As you’re probably aware, Pingdom has been growing a lot over the past couple of years, and as more and more users have joined our service, our infrastructure has grown as well. Our monitoring network now consists of 40 locations that test your sites and servers.
Starting this week, we’re going to change how these monitoring locations monitor your sites. It’s not something most of you will notice, monitoring will be as reliable as ever, but we thought it was worth a blog post explaining what we’re doing and why to avoid any potential confusion.
How monitoring has worked until now
The way it has worked until now, tests of a site have been rotated through all our monitoring locations.
In other words, assuming today’s 40 locations and 1-minute monitoring, each location tests your site once every 40 minutes. If you use 5-minute monitoring, that gets spread out even more, each location testing your site once every 200 minutes (3h 20m).
The problem with our historical approach
As you can imagine, one problem with this is that it makes it increasingly difficult to see trends per location as our monitoring network grows. With a bigger network, each location tests your site more seldom. And we keep adding more locations.
This would be fine if all we cared about was pure uptime monitoring, i.e. testing if a site works or not. But we also care about trends, including performance trends.
In short, our old approach doesn’t scale in the long term, and Pingdom will have many more monitoring locations. We’re already getting too big for our old approach.
That’s why we’re now making a change.
New approach, grouping our monitoring servers
We’re going to start grouping our monitoring servers into slightly smaller groups (subsets) that will each act as a separate monitoring network. Each site or server monitored by Pingdom will be monitored from one of these groups. It will all be handled automatically, there are no settings our users have to worry about.
The locations in each group are still geographically dispersed, but now any “trend tracking” you do will be more precise, since each location will test your site more frequently.
It’s also worth pointing out that there are still enough locations in each group that monitoring reliability will not suffer at all.
Initially we’re grouping our existing 40 locations into 4 groups of 10 monitoring locations each (the exact numbers aren’t important, and will grow and change over time). Just as today, each group contains a mix of locations from North America and Europe.
Added bonus & final words
An added bonus is that it now also becomes easier for you to filter by location (for example in the response time report) since you don’t have to wade through quite as many locations. And when you do select a location, since tests are performed more often by each location, your results will have a higher resolution.
This location grouping also has some nice benefits on Pingdom’s backend and the way we aggregate data, which will help us introduce some pretty cool new features in the future. We can’t tell you about these plans quite yet, but they’re coming.
As we said in the beginning, most of you won’t notice anything at all since this is a “behind-the-scenes” change, but we like keeping you guys posted on what’s going on.
And as always, if you have any questions whatsoever, please don’t hesitate to get in touch with our support (support (at) pingdom.com).