Load Balancing Metrics
MRTG is an extremely useful tool that combines data collection with graphical trending. In its simplest configuration, you can graph the bandwidth in and out of any SNMP enabled network device. This includes, but is not limited to, servers, routers, and switches.
Also capable of being graphed are statistics and metrics for load balancers. I've compiled various objects and MRTG configuration files for a variety of load balancers and related products. Depending on the load balancer vendor, with MRTG and the correct configuration, you can graph the following:
Bandwidth in and out in bits or bytes per second
Connection rate (in connections per second)
Bandwidth in an out of a particular VIP/virtual server or node/real server
Total number of concurrent sessions
Any parameter that has an SNMP counter or gauge object/OID
Many, many more
I'm alwys looking for more objects, vendors, and products. If you have information on a load balancer, SSL Accelerator, or related product, feel free to submit it.
A Picture Is Worth 1,000 Words (or data points)Here is an example of what MRTG configured for load balancers can do. This is a graph showing connections per second, such as HTTP connections per second:
This image clearly shows how activity has progressed over the past 36 hours, and does so in a way that even your manager can understand. Without, you know, boring them to death.
Setting up MRTG for your load balancer is easy, cheap (MRTG is free, the hardware to run it can be very modest) and takes very little time. The return on that investment can be enormous. For more on my somewhat fanatical endorsement on using MRTG, read my SLB and MRTG Manifesto.