Happily upgrading Ruby on Rails at production scale
webuild.envato.com
Happily upgrading Ruby on Rails at production scale
1–6 of 6 posts
Re: Happily upgrading Ruby on Rails at production scale
#2Would you mind to elaborate on
- how many of those requests are responded from cache
- the underlaying deployment architecture (type & number of servers and/or VMs)
- and you average server response time (uncached)
Thanks
EDIT: Why the downvote?
Re: Happily upgrading Ruby on Rails at production scale
#3"8000 requests per minute" sounds quite a lot for a RoR system. Would you mind to elaborate on - how many of those requests are responded from cache - the underlaying deployment architecture (type & number of servers and/or VMs) - and you average server response time (uncached) Thanks EDIT: Why the downvote?
Re: Happily upgrading Ruby on Rails at production scale
#4"8000 requests per minute" sounds quite a lot for a RoR system. Would you mind to elaborate on - how many of those requests are responded from cache - the underlaying deployment architecture (type & number of servers and/or VMs) - and you average server response time (uncached) Thanks EDIT: Why the downvote?
So: the downvotes are for calling ~65 req/s "quite a lot for a RoR system".
Re: Happily upgrading Ruby on Rails at production scale
#5"8000 requests per minute" sounds quite a lot for a RoR system. Would you mind to elaborate on - how many of those requests are responded from cache - the underlaying deployment architecture (type & number of servers and/or VMs) - and you average server response time (uncached) Thanks EDIT: Why the downvote?
I mean, 8000 requests per minute on a quad-core systems is 33 request/second per core, or 100 million CPU cycles per request on a 3.3GHz CPU.
That seems enough for the web server to take a vacation in the middle of responding.
Re: Happily upgrading Ruby on Rails at production scale
#6"8000 requests per minute" sounds quite a lot for a RoR system. Would you mind to elaborate on - how many of those requests are responded from cache - the underlaying deployment architecture (type & number of servers and/or VMs) - and you average server response time (uncached) Thanks EDIT: Why the downvote?
- None of those requests are served from cache. That's the number hitting the Rails stack and being served dynamically (there are some cached pages on top of that. not sure of the number of RPM off hand)
- Deployed on nginx/unicorn running on linux servers. I think we've got 17 web servers at the moment. Not sure of the specs sorry.
- average server response time is 155ms (again, not using caching)