Marathon beats VMware to the SMP fault-tolerance punch

One of the basic concepts of server virtualization is the ability to abstract servers from the underlying hardware, thus preventing vendor lock-in and giving consumers the ability to mix and match commodity hardware in their environments. Marathon Technologies has taken that a step further by providing consumers with a software-based fault-tolerant solution that runs on commodity servers, not specialized equipment, thereby also preventing hardware vendor lock-in.

Marathon’s everRun MX is designed to help small to midsize organizations affordably ensure uptime for applications running on industry-standard Windows servers. With this latest version Marathon is able to take advantage of today’s modern server hardware, which has multicore CPUs and is usually configured for symmetric multiprocessing (SMP).

Customer demand for always-on environments has outpaced the ability of companies to protect themselves against a system fault, according to Marathon. As a result, high availability (HA) is no longer “good enough.” Customers want more than the ability to quickly recover from a system failure. They want fault tolerance.

everRun MX provides fault tolerance to applications that require 100 percent uptime. It does so by combining two standard Windows servers into a single operating environment with complete redundancy of all underlying hardware and data. everRun MX then presents these redundant servers as a single operating environment to keep applications running in the event of component or system failures. It operates in lockstep, ensuring redundancy of the hardware, data, and networks, to provide automated fault management. Marathon says this ensures true continuous availability for applications that simply cannot have any downtime or lose “in-flight” transactions.

Marathon claims everRun MX is the industry’s first software-based fault-tolerant solution for symmetric multiprocessing and multicore servers and applications. The company also says everRun MX removes the price/performance barrier to continuous uptime, which not only eliminates technology trade-offs but also makes fault tolerance more obtainable by any organization, regardless of its size.

Software fault tolerance itself is not new in the virtualization community. Virtualization giant VMware also provides a software FT solution for its own platform. But both VMware and Marathon have had to settle on protecting only those virtual machines that were configured with a single virtual CPU. Until now, keeping two virtual CPUs in lockstep has proven to be quite the challenge.

Dave Bartoletti, a senior analyst with the Taneja Group, believes Marathon is out ahead of VMware fault tolerance with its everRun MX software. “[Fault-tolerance] has always been a niche market, because you either needed very expensive fault-tolerant hardware or complex fault-tolerant software that wasn’t actually much cheaper. Marathon is cheaper than hardware and legacy software fault-tolerance solutions, and they support SMP and multicore servers and, critically, the apps that require them,” Bartoletti added.

By only supporting fault tolerance on a single virtual CPU VM, virtualization expansion in the production environment has halted or at least slowed down for many organizations. One of the reasons many IT shops have hit the VM Stall mark of around 30 percent is because they haven’t been able to provide that high degree of fault tolerance to mission-critical applications. While virtualization platforms have supported multicore and SMP virtual machines for some time, they have not been able to provide SMP fault tolerance and therefore could not guarantee uptime for these business-critical applications.

Based on many conversations with his own clients, Bartoletti agrees. “Now that the easy stuff is mostly virtualized, the hard stuff isn’t limited by consolidation requirements but by availability. Those mission-critical analytics servers won’t get virtualized without a high degree of fault tolerance, and more and more they are leveraging multi-core servers,” he said.

Bartoletti also believes Marathon’s solution can help change what he calls the “FT 90/10 rule,” where 90 percent of a fault-tolerance budget is spent protecting only 10 percent of the workloads. If fault tolerance is made more affordable and available for multicore servers, he believes they can help push that number to 80/20 or perhaps even to 70/30.

According to Marathon, everRun MX only works with Intel systems today. However, the company is working on a similar product that will work for AMD processors. everRun MX can operate with any 32- or 64-bit version of Microsoft Windows and any application that runs on them, but the latest release is limited to the Citrix XenServer virtual appliance. The company is working on Linux and VMware compatibility down the road.

A complete everRun MX configuration starts at under $10,000 with licenses for two systems that can be paired together with any configuration (core/sockets/memory) and includes one year of support and maintenance. For more uptime guarantee, everRun MX can also provide N+1 protection when three or more active servers are used.

This article, “Marathon beats VMware to the SMP fault-tolerance punch,” was originally published at InfoWorld.com. Follow the latest developments in virtualization and cloud computing at InfoWorld.com.