The majority of monitoring tools we rely upon report to a granularity of 1 second. For instance, measuring bandwidth in bits per second and server performance in messages per second. On the face of it this makes complete sense, because as humans we’re familiar and comfortable with the concept of a second. But how about something smaller, like tenths of a seconds? Could you accurately judge the difference between 0.2 and 0.3 seconds? I know I couldn't.
In contrast, machines (computers to be specific) live in a world that's incomprehensible to us. Processors are clocked billions of times every second. Applications process hundreds of thousands of transactions every second. And as these systems communicate with each other, huge volumes of data are produced within thousandths of a second.
Our networks, which facilitate these transfers, absorb the bursts with buffers as they forward the data onto network links. And because the buffers can fill much quicker than they empty the bursts have long lasting effects, specifically in terms of introducing latency and also packet loss when the buffers overflows.
So do one second time frames still cut it when managing real-time applications?
No they don’t.
At Corvil we refer to these transient bursts as “Microbursts” and we have discovered them to be the root cause of intermittent user experience and application performance brownouts. But because microbursts only last for a fraction of a second they fly under the radar of most monitoring tools which average out activity over a full second. Many operations teams who are unaware of microbursts find themselves chasing their tails trying to diagnose intermittent issues, wasting time and pointing fingers at other teams.
Essentially Corvil allows you to see and understand the impact of microbursts in real-time because every packet is seen and analysed. This approach allows the instantaneous rate of traffic to be continuously calculated, rather than waiting a full second to produce a diluted version of the truth.
To illustrate this point...
A great example of a tool that reports instantaneous rate is the speedometer in your car because it continuously reports the speed you are travelling in miles per hour. But imagine a world where speedometers behaved like a traditional monitoring tools and only reported your speed once an hour every hour. If your journey only took 20 minutes, you’d have to sit in the car and wait another 40 minutes for your average speed to be displayed. It would be totally useless and there’d be chaos on the roads with high speed collisions and unpredictable journey times – much like a computer network! You’d probably also receive a string of speeding tickets because you don’t know the instantaneous speed of the car.
So how did the traffic cops measure my speed? That’s easy - their radar guns are using Corvil technology!