[sldev] Linden Lab and unfair business practices

[David M Chess]

Boy, am I not liking this thread's Subject line.  :)

> From: "Gareth Nelson" <gareth at litesim.com>

> On the technical side of things, just saw this on the JIRA:
> Dale Innis - 28/Oct/08 12:27 PM
> Seems like it should be possible to prevent "abuse" (i.e. people
> loading their OpenSpace sims to an extent that hurts the performance
> of the others on the same machine) by appropriate fiddling with the
> operating system or hypervisor or whatever. Talk to the techies, and
> put in CKRM policies or whatever to ensure that each sim process gets
> at least its fair share of the CPU. (If it's about disk access or
> network bandwidth rather than CPU, there are perhaps similar things
> that would let you do similar things for those.)
>
> I'd go further by dynamically renicing "abusive" sim processes down,
> but that's just me.......

Thanks for the quote, Gareth; I was just looking through sldev to see if 
this technical question had come up.

It actually seems likely to me that they've considered using CKRM policies 
or hypervisor settings or some equivalent to insulate the OpenSpace sims 
from each other, and to keep them from using more than one-quarter of a 
machine when there's contention, and it just turned out not to work for 
some reason.  The technical folks who discovered that it didn't work 
probably aren't quite sure of exactly why, and they probably haven't 
described the issues in enough detail up the chain that it can come back 
to us via Jack with any fidelity at all; information flow is a difficult 
thing.

BUT, just in case, because stranger things have happened, I do hope that 
someone reading this forwards a quick note to someone on the ground, 
saying "are you guys aware of the severity of this OpenSpace sim resource 
contention problem, and have you looked into using the OS or CKRM or the 
hypervisor or whatever to limit the OpenSpace processes from using more 
than a quarter of the machine when there's contention?", and then maybe 
even lets us know what the reply is. 

It could be, like, "oh, is that still a problem? sorry, I was busy with 
something else; we'll give CKRM a try and let you know; it'll probably 
work."  And think how much pain that could avert!  :)

Of course people will (still) complain when it turns out their OpenSpace 
sims get really slow under loads that don't stress a full sim at all, but 
well... TANSTAAFL an' all.

-- Dale Innis (DaleInnisEmail at gmail.com)
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.secondlife.com/pipermail/sldev/attachments/20081029/fb65607f/attachment.htm