Snap – A telemetry framework • Hacker News Board

Impressive. Are the collector plugins meant to be run on bare metal Intel CPUs? Are virtual Intel cores supported (running collectors in VM guest)? Are other non Intel CPUs theoretically supported? (most cloud hosters use Intel, though some may use something else)

Re: Snap – A telemetry framework

#6

Post by danielscottt » Thu, Dec 10, 2015, 5:53 PM UTC

frik wrote: post #5

Impressive. Are the collector plugins meant to be run on bare metal Intel CPUs? Are virtual Intel cores supported (running collectors in VM guest)? Are other non Intel CPUs theoretically supported? (most cloud hosters use Intel, though some may use something else)

1. The plugins can be _anything you want_. :) They do not have to be Intel-centric. Ultimately, the goal was to make it easier to get Intel telemetry out of the system and be able to correlate that with whatever data you may already be collecting.

2. That depends on the plugin. A plugin which is reading a MSR* probably would not, like our PCM plugin for example, but a plugin which is OS based like say, the perfevents plugin, would totally work on a virtual guest.

3. Your third question could also be addressed by #1 above. All it would take to collect data from other architectures or platforms would be the right plugin!

*https://en.wikipedia.org/wiki/Model-specific_register

Re: Snap – A telemetry framework

#7

Post by lynxbat » Thu, Dec 10, 2015, 8:05 PM UTC

frik wrote: post #5

Impressive. Are the collector plugins meant to be run on bare metal Intel CPUs? Are virtual Intel cores supported (running collectors in VM guest)? Are other non Intel CPUs theoretically supported? (most cloud hosters use Intel, though some may use something else)

Plugins can also call remote API in a proxy mode. This is especially useful for hypervisors like ESXi.