Industry standard Performance

[Linda Chan]

Hi all,

I'm curious to know if there's any sort of industry standard metrics on API performance (e.g., availability, response time, error rate)? 

Thanks.

[Lukas Rosenstock]

Hello Linda,

there is, to my knowledge, no industry standard, however there are some suggestions and best practices on what should be monitored on APIs.

I find the section on monitoring in http://nordicapis.com/three-best-practices-to-achieve-release-when-ready/ quite helpful.

In a nutshell, it suggests the following metrics:

• Request volume
• 2xx/3xx/4xx/5xx rates
• Server load
• Error log rate

To quote the procedure from the article: "Once you’ve identified these norms and an acceptable range around them, you can then set up alerting for when they are exceeded. This can be as simple as a threshold (we’re over 2% 400/500 rate = alert) to more complicated setups (we’re seeing a larger-than-normal request volume from IP address X over the past 10 mins — is this DDoS?).“

Hope this gave you some hint on how to monitor API performance.

Regards,

Lukas Rosenstock

https://cloudobjects.io/

--
You received this message because you are subscribed to the Google Groups "API Craft" group.
To unsubscribe from this group and stop receiving emails from it, send an email to api-craft+...@googlegroups.com.
Visit this group at https://groups.google.com/group/api-craft.
For more options, visit https://groups.google.com/d/optout.

[Lukas Rosenstock]

Hello Linda,

there is, to my knowledge, no industry standard, however there are some suggestions and best practices on what should be monitored on APIs.

I find the section on monitoring in http://nordicapis.com/three-best-practices-to-achieve-release-when-ready/ quite helpful.

In a nutshell, it suggests the following metrics:

• Request volume
• 2xx/3xx/4xx/5xx rates
• Server load
• Error log rate

To quote the procedure from the article: "Once you’ve identified these norms and an acceptable range around them, you can then set up alerting for when they are exceeded. This can be as simple as a threshold (we’re over 2% 400/500 rate = alert) to more complicated setups (we’re seeing a larger-than-normal request volume from IP address X over the past 10 mins — is this DDoS?).“

Hope this gave you some hint on how to monitor API performance.

Regards,

Lukas Rosenstock

https://cloudobjects.io/

Am 08.08.2017 um 01:33 schrieb Linda Chan <lind...@gmail.com>:

[Diogo Lucas]

No standard, AFAIK. Most APM tools out there will highlight availability and response time metrics (some, such as NewRelic, combine both in their 'Appdex' metric) and will provide further ones (response code rates, endpoint usage, etc) as you drill down. If you are looking for something to use as a reference, you can start there...