Marathon CTO Jerry Melnick: A Fault-Tolerant Approach to Virtualization

In March of this year, Marathon Technologies announced everRun VM, a first-of-its kind, fault-tolerant, high availability software package for server virtualization. This product picks up where its first-generation kin — everRun — left off in providing companies the ability to prevent outages and data loss in Citrix XenServer virtual infrastructures.

The chief architect of this new software strategy, CTO Jerry Melnick, saw a void in the marketplace left open by leading vendors in the server virtualization space. More recently, Melnick has largely been responsible for earning the company a number of high-profile awards for his leadership in developing fault-tolerant, high availability and disaster recovery software for server virtualization. The company won a 2007 VMworld New Technology award for bringing fault-tolerant availability to virtual servers.

The significance of this new software for the IT industry is threefold. First, it enables companies to run high-value production applications in virtual machines. This provides them the benefits of virtualization across a much broader range of applications.

Second, it makes it practical for companies to make high availability and disaster recovery a standard part of the IT infrastructure for midsize and larger companies. Three, it creates the platform to move virtualization technology into the next wave of server virtualization adoption.

Another potential outgrowth of Melnick’s leadership might be encouraging many companies to take a critical step in migrating their highest value, business-critical applications to virtual servers. The reluctance for some companies to deploy virtualization can often be attributed to performance concerns and a lack of true high availability.

everRun protects virtual workloads by providing redundant virtual machines (VMs) and synchronized mirroring of the whole system.

Through tight integration with Citrix XenServer, everRun VM breaks through the availability bottleneck by providing completely automated VM availability. The product automates VM availability setup and configuration. IT managers can usually deploy it in less than thirty minutes, according to the company. It also automates fault and policy management so that the software maintains itself.

“Based on our research, high availability solutions, including clustering software and proprietary hardware, have never exceeded 20 percent deployment rates,” said John Humphreys, program vice president of virtualization software for research firm IDC. “The reason: they are too complex to set up and too complex to manage. The combination of easy-to-use virtualization and availability software could bring availability to the masses.”

LinuxInsider met with Melnick for an in-depth look at the everRun process and its potential to accelerate the virtualization trend.

LinuxInsider: Numerous industry reports show that the rush to adopt server virtualization is run mostly by very large corporations with a large bank of servers. What part of the market is your company targeting? Are you following the big money trail?

Our product is designed to work like a standard application taken off the shelf and installed right out of the box. Existing availability systems can cause more problems than they solve. Our system is different. That is precisely what midsized companies need.

LI: What barriers does everRun VM take down that you expect to drive a company’s need for what you offer?

For one, there is a cost advantage over high-end proprietary products. On the other end, data replication solutions don’t provide all the functions needed. Replication is managing data only. The comprehensiveness of our solution is what makes it unique. The application is unchanged in our environment. In addition, our solution does not need lots of setup.

LI: In order to make a real impact on server virtualization, your software release has to be more than just an upgrade from earlier versions of everRun itself. What new technology did you introduce to make everRun VM revolutionary?

We took the same product [everRun] and revised the degree of availability functionality. With virtualized servers, the needs for availability differ. Companies until now had to accept a trade-off. But full tolerance is not what everybody needs. What we built into everRun VM is the ability for IT to dial into the software the degree of the goals they want on reaching a balance between availability and recovery. Within that strategy, we use the same technology with the same skill set.

LI: Marathon already offers these earlier software products for server virtualization protection. How is what you created in everRun VM so innovatively different?

What we provide in everRun HA is component-level fault tolerant protection for essential applications. If any I/O (input/output) component fails, everRun HA automatically redirects I/O away from the failed component to components in the other server. If a server fails, everRun HA automatically and immediately restarts on the secondary server.

everRun FT delivers system-level fault tolerant protection for mission-critical applications that simply can’t afford even a brief interruption in service. With system-level fault tolerance, the Windows Server application runs redundantly on both servers, so if a server fails, you get continuous availability with no loss of application state or loss of data.

Marathon’s partnering with Citrix, the developer of XenServer, provided a product that integrates Marathon’s everRun with Citrix XenServer Enterprise Edition. This lets businesses run their critical applications in virtual machines with reliable protection by providing redundant virtual machines and synchronized mirroring of the network, storage and data.

LI: What was the starting point for this new approach?

Two or three years ago, I saw the need for using multiple applications in virtualization. I realized that this would be a new direction for virtualization. The rest of the world was building out consolidation through virtualization. This process compounds availability concerns. If we married our technology to industry standards, we could bring virtualization to a new level.

LI: What are you thinking about beyond the accomplishments of everRun VM as the next virtualization enhancement?

Now we have taken the process to the next leap. What I mean is we can now take two servers with hardware virtualized and build a Xen server pool and manage it as a single entity. We can build a virtual machine in Windows by installing out software and give it an interface of all running servers. Our next vision is to have one mechanism to do availability with multiple uses. It will take six months to one year. This will add new levels of choices to select plug-ins for desired results. All of this will be possible through one product.

LI: How do you see this vision changing the industry?

We’re breaking through limitations. We are in a position to expand use. Availability is something that everybody is talking about. We are providing ways of doing it.