Web Operations

Conclusion

Gathering, storing, and displaying metrics should be considered a mission-critical part of your infrastructure. Whether it's for troubleshooting on-the-fly, capacity forecasting, product launch planning, or application feedback mechanisms, you will be lost without the right metrics to give you a full view of what your infrastructure is doing. At Flickr and Etsy, we depend on these metrics on a daily basis to give context to our work.

All good metrics collection systems share the same traits:

• They can collect metrics while keeping collection and storage costs low.

• They can aggregate and organize metrics once they're collected.

• New devices can automatically be added to the system without disruption or overhead.

• New metrics can easily be added without disruption.

• Metrics can be pulled out as easily as they are put in, allowing for correlation with external sources of data.

• Metrics collection can be done on a global and physically separate scale while maintaining integration.

These traits allow for your metrics system to scale and provide the flexibility to measure and record all of the things you'll need to manage your site.

Matt Massie leaves you with this advice: think about security when you design how data flows through your system, and make it easy to export data to other applications. In the end, you want to create a system that people trust and can be creative with. Once you have your operations collecting metrics data, you may even find that tracking the data is actually ...