or on weekends, when traders were off the floor. But now the bank can conduct business while upgrading or repairing systems--something it couldn't do without virtual servers.
"Virtualization is a modernization catalyst," contends Thomas Bittman, a Gartner VP and analyst. Even if an initiative's original driver is savings, extending the life of existing hardware by using software to create virtual servers and desktops can lead to fundamental changes in architectures, processes--and even maintenance.
The Natixis project began in 2004, when the firm's 5,000 square-foot data center in midtown Manhattan was running out of space and power. "It was cramped to the max," says Habibulla.
To continue expanding while pulling out stacks of hardware, the bank began replacing some of its 240 physical servers with virtual copies of machines, operating systems and applications that run on interchangeable racks of IBM Corp. blade servers. Today, only 70 servers are dedicated to specific applications and operating systems.
Even with all the virtual machines dedicated to trading, there is enough capacity left over to shut down, start up, upgrade or move virtual servers and their applications, without interrupting service. Habibulla says he leaves about 30 percent of the available capacity free, allowing him to set up replacement machines on different servers and switch over instantly.
When the project started, Natixis was building a facility in New Jersey for redundancy, which it expected would replace the New York center. But the virtualization of its infrastructure extended the old facility's life span.
Seamless Switchover
Natixis, which has 900 staffers, looked at virtualization products from Citrix Systems and Microsoft Corp. before opting for VMware. The biggest reason? VMotion, a feature that lets companies copy a virtual machine from one host server to another and then switch to the reproduction without interruption. That allows clients to repair a fan, hard drive or other components of a physical machine without any effect on operations, notes VMware senior director Jerry Chen.
Though that was radical at the time, other vendors have since added the capability, says Habibulla. And VMware itself now moves virtual machines behind the scenes automatically, to balance loads on physical servers, according to Chen.
Typically, servers dedicated to particular tasks last about three years, says Habibulla. With virtualization, Natixis has been using the same hardware for five years. For the trading servers alone, the savings amount to $2.1 million, or $7,000 for each replacement machine, he estimates. The company has also saved $200,000 or $300,000 a year in operating costs, adds Habibulla--all with a two-person virtualization team.
In 2006, the firm began reformatting its network host computers so they could appear on any server at any physical location, essentially virtualizing Natixis' internal communications network. Using supervisory software from Cisco Systems, which also supplies its primary networking gear, Natixis can move hosts or reassign communications channels between devices on the fly, says Henry Yin, deputy manager of the core infrastructure group. Slated for completion by year-end, the network virtualization will increase the availability of resources and make it easier to manage them, says Yin.
