As virtualisation becomes more pervasive, the burden on the physical server hardware becomes a critical concern. And the burden on the physical server is not simply a performance issue — fault tolerance and redundancy features become ever more important as a single physical server or blade failure can bring down mission critical virtual servers.
A poor choice of virtual application and resource allocation can result in disaster or a clunky and poorly performing server network, even on the most powerful and fault tolerant physical server.
The servers in this review are aimed at medium-sized businesses. They range from low-cost 1RU servers with relatively little in the way of fault tolerance or redundancy, such as the Xenon Radon R1150, up to the HP c3000 blade enclosure, which offers fault tolerance and redundancy second to none. It is worth noting that all the servers, including the Xenon Radon R1150, support memory mirroring.
Don’t, however, dismiss the Xenon Radon R1150, despite its minimal fault tolerance and redundancy. A handful of R1150 servers in a rack with virtual application servers and their virtual secondaries strategically distributed between physical servers can arguably provide more fault tolerance than the HP c3000. The overall performance and robustness of your virtual servers rely on more than just the physical hardware and resources — the application deployment procedures and control are just as critical.
Read the Reviews:
- Fujitsu Primergy RX300 S5
- HP c3000 Enclosure and BL490c Blade
- HP Proliant DL380 G6
- Lenovo ThinkServer RD220
- Xenon Radon R1150 1U
- Xenon Radon R4550 Quad 2U
The hardware equation
In a perfect world, the performance and allocation of hardware resources will ultimately determine the performance and reliability of your virtual environment. As far as performance is concerned, a potent physical server should have plenty of resources to share amongst the virtual machine (VM) applications including central processing units (CPUs), size and number of drives and LAN I/O. The more the merrier.
Of course, if a critical component in a physical server fails, you lose the whole shooting match so redundancy is undoubtedly critical. The simplest way of achieving redundancy is using a collection of relatively low performance physical servers; if one dies, there are plenty of others to take over. A more elegant, and often more costly, solution is to build redundancy into the server in the form of independent blades or nodes — these can be processing, storage or LAN I/O — and connect them to fault-tolerant backbones and redundant power supplies.
All of the servers in this review are multi-socket and multi-core — perfect candidates for virtualisation, with most based on a typical server architecture of a single motherboard embodying most, if not all, of the processing, memory and I/O. Two of the servers, however, provided a different route. The HP c3000 enclosure is an all-singing-all-dancing blade server that takes fault tolerance to a much higher level. The Xenon Radon R4550 has four independent processing nodes, pretty much blades for all intents and purposes.
Next: Processor, Chipset, VM management