Original URL: http://www.theregister.co.uk/2009/08/05/liquid_computing_update/

Can Liquid Computing ride Cisco's California coattails?

Or will it get et?

By Timothy Prickett Morgan

Posted in Servers, 5th August 2009 23:35 GMT

Taking on a megacorp with over 65,000 employees and $8.5bn in annual sales may seem a daunting challenge, but the focused folks at unified-computing shop Liquid Computing have Cisco in their sights.

Virtualized servers with unified server and storage management, plus scalable and reconfigurable infrastructure, all controlled through out-of-band system management tools, were supposed to be all the rage by now. And many of them were supposed to have a Liquid Computing brand slapped on them.

Others were supposed to have the Fabric7 brand on them, but that company went the way of all flesh back in April 2007, after a year and a half of being in the market - and it couldn't even blame the economic meltdown for its demise.

But Liquid Computing is still in the game, even though it has been forced to rejigger its technology and make some other changes to pursue business - just in time for Cisco Systems to start talking up its "California" Unified Computing System, which embodies some of the same principles that Liquid Computing has been trying to peddle with its LiquidIQ iron for the past three and a half years.

"Finally, there are some pretty big companies out there that are dismantling silos," says Vikram Desai with a laugh. Desai was tapped to be president and CEO of Liquid Computing in January 2009 following a revamp of the LiquidIQ product line. "We released our eleventh software release before Cisco even got its first one out the door," he told The Reg

Liquid Computing made a big splash when it entered the server arena in November 2005, at essentially the same time that Fabric7 launched its Opteron-based NUMA servers.

Liquid Computing had a much more sophisticated design than Fabric7's - which is what you would expect from a bunch of telecom engineers from Nortel who also built supercomputers for the U.S. government's Defense Advance Research Project Agency. Their team's experience may explain why Liquid Computing is still alive and kicking - mostly in the direction of Cisco these days.

That and $45m in two rounds of venture funding from VenGrowth Capital Partners, ATA Ventures, and Newbury Ventures. A third round is expected sometime this year to keep the company rolling.

Liquid Computing started shipping its initial LiquidIQ servers in October 2006, which spanned up to 17 chassis of blade servers for a total of 960 single-core Opteron processors that could be configured, using the proprietary IQInterconnect, into any number of virtual SMP servers.

With the LiquidIQ 2.0 release announced last October, the IQInterconnect secret sauce which allowed flexible virtual SMP setups was replaced with plain old Gigabit Ethernet and 10 Gigabit Ethernet ports. The IQInterconnect was very slick in that it offered 100 GB/sec (that's bytes) of bandwidth between compute nodes in the system, which is how it could create an SMP of many boards. The Ethernet backbone that replaced it can only handle 84 Gb/sec (that's bits) of bandwidth.

The basic blade architecture is the same for the 2.0 machines, with a chassis supporting 10 blades in the front and 10 blades in the back, although the box now supports two-socket as well as four-socket Opteron blades.

It has taken a few months to reorient Liquid Computing, but Desai says the company is pressing ahead. In March, it announced the LiquidIQ 3.0 platform, which will feature Intel's Nehalem EP Xeon 5500 processors and which Liquid Computing has dubbed a "unified computing system."

While this machine is in preview and is not expected to start shipping until the end of this year or early next, the Opteron version of the LiquidIQ 3.0 system supports the latest six-core AMD Istanbul Opterons as well as the vSphere management stack and the ESX Server 4.0 hypervisor from VMware.

As far as operating systems go, the LiquidIQ box can support Microsoft Windows Server 2003 or 2008, Red Hat Enterprise Linux 4 and 5, Novell SUSE Linux 10, and Sun Microsystems Solaris 10. Microsoft's Hyper-V hypervisor is supported, as well, as are older ESX Server 3.0 and 3.5 releases from VMware. Oracle's clone of RHEL and its clone of Xen are also supported. And like Cisco with its California blade boxes, LiquidIQ has a tight partnership with NetApp for iSCSI storage.

With the virtual SMP capability lost, Desai says that Liquid Computing can still compete because of the out-of-band management capabilities of the LiquidIQ platform. "We make all of this software and hardware complete," says Desai. "We allow the virtual environment to talk to and change the underlying hardware."

Still, Liquid Computing faces a pretty big uphill battle, with Cisco making all its noise about California blades and their converged networks and storage and integrated management and virtualization, and with Hewlett-Packard making a lot less noise (but probably more sales) with its BladeSystem Matrix, a counterpunch to California that's based on HP blades, storage, and integrated management tools. IBM can field similar boxes and marketing, but Sun is so confused right now that it's hard to say what it will do until Oracle tells it the plan.

For now, Liquid Computing is doing about as well as Cisco with this unified computing thing, with "under a dozen customers in production," according to Desai, who adds that "they are all paying customers" and that "nobody is getting free iron."

One of those customers, Seattle, Washington-based travel network operator Virtuoso, which supports luxury travel accommodations for 300 agencies in 22 countries, has standardized on the LiquidIQ machines to support both bare-metal and virtualized Windows and Linux instances and has seen a 10 to 1 reduction in its operational costs since ditching standard x64 boxes.

Despite the tough economy, Virtuoso says it plans to have a ten-fold increase in revenue in the next five years, but it still needs to radically cut costs of operating its systems. Maybe the rich really do live differently from the rest of us, but the data centers supporting their travel whims apparently have to do more for less.

It is hard to imagine that Liquid Computing will be able to compete toe-to-toe with Cisco in terms of research and development, marketing budget, and so forth. Perhaps the best thing that could happen to LiquidIQ is that a company sorely in need to get out of its box mentality and short on management and virtualization tools - say, for example, Dell - swoops in and picks up LiquidIQ and takes the technology mainstream.

It would not be surprising to see HP or IBM snap up Liquid Computing, either, and Oracle might do it by accident as it tries to eat the entire IT industry. Then again, if Liquid Computing can stay in the game and fight all by itself for customers, that's fine, too. Competition is a good and necessary thing. ®