On 11/04/2014 11:30 AM, pilger wrote:
@yun
What if they guarantee *my VPS* will have the 100 IOPS?
You've reached the limits of my knowledge :]
BTW I neglected the read ops before. I don't have useful data for a
single TF2 instance (I typically run ~4 TF2 servers per VPS) but one of
my VPS bursts up to 480 read ops/sec. Write ops rarely go above 10/s,
even with replays + logging. In terms of throughput reads burst over
16MB/s but writes are <100kB/s.
But I stress that these numbers include 2 x MvM servers + 1 x 24p all
full at the time. I can't break out the numbers for the 24p on its own
because the per-process data that I have shows the read reqs made by
srcds but a bunch of that would be satisfied by OS caching so it doesn't
translate to actual disk I/O.
Which points at another issue with trying to just run the numbers --
there's caching happening within your VM, and possibly underneath as
well, and both would be dynamically changing depending on what you &
your neighbours are doing.
It keeps coming back to -- try it and see :]
BTW I understand your pain. Australia is also expensive (relative to
US/Europe but I think cheaper than Brazil) and over 16 months of running
TF2 servers we have been through 4 providers and migrated hosts within
providers umpteen times. That's how I got to the point of monitoring
crazily and to the conclusion that you have to talk to the hosts + be
prepared to trial/move a lot.
_______________________________________________
To unsubscribe, edit your list preferences, or view the list archives, please
visit:
https://list.valvesoftware.com/cgi-bin/mailman/listinfo/hlds_linux