Hardware-Based HA for the Masses
Stratus Technologies announced that the Avance software-only high availability product gained certification on server systems offered under the Intel Enabled Server Acceleration Alliance. In the land of usability certificates commonly inhabited by VARs and ISVs, industrial-strength Stratus Avance software can be stamped with a seal of approval that makes it easier to put the product in branch offices and smaller organizations.
The whole relationship between Stratus and Intel's ESAA (Enabled Server Acceleration Alliance) is a fairly complex chain of endorsements that boils down to two basic questions: "How much will this cost?" and "how is this measured on my service contract?"
Using Avance, a high-availability link can be created between two certified-but-otherwise-general-purpose Intel-based servers. According to Stratus, this HA platform can support off-the-shelf applications from Microsoft and Red Hat. The Avance software, along with an online monitoring service from Stratus, automatically monitors the two hardware systems for faults, switching processing to the surviving hardware system if one part of the pair fails. Stratus and the reseller can work together in a number of ways to service customers who use Avance to survive a failed physical computer.
In many ways this sounds more like the world of yesterday rather than the world of tomorrow. The Avance system running in this certified environment limits server choices to Intel-only products. When I talked to Stratus officials, they had little to say about using Avance in a computing environment running on servers based on AMD chips.
I would add a third question to the two I posed above: "What happens when a physical high-availability product such as Avance meets a virtual high-availability product such as VMware?" VMware is an Intel ESAA member, as is Microsoft, Citrix and Red Hat. The likelihood that one or more of these virtualization tools will show up in the mix that would benefit a small or branch office seems very possible. Aside from the messy technical details of what might happen if both systems were attempting to manage where a VM (virtual machine) would end up in the case of a hardware failure, the fundamental question might be, "why should the customer care?"
Breaking the tie between workloads and underlying physical servers even in small offices looms on the horizon. And breaking the tie between workloads and privately owned compute resources, especially for an office that makes extensive use of technology but has already made the decision to outsource IT operations, looms even more so. Thus, while I'm fascinated by the intricate signaling and thoughtful coding that makes a single application run seamlessly on two physical computers, I'm left with the notion that I'm one of the people who witnessed John Henry beat the steam shovel. It can be done, but which method will prevail?