Egenera spends a lot of time telling the world that our servers are “stateless” and that it really, really matters…Why? The answer lies in uncovering the complexity in today’s datacenters. Ever wonder why it costs more to manage servers than it does to buy them? Blame state. It’s one of the main culprits.
What is server state? State is anything that gives a server a permanent identity. An operating system on a local disk. A MAC address on an Ethernet NIC. A World Wide Name on an HBA. An association with a particular terminal server port. All of these put rigid controls on the server and define its use going forward. That makes it difficult to configure and maintain a server, but that’s really just the tip of the iceberg. It gets more complicated when these servers are mission critical and have to be configured in clusters for failover and disaster recovery. This is where state really can create havoc.
There are thousands of lines of code in mission critical management software that were written just to figure out how to transfer state of a failed server to its backup server, not to mention the un-wiring and re-wiring of network, console, and SAN connectivity. This code is complex and prone to errors given the combinatorial testing problems. Imagine the complexity when a hypervisor and virtual machines are thrown into the mix.
So, what does this all mean? It means that removing state from the server is critical to solving the complexity problem enterprises face today managing the explosion of servers in their datacenters. When state is removed, it unbinds the server from any particular processing hardware, providing an elegant solution to dynamic provisioning and re-purposing, automatic failover, and managing and maintaining mission critical servers. It reduces the number of physical compute resources needed far beyond what virtual machines provide on their own. It removes the complexity at its core.
Brilliant! (to quote the Guinness guys). Well…Actually it’s back to the future by applying a mainframe architecture to commodity processing to get the best of both worlds—best in class manageability combined with best in class price/performance.