On Monday 04 May 2009 21:21:50 Victor Denisov wrote: > Hello, > > Since my node autoupdated to 1208, my system's performance degraded so > much as to render it completely unusable while Freenet is running. I > can't perform simplest tasks, such as surfing or typing a document, as > switching between tabs in Opera can take 10+ seconds and delay between > typing a letter and it showing up in Word can be several seconds as > well. Anything more stressing (such as running a game or developing an > application) is simply not possible. > > From what I can see, the reason for this is that Freenet makes hundreds > of small disk i/o ops per second, basically blocking the OS from > accessing the hard drive for swapping and such. > > The above is definitely affected by the queue size. First, I tried > adding ~ 100 random files from Thaw when the node first updated, but > hadn't had the patience to wait for the request to complete (I think I > waited at least 20 minutes, perhaps more - with the system being nearly > paralyzed by the constant HDD thrashing). With just 3 or 4 files being > put in the queue the system starts to stutter noticeably, provided that > files start downloading and not hang at 0%. The only time when I can run > Freenet as a real background app is when I don't have any files in the > queue and FMS isn't running (which seems more or less pointless to me :-(). > > My system is set up as follows: > > AMD Athlon 64 X2 6000+ > 4 Gb RAM > 2x250 Gb 7200 rpm SATA2 HDD (mirror via mobo's built-in nVidia 570 RAID > controller) > Windows XP x64 > Java 1.6.0_07 64-bit > Of course, all the latest updates/patches/drivers, etc. > Freenet uses 5 Gb datastore on an unencrypted partition. > > Interesting thing I noticed was that Freenet significantly underutilized > the memory I provide it with. From 320 Mb heap memory available, I > hadn't seen it allocate more than ~ 80 Mb. > > Any thoughts on why this could be happening? I hadn't seen anyone > complain about the performance of 1208/09 yet, so it is probably > something with my machine :-(, but it beats me what it could be :-(.
This is the downside of db4o. If it is a widespread problem, we're gonna have to revert it. Which means throwing away more than 6 months work largely funded by Google's $18K. My database queue is usually pretty empty, even with queued downloads, but I have 8G and fast mirrored disks... On Tuesday 05 May 2009 03:22:06 Juiceman wrote: > > I see it too. I have my node installed on it's own separate disk yet > it stalls my quad-core system. It's as if the HDD controller chip is > so swamped with disk ops that the OS and other apps are starved for > resources. I'm seeing disk queues exceeding 500 vs the 40 my > workgroup server hits. > I also see Freenet using 2650 handles. That twice the next highest > app which is my antivirus and 10x the average app. Heck, even the > core Windows processes don't use more than 300-600 usually. > > Any recommendations for improving performance? Is BDB datastore any > better (more stable or less disk IO) than salted-hash? I could try > going back to that... 2650 handles is strange, on unix we are generally limited to 1024 and generally we don't exceed that. Both of your problems may be caused by flaky hardware, but frankly we do need to run on flaky real world hardware. :| No, changing the datastore won't help. On Tuesday 05 May 2009 10:30:56 Victor Denisov wrote: > > I see it too. I have my node installed on it's own separate disk yet > > it stalls my quad-core system. It's as if the HDD controller chip is > > so swamped with disk ops that the OS and other apps are starved for > > resources. I'm seeing disk queues exceeding 500 vs the 40 my > > workgroup server hits. > > Same here. Enormous disk queues. I've also compared i/o counts with i/o > bytes read/written - that's how I know that i/o operations are small. In > the statistics screen, I routinely see 100+ outstanding database jobs. > It can't be good. This just confirms that disk I/O is the problem ... and almost certainly caused by db4o as it goes away if nothing is queued. :( Opinions of other folk? Does this happen to you? Are there lots of complaints of similar problems on FMS? Is the recent increase in the performance of automated tests the result of the network shrinking due to lots of people uninstalling? -------------- next part -------------- A non-text attachment was scrubbed... Name: signature.asc Type: application/pgp-signature Size: 835 bytes Desc: This is a digitally signed message part. URL: <https://emu.freenetproject.org/pipermail/support/attachments/20090506/0909a573/attachment.pgp>
