VMware and ISVs Including Marathon Square Off on Fault Tolerance

CIO — Among the consequences of VMware's battery of vSphere 4 announcements Tuesday about its new virtualization infrastructure and add-on components, third-party software vendors specializing in fault-tolerance and high-availability are feeling new pressure now that VMware has added basic versions of those functions to its core product.

Marathon Technologies, for example, rolled out the newest version of its well-respected fault-tolerance applications one day before VMware's product-announcement blowout, pressing Marathon's point of view that the fault-tolerance features VMware has long said it would add to VMware won't cut it for demanding customers.

Keep in mind, Marathon's technology works with Citrix VMs, but not with virtual servers created using VMware technology. In January, 2009, Marathon announced a partnership with Microsoft to expand its capabilities to virtual infrastructures running on Windows Server 2008 with Hyper-V.

The fault tolerance capability VMware is building into its vSphere infrastructure is fine for basic failover protection, but doesn't have nearly the kind of flexibility and configurability that Marathon's new everRun 2G offers, says Marathon CEO Gary Phillips.

Marathon's new fault-tolerance application for virtual servers, everRun VM Lockstep, allows the company to give customers the ability to set the level of protection for each specific application within their network of virtual servers, Phillips says.

"If you're running Microsoft Sharepoint and Exchange on the same VM, that may not rise to the point that you need HA functions or FT. But if you run several instances of each one on that same physical server, so all your e-mail and documents are going through the same physical machine, they may collectively make that server mission-critical enough that you want to protect them," Phillips says.

everRun VM Lockstep is designed to monitor virtual servers and shunt applications or workloads from a server that's showing signs of a crash to one that is stable.

An additional layer of protection is available for the physical server from everRun 2G, a combination of Marathon's existing everRun HA—which monitors each component in a pair of servers and redirects traffic to the other server when a component of one fails, and everRun FT—which runs the same applications on two servers simultaneously so one can take over if the other fails.

Customers can choose whether they want high availability for a single VM, all the VMs on a server, the server itself, or full fault tolerance for the full physical server so that not a second's worth of data or transactions are lost in the case of either hardware or virtual-server problems, according to Mark Bowker, analyst for the Enterprise Strategy Group.

"Some people we've talked to are switching over to virtualization just to get this kind of failover capability in software," Bowker says.

HA and fault-tolerance aren't specialties a vendor can learn or build overnight, Bowker says.

"VMware announced a lot about high availability, mainly automatic restart of a VM on another machine," Bowker says. "That's very difficult in the physical world. It requires a lot of hardware and special configurations. By teaming up with the guys at Citrix, Marathon is able to take applications that are being virtualized and provide high availability without as much effort on the hardware side."

That's true to a certain extent, but Marathon's drawbacks and partnerships could hurt its chances with new customers as quickly as if its capabilities were not more proven than VMware's, says Gordon Haff, analyst at Illuminata.

"There are cases in fault tolerance where there is some small issue over here and another over there that add up to a bigger problem, and Marathon may well be able to handle that better than VMware can," Haff says. "The fact is that VMware is going to be in a lot of companies anyway, and now it has a fault-tolerance solution. The question is how many companies are going to go looking for the gold standard of fault tolerance from companies like Integrity Nonstop or Tandem, and for how many is VMware HA going to be good enough?"

Cost may be one issue. Marathon charges $9,000 per physical server for everRun Lockstep and for everRun 2G delivering high-availability levels of service. Full fault tolerance costs another $9,000.

VMware charges and additional $2,245 per processor for the Advanced version of its vSphere infrastructure, which includes HA capabilities along with VMware VMotion server migration, VMware Data Recovery for backup and VMware vShield Zone for security.

Marathon's decision to base everRun on Citrix Xen Servers may also be a factor, Haff says. The virtual-server market has come down to a competition between Microsoft and VMware, he says. Marathon will be more attractive to companies who don't already use VMware, which means Hyper-V users primarily, but being available only on Citrix limits Marathon's appeal, especially to companies that don't want to have to deal with more than one vendor's software in their virtualized environment, Haff says.

"It's going to get a lot harder to play in the virtualization market as, essentially, a point-product ISV," Haff says.

Follow everything from CIO.com on Twitter @CIOonline