two
Volume 3, Number 13 -- April 19, 2006

Marathon Introduces High Availability Software for Windows

Published: April 19, 2006

by Alex Woodie

Marathon Technologies this week introduced new high availability software designed to protect Windows servers from going offline due to hardware failures. Called EverRun HA, the new product is based on Marathon's flagship product, FTvirtual Server. It makes a slight trade-off with its bigger brother exchanging failover speed for a more affordable price tag and greater tolerance for differences in hardware and configurations.

About two years ago, Marathon Technologies introduced a product that made new use of virtualization software. While X86 virtualization products such as Microsoft Virtual Server and VMware GSX Server and ESX Server have helped users drive higher efficiencies by carving their commodity boxes into multiple virtual machines, Marathon saw that virtualization software could also form the basis for fusing two boxes and tricking the operating system and applications into thinking they were a single machine.

FTvirtual Server works by implementing a single instance of Windows Server 2003 (Standard or Enterprise editions) on top of two Windows Server 2003 boxes, for a total of three copies of Windows. The customer foots the bill for all three, however, because only one copy of the application is in use at any time, the user only needs to buy one application license.) If a disk crashes or a network connection is lost on one of these two servers, FTvirtual Server running on the overseer version of Windows directs user requests and I/O to the second server, which was kept in "lock-step" with the first, which means all I/O, processing, and storage is duplicated in real-time on both machines, to create two perfect replicas. Processing, storage, and network are segregated, so that the Marathon software can bypass a failed component on one system while still using the good components on the other. Marathon also sells a product called SplitSite, which provides data replication across a WAN for two servers separated by up to 100 miles.

Compared to other vendors of fault-tolerant gear, such as Stratus and NEC Solutions America, Marathon's approach had the benefit of giving customers their choice of Wintel gear, be it Dell PowerEdge, Hewlett-Packard ProLiant, IBM xSeries, or some other vendor. The one catch is the hardware has to be identical for FTvirtual Server to work. And compared to clustering, which is the other way Windows shops prefer to improve their systems' availability, the lock-step approach of Marathon and its ilk is much simpler to implement and isn't as susceptible to application code changes, which is a thorn in the side of the clustering folk, including Microsoft's SQL Server Clustering Services, EMC's Legato, Symantec's Veritas, and Oracle's RAC.

Marathon started shipping FTvirtual Server in January 2004, and today it boasts 4,500 licenses at about 1,000 customers running a variety of mission-critical Windows applications, including 911 emergency dispatching systems, process control systems for five of the top 10 drug companies, and television broadcasting systems (perhaps the most critical application of all). Last year was a particularly good year for Marathon, which doubled its customer base and grew revenue by 80 percent. The spate of hurricanes and other high-profile disasters didn't hurt either, as disaster recovery and high availability have moved further up CIOs' lists of priorities.

Last week, the Littleton, Massachusetts, company shifted its strategy slightly down market with the introduction of EverRun HA. It also rebranded FTvirtual Server as EverRun FT, and SplitSite as EverRun SplitSite.

EverRun HA

EverRun HA is based on FTvirtual Server (now EverRun FT), but with a few key differences. The most important difference has to do with how the lock-stepping occurs between the two nodes, and how the overseer OS--that third version of Windows controlling the other two--reacts to hardware failure.

With EverRun FT, there are direct and real-time links between the overseer OS and the ones doing the actual work, so if a hiccup happens, the connection to the failed node is severed, and work continues, uninterrupted, on the second node. With this approach, users will not even be aware that there was a serious hardware problem.

With EverRun HA, there are no real-time links between the master OS and the nodes, which Marathon calls co-servers. If a hiccup happens on the active co-server, the EverRun HA shifts the active environment to the backup environment, and redirects users to the backup co-server. This redirection (the company says it's not the same as a traditional failover) entails a restart of the operating system and the application, and can take a few moments, which the users will notice.

Michael Bilancieri, director of product marketing for Marathon, says EverRun HA and EverRun FT share the same code and architecture, and are 99 percent alike. "The EverRun HA product was derived from EverRun FT," Bilancieri says. "The only difference is we don't lockstep the processing in real time. Users will have a brief interruption [with EverRun HA]."

Because EverRun HA doesn't deliver total fault tolerance, and redirection entails some downtime, the application is being targeted at Windows shops that can tolerate some application downtime. It is also less expensive, at $7,500 for two servers (which includes one year of support), compared to EverRun FT, which costs $16,000 for two servers. SplitSite, which works with both the HA and FT products, sells for $10,000 for two sites.

There are also differences in hardware support. EverRun HA can run on one-, two-, four-, and eight-way X86 or X64 servers, in addition to servers running multi-core processors, and has greater tolerance in differences in hardware and configurations compared to EverRun FT, which requires the two servers to be practically identical. EverRun FT also only supports one- and two-way servers, and doesn't yet support multicore processors.

Gary Phillips, Marathon's president and CEO, says EverRun HA marks a significant breakthrough for the Windows high availability industry.

"The high availability market, while it's evolved over the last few years, has really been devoid of any technology breakthrough," he says. "Now what we're doing with the EverRun HA product is extending the majority of that EverRun FT technology in a low-cost, mainstream solution to make it simple enough [for smaller Windows shops to use]. . . . What we're providing to customers for Windows application is a simple, comprehensive, and automated availability solution."



Sponsored By
OPENLOGIC

OpenLogic
Solving the Open Source Dilemma

Open source tools are constantly evolving, but their strengths can cause your headaches. Unpredictable updates and ongoing maintenance can turn productive developers into frustrated system administrators and consume untold hours. OpenLogic can help.

Our flagship product, BlueGlue, enables you to install, integrate, test, manage, and learn over 120 open source projects in minutes.

Take a guided tour and get the free demo at www.openlogic.com



Editor: Alex Woodie
Contributing Editors: Dan Burger, Joe Hertvik,
Shannon O'Donnell, Timothy Prickett Morgan
Publisher and Advertising Director: Jenny Thomas
Advertising Sales Representative: Kim Reed
Contact the Editors: To contact anyone on the IT Jungle Team
Go to our contacts page and send us a message.

Sponsored Links

Vision Solutions:  Get facts on managed availability and business continuity to eliminate downtime
Wolf Computer Consulting:  Reliable service and affordable rates for business computing needs
COMMON:  Join us at the Fall 2006 conference, September 17-21, in Miami Beach, Florida

 


 
Subscription Information:
You can unsubscribe, change your email address, or sign up for any of IT Jungle's free e-newsletters through our Web site at http://www.itjungle.com/sub/subscribe.html.

Copyright © 1996-2008 Guild Companies, Inc. All Rights Reserved.
Guild Companies, Inc., 50 Park Terrace East, Suite 8F, New York, NY 10034

Privacy Statement