Return to BSD News archive
Path: euryale.cc.adfa.oz.au!newshost.carno.net.au!harbinger.cc.monash.edu.au!news.rmit.EDU.AU!news.unimelb.EDU.AU!munnari.OZ.AU!uunet!in2.uu.net!192.207.105.50!prodigy.com!darkstar.prodigy.com!davidsen From: davidsen@tmr.com (bill davidsen) Newsgroups: comp.os.linux.misc,comp.os.linux.networking,comp.os.linux.setup,comp.unix.bsd.misc,comp.os.ms-windows.nt.advocacy,comp.os.os2.advocacy Subject: Re: Linux vs whatever Date: 5 Feb 1997 23:17:05 GMT Organization: TMR Associates, Schenectady NY Lines: 44 Message-ID: <5db4dh$38gi@usenet1y.prodigy.net> References: <32DFFEAB.7704@usa.net> <5cphaj$qvg@cynic.portal.ca> <5d7rtu$ao9@rzstud2.rz.uni-karlsruhe.de> <5d81q1$8ls@cynic.portal.ca> NNTP-Posting-Host: darkstar.prodigy.com Originator: davidsen@darkstar.prodigy.com Xref: euryale.cc.adfa.oz.au comp.os.linux.misc:156280 comp.os.linux.networking:67332 comp.os.linux.setup:95585 comp.unix.bsd.misc:2261 comp.os.ms-windows.nt.advocacy:51951 comp.os.os2.advocacy:265810 In article <5d81q1$8ls@cynic.portal.ca>, Curt Sampson <cjs@cynic.portal.ca> wrote: | Try an NFS test with 8 KB blocks, and you'll see just how broken | the TCP/IP implementation is. There's a very good reason that Linux | uses 1 KB blocks for NFS; the stack will crawl on 8 KB blocks | because it has to do an extra copy of all the data every time it | reassembles the fragments. (I described this in detail in another | post.) Okay: darkstar:/etc# l /mnt -rw-r--r-- 1 65534 1474560 Nov 6 13:57 testdev darkstar:/etc# time sum /mnt/testdev 00000 1440 0.35user 0.99system 0:08.66elapsed 15%CPU (0avgtext+0avgdata 0maxresident)k 0inputs+0outputs (0major+0minor)pagefaults 0swaps darkstar:/etc# umount /mnt darkstar:/etc# mount -r -t nfs newsppc2-int:/usr/news/misc /mnt darkstar:/etc# time sum /mnt/testdev 00000 1440 0.34user 0.42system 0:28.45elapsed 2%CPU (0avgtext+0avgdata 0maxresident)k 0inputs+0outputs (0major+0minor)pagefaults 0swaps darkstar:/etc# Your point is well taken about the CPU time, although this system is not blindingly fast (P5-90). The connection is 16Mbit TokenRing, I would expect to see better from ethernet. But for all of that, the speed is notably higher, 166kB vs. 50.6kB, over a WAN which I believe included 60-70 miles of T1, three bridges, two routers (and a partridge in a pear tree). I would love to try this over a short, unloaded, well controlled network, but I sure don't have one handy. I did one test on a machine on my ring, but it has a load average in double digits and was slower than the machine in test one, which is totally unloaded. Let's say that the NFS runs faster with larger buffers, as expected, and that CPU use is excessive as claimed. Not to the point that I would hesitate to use large buffers, though! -- -bill davidsen (davidsen@tmr.com) "As a software development model, Anarchy does not scale well." -Dave Welch