Which KPIs do you use to measure network performance?
Key performance indicators (KPIs) aren’t just for evidencing performance levels to others; they’re an invaluable tool for you, enabling you to measure and assess the quality and capability of your network.
Managing, updating and scaling your network is pretty much impossible unless you can visualize what it’s doing – and this is especially true with software defined networks (SDNs).
Key performance indicators (KPIs) aren’t just for evidencing performance levels to others; they’re an invaluable tool for you, enabling you to measure and assess the quality and capability of your network. While there isn’t a defined set of universal standards, the following categories tend to be the ones most commonly used.
The most common KPIs are CPU and memory allocation, memory utilization, temperature and fan status. You should be able to obtain these values (or similar) from every physical device on your network.
Ideally, you’ll want to be able to see the availability of every device on your network, in as near to real time as you can get it. If you only become aware that a router or a managed switch isn’t working when someone calls you to complain, that’s too late.
Latency and packet loss.
Measuring device latency and packet loss gives you an early warning of any problems and can tell you a lot about the health of a device and that part of your network. If latency is increasing or generally bad, you’ll know that you have a problem. If you can see packet loss, then something unacceptable is going on. It’s time to investigate.
Using SNMP polling you can see common KPIs, such as volume of network traffic. You can also see errors and discards per interface, inbound and outbound. You can poll an interface for availability and – like a device – if it doesn’t answer its ping, you can assume that it is down or broken.
You can also use link bandwidth data to assess the capacity of the network and predict the need for capacity upgrades.
Link statistics tell you what is occurring on the link itself – particularly important when you use the cloud for some portions of your network or lease a network segment from a vendor. Running tests will tell you if the link is working; and using link latency and packet loss jitter as KPIs will give you a real-time view into that segment of your network.