00:48:17 *** Kyle_AmesLa1 has quit ("Leaving.")
01:08:01 *** Kyle_AmesLa1 (n=kschoche@ohm.scl.ameslab.gov) has joined #pvfs2
01:15:33 *** Kyle_AmesLa1 has parted #pvfs2 ()
01:32:33 *** yfttyfs (n=yf@125.33.200.68) has joined #pvfs2
05:09:49 *** yfttyfs has quit (Remote closed the connection)
07:28:51 *** yfttyfs (n=yf@123.112.17.44) has joined #pvfs2
07:57:04 *** yfttyfs has quit ("Leaving.")
07:58:56 *** yfttyfs (n=yf@123.112.17.44) has joined #pvfs2
08:08:31 *** yfttyfs has quit (Remote closed the connection)
08:53:04 *** _icebon4 (n=icebone@pD9E34D11.dip.t-dialin.net) has joined #pvfs2
09:07:46 *** _icebon4 has quit (Remote closed the connection)
09:08:17 *** _icebon4 (n=icebone@pd9e34d11.dip.t-dialin.net) has joined #pvfs2
09:09:39 *** _icebon4 has quit (Remote closed the connection)
09:10:13 *** _icebon4 (n=icebone@pd9e34d11.dip.t-dialin.net) has joined #pvfs2
09:10:40 *** _icebon4 has quit (Remote closed the connection)
09:11:12 *** _icebon4 (n=icebone@pd9e34d11.dip.t-dialin.net) has joined #pvfs2
09:14:00 *** Kyle_AmesLab (n=kyle@12-216-6-144.client.mchsi.com) has joined #pvfs2
13:28:32 *** Kerrek (n=chatzill@c-71-196-234-180.hsd1.co.comcast.net) has joined #pvfs2
13:31:45 *** pcarns (n=pcarns@adsl-219-143-139.asm.bellsouth.net) has joined #pvfs2
13:41:22 *** yfttyfs (n=yf@125.33.200.164) has joined #pvfs2
15:10:45 *** yfttyfs has quit (Remote closed the connection)
19:37:18 *** |icebone| (n=icebone@pD9E34C65.dip.t-dialin.net) has joined #pvfs2
19:53:13 *** _icebon4 has quit (Read error: 110 (Connection timed out))
20:14:43 <Kyle_AmesLab> is Alt_AIO server-side only?
20:35:07 <pcarns> yep
20:35:23 <pcarns> that just toggles how the server does its reads and writes
20:35:53 <Kyle_AmesLab> pcarns: any idea where we can pick up the AIO branch?
20:36:13 <Kyle_AmesLab> pcarns: we finally got everything working, and cant scale above 8 proc's..
20:36:56 <pcarns> Kyle_AmesLab: not sure actually, I know it is in trunk but I don't know which releases have it
20:37:11 <Kyle_AmesLab> pcarns: ahh, okay. thanks 
20:37:18 <pcarns> Kyle_AmesLab: you have to use the TroveMethod config option to turn it on or off
20:39:19 <Kyle_AmesLab> pcarns: its not on by default
20:39:21 <Kyle_AmesLab> pcarns:  ??
20:41:30 <pcarns> Kyle_AmesLab: no, as far as I know it uses the "normal" dbpf method by default
20:41:37 <robl> fs.conf
20:48:43 *** _icebone_ (n=icebone@pD9E3508B.dip.t-dialin.net) has joined #pvfs2
20:50:34 <Kyle_AmesLab> robl: do I have to set it in both StorageHints as well as Default?
20:54:29 *** |icebone| has quit (Read error: 60 (Operation timed out))
20:58:50 <robl> Kyle_AmesLab: http://www.pvfs.org/pvfs2-options.html#TroveMethod
21:00:52 <Kyle_AmesLab> robl: I was looking at that
21:00:58 <Kyle_AmesLab> robl: it says I can pick either
21:01:12 <slang> Kyle_AmesLab: set it in StorageHints
21:01:16 <Kyle_AmesLab> robl: but not whether or not I have to put in default if I also want it in storage
21:01:33 <slang> Kyle_AmesLab: its not inherited
21:02:42 <Kyle_AmesLab> slang: awesome, thanks
21:07:38 *** _icebone_ has quit (Read error: 110 (Connection timed out))
21:23:50 <hozer> hello all
21:23:59 * slang  waves
21:24:14 <Kyle_AmesLab> heya
21:24:21 <Kyle_AmesLab> bout time you woke up hozer
21:24:31 <hozer> I am trying to write a proposal 
21:24:43 <Kyle_AmesLab> blah, write a proposal that fixes my tooth, I just chipped it
21:24:47 <hozer> doh
21:24:54 <hozer> I'm trying to get on the bioeconomy gravy train
21:25:43 <hozer> so, there is native linux AIO in the trunk?
21:26:22 <Kyle_AmesLab> in A branch, I believe
21:26:43 <hozer> Kyle_AmesLab: brett and I were also talking about having Brett's shim layer do async pvfs operations, as well as using a single file instead of 1 file per process
21:27:20 <Kyle_AmesLab> hozer: bbode mentioned something like that
21:30:00 <Kyle_AmesLab> hozer: but I didnt follow up on it
21:30:14 <Kyle_AmesLab> hozer: he didnt say anything about the single file thing tho
21:31:43 <hozer> I was going to try to figure out how to do pvfs async in NetPIPE
21:46:13 <Kyle_AmesLab> that sounds icky
21:47:25 <robl> it shouldn't be too bad
21:47:37 <robl> we don't have a lot of async users, so that would be a great test
21:47:49 <robl> our model is a lot like MPI
21:58:00 <Kyle_AmesLab> upc all the way
22:02:13 <Kyle_AmesLab> hozer: I put aio hints onto those servers, run 15,000 clients and tell me if its faster
22:11:22 <robl> tell us too
22:11:42 <robl> it should be, iven what we know of glibc aio
22:11:43 <robl> (assuming you are writing to a single file)
22:11:56 *** pcarns has parted #pvfs2 ()
22:16:00 <hozer> heh
22:16:14 <robl> Kyle_AmesLab: you do much work with upc?
22:16:16 <hozer> robl: any docs I should start with ? 
22:16:30 <hozer> (for the async stuff)
22:16:36 <robl> Kyle_AmesLab: you and darius can hang out when you are at argonne and talk about it
22:17:48 <robl> hozer: pvfs async?  heh
22:17:51 <robl> hozer: http://www.pvfs.org/doxygen/group__sysint.html
22:17:57 <Kyle_AmesLab> robl: hah, no I dont I just heard about it from  a collegue
22:20:39 *** Kerrek has parted #pvfs2 ()
22:24:48 <robl> hozer: the biggest users of the async interface are the syncronous calls (they call the async and block) and pvfs2-client-core
22:30:04 <hozer> heh, okay
23:00:50 <slang> 16:25 < hozer> so, there is native linux AIO in the trunk?
23:00:50 <slang> 16:26 < Kyle_AmesLab> in A branch, I believe
23:01:36 <slang> hozer: native linux aio has the same interfaces as glibc aio
23:02:16 <slang> hozer: so the pvfs support is there if the kernel/libaio support is there
23:02:35 <slang> Kyle_AmesLab: I don't know of any branches that are aio specific
23:08:09 <Kyle_AmesLab> slang: I swear someone said it existed, or was in a branch, in fact, the mailing list had someone doing tests on it at one point.
23:09:04 <slang> Kyle_AmesLab: sure
23:09:15 <slang> Kyle_AmesLab: but that doesn't require any code mods to pvfs
23:09:39 <slang> Kyle_AmesLab: if you compile kernel support for aio and install libaio, pvfs servers will use it automatically
23:13:45 <Kyle_AmesLab> slang: hrm.
23:14:08 <slang> Kyle_AmesLab: no good?
23:14:15 <Kyle_AmesLab> hozer: we had that bulposix aio port at one point..
23:14:24 <Kyle_AmesLab> slang: not sure what that means for client side stuff?
23:14:34 <Kyle_AmesLab> slang: we want to overlap computation with reads
23:14:52 * slang nods
23:14:57 <Kyle_AmesLab> slang: we only get likie 80% cpu utilization right now, and want higher heh
23:15:02 <slang> Kyle_AmesLab: on the client you're not going throught the vfs though, right?
23:15:15 <Kyle_AmesLab> slang: no.
23:15:16 <slang> Kyle_AmesLab: so you can just the PVFS_isys_* interfaces
23:15:32 <slang> Kyle_AmesLab: that's what robl was referring to I think
23:15:42 <slang> Kyle_AmesLab: that gives you the async functionality you're looking for
23:15:57 <Kyle_AmesLab> slang: bbode wrote a wrapper for our jobs so it uses libpvfs2
23:24:31 <slang> Kyle_AmesLab: cool
23:53:38 *** Kyle_AmesLab has quit (Read error: 60 (Operation timed out))