You may have no idea what Avocent does, or quite possibly, you have never heard of the company. It flies beneath most people’s radar. Avocent is the primary supplier of Embedded Service Processors to the hardware giants – including Dell, IBM, Intel, Lenovo and others. It has revenues of $617m for 2007 and it is growing fast, having nearly doubled its revenues in 2 years. That’s quite a clip really – and it means that it’s selling something that’s quite compelling. So what is a Service Processor?

Service Processors

A Service Processor is a micro-controller that you can embed on a server board, which enables the remote monitoring and management of the board. It has a completely separate life from the CPU, OS and any applications that might be running on the server and can be connected to a separate network (via ethernet) as illustrated in the diagram below.

You can think of a Service Processor as a nerve ending in the nervous system of the network. It complements management software agents that might run on the CPU. In a number of ways it is a preferable mechanism for determining the status of the server than running a systems management software agent. A software agent will disappear without trace if the OS dies for any reason, including the failure of the CPU, and it can also die of its own accord. The Service Processor only dies if the electricity to the board it lives on dies.

The rough distinction between a Service Processor and a software agent is that the Service Processor is best placed to monitor hardware elements and firmware, whereas the system management agent is well placed to cover everything above those layers of the stack. If you want the nitty-gritty detail of Service Processor capabilities, then you need to bone up on IPMI, the Intelligent Platform Management Interface, an industry standard which was designed specifically to improve server management in mixed server environments. It allows the Service Processor to monitor the board it lives on and send out alerts or to run diagnostics over the network.

Service Processors can carry out data logging, set platform event traps, keep a system event log and provide remote access to the server via Virtual KVM (Keyboard, Video, Monitor). It’s most important role though, is to monitor the health of its host device and support graceful shutdown, if it detects that a failure is probable.

Remarkable though it seems (to me at least) the average server uptime is not much better than 98.5%. Translate that into days and it means about 5 lost days per year per server. That’s not the real problem though, because some of that downtime is likely to be planned. The problems (and the costs) arise when the server dies while it’s busy. The Service Processor finesses a good deal of the problem here, because it can pre-empt some of the problems and provide the grace time needed to transfer workloads from one server to another if a hardware failure is inevitable.

In a nutshell Service Processors improve the management of systems by reducing the Mean Time To Resolve an outage. In a world where the availability demands of the IT Network are gradually rising each year (think SOA, virtualization, VoIP, unified communications, etc.), the need to diminish the impact of outages is obvious. And while the costs of servers seem to diminish every year, the cost of data center management has a tendency to rise.

The business argument for deploying Service Processors is clearly compelling and, as it happens, it’s an argument that has already been won. Right now 90% of servers shipped have Service Processors on board.