Locality of Metrics Collection and Storage

If you're an organization that has more than one physical presence, or you use a cloud infrastructure whose physical location isn't quite known, you'll run into the issue of having multiple collection systems and the question of how to aggregate those metrics into a single place for easy comparison and cross-location correlation.

This is commonly found in websites that are being served from multiple datacenters. Let's say we have web servers and databases in two different datacenters, and we're geographically load-balancing the traffic across the two. I'll definitely want to know what the total traffic (say, Apache requests) is quickly and easily, without having to look at two graphs or values and adding them up. Our tool should be able to collect metrics locally in each datacenter, to keep the cost of gathering and transporting metrics low. At the same time, you'll also want to be able to aggregate data from multiple locations. Figure 3-1 shows an example of this.

Flickr photo-serving requests, gathered and aggregated from six datacenters

Figure 3-1. Flickr photo-serving requests, gathered and aggregated from six datacenters

Whatever metrics collection tool you use, being able to get metrics in and out easily should be considered mandatory.

Get Web Operations now with the O’Reilly learning platform.

O’Reilly members experience books, live events, courses curated by job role, and more from O’Reilly and nearly 200 top publishers.