I used to love blade servers. Then I hated them. Now I love them again.
Blade servers went through a few phases before they became ready for virtual infrastructures. But today there are numerous advantages of blade servers when they are installed in a virtualization environment. And servers in a blade enclosure can improve virtual server load balancing and solve the problems surrounding server-generation differences.
Pre-virtualization blade server challenges
Blade servers arrived before the advent of virtualization. At that time, one of their major selling points was their automated management tools. I still remember the first hardware vendor whose blade presentation suggested literally taking one blade server out of a blade enclosure and replacing it with another. At that time, blade management tools would recognize the replacement and automatically rebuild the old blade server's operating system onto the new hardware.
The goal was to rapidly replace failed equipment. If one blade server failed, you could simply yank it from the blade enclosure and insert a replacement. Then the management tool took care of the rest.
But most of us never got to that fully automated nirvana, in part because getting there was difficult. Before virtualization, the automated rebuild process was still fairly nascent. It required a substantial time investment to create the proper OS images for delivery. These images changed over time, which further complicated the process with physical servers.
Early-generation blade servers: Vendor promises
Eventually, virtualization gained prominence, and with it came the second wave of blade server vendors. These vendors suggested that, for the same reasons as they had previously, a blade enclosure was the perfect fit for virtualization: One could rip and replace a blade server and watch the new one automatically rebuild with a virtual host's OS.
But this time, the important difference was that a virtual host's OS is relatively stateless. Since most changes occur in the guest virtual machine (VM), the host OS became an easily hot-swappable item.
During this second period, I still hated the blade concept. Despite vendor claims, the early-generation blade enclosure simply wasn't designed with virtualization in mind. It was limited to only 2 or 4 single-Gb network interfaces -- far too few for most virtual infrastructures' needs. Some blade servers were limited to Fibre Channel host bus adapters as well, creating a hardware bottleneck in the areas where performance is most important.
The advantages of blade servers with virtualization
Today these early port limitations no longer exist, and hardware vendors have finally come around. Some blades and blade enclosures are now designed specifically for virtualization, sporting multiple 1 Gb and 10 Gb connections in addition to plenty of Fibre Channel interfaces.
Blade management and automation have also significantly improved. With offerings such as VMware's near-stateless ESXi hypervisor and Microsoft's lightweight Windows Server Core, it's a piece of cake to lay down a new host OS when a blade server has been replaced.
More important, automation at the blade level has become more integrated with VM automation at the hypervisor layer. A blade server failure (or, in some cases, a pre-failure warning) can now quickly signal to VMware vSphere or Microsoft System Center Virtual Machine Manager that VMs must be migrated elsewhere.
Once virtual machines have been failed over to another host, it's simple to deploy a new OS on the original, now-empty host. That host returns to functionality quickly. Finally, the manual blade-server replacements of 10 years ago meet real-world utility.
Blade servers and the generational problem
When it comes to virtual machine migration and management, blade servers bring important advantages.
Today's blade servers present a solution to one of the most painful and least-understood long-term problems of private cloud computing: the limitations of technology generations. As your infrastructure ages, it becomes more difficult to live-migrate between new servers and older ones and to load-balance among different VMs on these servers.
In my recent book, Private Clouds: Selecting the Right Hardware for a Scalable Virtual Infrastructure, I discuss how hardware generations can create problems in a virtual infrastructure and private cloud: "Virtual environments … have a tendency to outlast typical physical servers due to their resource abstraction. The problem is that sometimes different hardware generations don't necessarily integrate well."
Hardware's generational problems particularly affect VM load balancing. Live migration capabilities work only when processors at the source and target host are relatively similar to one another. We all know, for example, that you can't live-migrate between AMD and Intel processors.
What you may not realize, though, is that even processors from the same manufacturer must be relatively similar. A VMware knowledgebase article indicates which processors can vMotion among one another. They're more limited than you'd expect.
Over the short term, the generational limitation isn't usually a problem, but as your virtual infrastructure ages, it can create problems. Every time you add additional servers, the likelihood that a new batch of servers can't live-migrate with your previous batches grows.
How a blade enclosure improves server load balancing
Here's where the advantages of blade servers come in. A blade enclosure can solve the generational problem. Blades within an enclosure are more likely than individual servers to be from the same generation. This similarity means that you should be able to load-balance virtual machines within an enclosure, though you may not be able to among enclosures. That's enough server load-balancing capacity for most organizations' needs.
As time progresses and hardware accumulates, the generation gap tends to worsen, so take a hard look now at your hardware vendor's virtualization-aware blade enclosures and other such offerings.
About the expert
Greg Shields is an independent author, instructor, Microsoft MVP and IT consultant based in Denver. He is a co-founder of Concentrated Technology LLC and has nearly 15 years of experience in IT architecture and enterprise administration. Shields specializes in Microsoft administration, systems management and monitoring, and virtualization. He is the author of several books, including Windows Server 2008: What's New/What's Changed, available from Sapien Press.