Network Performance Management - Dynamic Baselines at Work
Being the network administrator is one of the many hats I wear at SevOne. As the director of systems engineering, I have a growing number engineers to help. And I have my own set of customer accounts and I'm driving new SevOne demo methodology. Nonetheless, it is important to our fast growing company that our network performs correctly every day.
The other day I was doing a network performance management demo from our internal system and showing a customer how we create and graph a dynamic baseline (or display what is normal) against what is currently happening in the network, and I had pulled up a graph of the CPUs in our core switch. What I saw made me smile - we were running perfectly normal for that moment of the day.
Of course I knew this was the case because if we had been out of the “normal” range I would have received an alert on my iPhone but it was reassuring to see anyway. When I looked at the graph what I saw was the baseline visualization (dashed line) plotted against the actual data of today. There was a CPU spike to 85% utilization but the baseline data shows that at this particular time each Tuesday this occurs, and because I have never had a complaint about performance at 5:30 am on Tuesdays I know there is nothing to worry about.
Calculating baselines is no easy task. SevOne holds a baseline for every performance stat we monitor. Our baseline granularity is 15 minutes by default and it can be as granular as 5 minutes. What this means is there is a unique value for each 15 minute period of the week (that's 672 unique baseline values per performance indicator). So Monday from 8:00 - 8:15 is different from Monday 8:15 - 8:30 which is different from Tuesday at 8:15 - 8:30 and so on. For a large SevOne PAS, this means we are calculating over 600 million values a week and checking them every 3 minutes, so we can send an alert if a performance indicator is no longer in the normal range. That's a lot of computational power that "has my back" that just put a smile on this SE director's face!
This ability to automatically understand the “normal” behavior of my networks and servers and only alert me when things are not normal saves me a great deal of time, leaving me the time I need for the other hats I must wear daily. Do you have a favorite SevOne story to share with us? Who has your back?
Dave Hegenbarth, Director of Systems Engineering for SevOne