| |||||||||||||
prefetch and buffer bloatprefetch and buffer bloatPosted May 24, 2011 16:15 UTC (Tue) by grunch (subscriber, #16603)Parent article: The problem with prefetch
This seems like a microscopic analog to the buffer bloat situation: trying to improve performance in a way that ultimately negates improvements the underlying handlers (hardware in this case, TCP in the case of buffer bloat) can already handle.
I wonder if it's a similar cause.
Thanks for the details Jonathan!
(Log in to post comments)
prefetch and buffer bloat Posted May 24, 2011 16:22 UTC (Tue) by martinfick (subscriber, #4455) [Link]
Not really. The buffer bloat problem is due to the tradeoff between a marketing performance improvement in throughput versus latency which is rarely marketed or even considered (until now).
prefetch and buffer bloat Posted May 24, 2011 16:50 UTC (Tue) by nye (guest, #51576) [Link]
I think that's an overly cynical viewpoint. It's likely that the great majority of cases of too large buffers are simply due to developers thinking "I don't know what's an appropriate value; let's just choose something big because it can't hurt and might help sometimes".
prefetch and buffer bloat Posted May 24, 2011 17:11 UTC (Tue) by martinfick (subscriber, #4455) [Link]
While the marketing part might be cynical, I believe the tradeoff is important to recognize. In the case of bad prefetch, there is no tradeoff that I can see, it simply degrades the performance.
prefetch and buffer bloat Posted Jun 2, 2011 21:54 UTC (Thu) by jch (guest, #51929) [Link]
Sorry if this is off topic for this discussion, but bufferbloat is not just a political issue -- it's a difficult technical one. Just reducing the size of buffers won't do, since the right amount of buffering depends on a lot of factors such as throughput, RTT, the transport- and application-layer protocol being used, etc.
Bufferbloat is not about reducing the amount of buffers in routers; it is about designing algorithms to make sure that routers only use as much of their buffers as necessary, and getting the router vendors to deploy such algorithms.
--jch
prefetch and buffer bloat Posted May 24, 2011 17:27 UTC (Tue) by nicooo (guest, #69134) [Link] I doubt it's a marketing issue. Having their current clients telling their potential clients that the service sucks can't be good for business.
prefetch and buffer bloat Posted May 24, 2011 22:58 UTC (Tue) by Lennie (subscriber, #49641) [Link]
The problem is with the numbers that come out of the tests the manufactures do to show how great their hardware performs. Those get better with bigger buffers.
So again you have manufactures doing non-real-world test (which might have been a good test a long time ago) for marketing purposes and optimising for that case.
prefetch and buffer bloat Posted May 31, 2011 20:43 UTC (Tue) by marcH (subscriber, #57642) [Link]
> The problem is with the numbers that come out of the tests the manufactures do to show how great their hardware performs. Those get better with bigger buffers.
... up to a size after which throughput does not get better. Yet we can sometimes see buffer sizes way past this point (e.g. 1 second), which proves that some manufacturers do not bother trying to optimize anything at all.
prefetch and buffer bloat Posted May 31, 2011 21:05 UTC (Tue) by dlang (✭ supporter ✭, #313) [Link]
something to remember is that memory comes in standard sizes. it's not always possible/reasonable to put less ram in the device.
since they have the ram anyway, and buffers that are too small can cause problems. the logic then follows 'why not just use the ram in the device as a buffer'
this causes other problems, but these other problems were not well described until recently.
prefetch and buffer bloat Posted May 24, 2011 17:53 UTC (Tue) by clugstj (subscriber, #4020) [Link]
Yes, it is marketing - whether from the marketing department or development engineering. You can't sell someone a faster ethernet card unless you can demonstrate higher throughput with no dropped packets, and the easiest way to do that is to increase the number of buffers.
|
|||||||||||||