home *** CD-ROM | disk | FTP | other *** search
- Newsgroups: comp.protocols.nfs
- Path: sparky!uunet!haven.umd.edu!darwin.sura.net!spool.mu.edu!think.com!ames!biersch!aga
- From: aga@Comtech.com (Alan G. Arndt)
- Subject: Re: Sun PC-NFS performance (again)
- Message-ID: <1992Dec15.015951.20329@Comtech.com>
- Organization: Comtech Labs Inc, Palo Alto
- References: <1fooqcINN52p@seven-up.East.Sun.COM> <1992Dec7.173718.12792@Comtech.com> <1g3673INNa64@seven-up.East.Sun.COM>
- Date: Tue, 15 Dec 1992 01:59:51 GMT
- Lines: 146
-
- I did some more testing with PC-NFS (4.0a) this weekend. As everyone
- seems to be harrasing me about how my SS1+ really wasn't a server and
- I should get a REAL machine I went to a friend site with my pc. I
- don't see how you can't call a SS1+ a capable machine when it is 4+
- times the machine of a 386 server running Novell. Yet it only
- performs at 1/4 the speed! Anyhow I connected up to a 4/670-140-64.
- Something of a REASONABLE server class machine wouldn't you say?
- Well suprise suprise I got EXACTLY the same performance, a paultry
- 277.7 KB/sec! And in fact on HIS SS1+ I got 294.2 KB/sec. The only
- reason I can figure his SS1+ was faster is it was on the SAME tap on
- the E-NET v.s. the 4/670 being 50+ Meters away through two repeaters.
-
- The fully gorry details of the performance numbers are as follows and
- all these performance numbers are on an IDLE, or nearly so Server AND
- LAN:
-
- Read/Write Performance Servers
- In KBytes/sec Ours Friends Machines
- SS1+ w/40MB 4/670 SS1+ w/24Meg
- Clients 1.3 GB, 5MB/sec 1.3 GB, 5MB/sec 440 Mb, 5MB/sec
- SunOS 4.1.1 SunOS 4.1.3 (STOCK) SunOS 4.1.1
-
- 486-DX50
- SMC8013 277.7 / 166.7 277.7 / 18.4 294.1 / 14.2
- 3C509 277.7 / 166.7 277.7 / 18.4 294.1 / 14.2
-
- 486-DX33
- SMC8013 277.7 / 166.7
-
- 486-DX25
- SMC8013 277.7 / 166.7
-
- 386-DX33
- SMC8013 277.7 / 166.7
- 3C501 160 / 90
- 3C503 160 / 90
-
- SS1+
- File Copies 'cat big || cat > /dev/nul' where big is 10MB
- 14.9 Seconds 15.8 Seconds
- 671 KB/Sec 632 KB/Sec
-
- All PC benchmarks are done with Norton Utilities 6.0 SI Network disk
- test. The same results are also obtained with a simple PC copy to
- nul: (within a few percent, i.e. 268 KB/sec by hand on my stopwatch).
-
- The WD8013 Cards obtain the same performance with a PC-NFS specific
- driver AND the NDIS drivers. The PC-NFS specific drivers are smaller
- then the NDIS drivers for memory usage. The 3C509 only works with a
- NDIS driver.
-
- Do we notice a PATTERN here? It CAN'T go faster then 277.7 KB/sec
- except in one circumstance which is belived to be because of the
- closeness of the machine.
-
- So the PC isn't the limiting factor, the PC's card isn't the limiting
- factor, the network isn't the limiting factor and the SERVER ISN'T
- the limiting factor, the protocol isn't the limiting factor, in fact
- the LOW Level drivers aren't even the limiting factor. So that only
- leaves ONE item, PC-NFS itself.
-
- Now that I have shown all the data I have I will go into what we
- found out. We then used etherfind to watch the packets going across.
- Years ago (5+) with PC-NFS V2.x I watched this transaction and I
- could swear it the Sun was sending 8KB blocks in 5 packet bursts and
- the pc would receive them and request another block. HOWEVER,
- yesterday we noticed that the PC was only requesting 1K blocks (1166)
- bytes. The request was a healthy 218 bytes as well. When one
- starts to add up the requests and individual 1K bytes I can belive
- that the performance is ONLY 277 KB/sec and that the machines are
- doing their BEST the get that. That is why there is a noticeable
- difference between 50 Meters away and a machine RIGHT next to it's
- client.
-
- We also noticed that the PC was only creating WRITES of apparently
- 512 bytes (734).
-
- As can be seen bye the last entry the other Suns certainly use the
- block size of 8K to the fullest advantage and of course get very
- acceptable performance.
-
-
- raritan = 486-50 PC-NFS
- durable = 4/670
- certes = 4/65
- picasso = 4/65
-
- lnth proto source destination src port dst port
- Reads
- 12.66 218 udp raritan durable 127 2049
- 12.66 1166 udp durable raritan 2049 127
- 12.66 218 udp raritan durable 127 2049
- 12.67 1166 udp durable raritan 2049 127
- 12.67 218 udp raritan durable 127 2049
- 12.67 1166 udp durable raritan 2049 127
- 12.67 218 udp raritan durable 127 2049
- 12.67 1166 udp durable raritan 2049 127
-
-
- Writes
- 9.62 734 udp raritan certes 127 2049
- 9.65 138 udp certes raritan 2049 127
- 9.65 734 udp raritan certes 127 2049
- 9.68 138 udp certes raritan 2049 127
- 9.68 734 udp raritan certes 127 2049
- 9.72 138 udp certes raritan 2049 127
- 9.72 734 udp raritan certes 127 2049
- 9.75 138 udp certes raritan 2049 127
- 9.75 734 udp raritan certes 127 2049
- 9.78 138 udp certes raritan 2049 127
-
-
- Reads
- 1.16 194 udp picasso durable 1021 2049
- 1.17 1514 udp durable picasso 2049 1021
- 1.17 *1514 udp
- 1.17 *1514 udp
- 1.17 *1514 udp
- 1.17 *1514 udp
- 1.17 * 934 udp
- 1.17 194 udp picasso durable 1022 2049
- 1.18 1514 udp durable picasso 2049 1022
- 1.18 *1514 udp
- 1.18 *1514 udp
- 1.18 *1514 udp
- 1.18 *1514 udp
- 1.18 * 934 udp
-
-
- Now we looked at the manuals and the only option I found was TSIZE
- which WAS set at 8Kbytes. The TSIZE is supposed to be only for
- writes but as there is only one option I assumed it might also be
- used for the read request size. It is however obvious that PC-NFS
- pay's NO ATTENTION to this parameter and it wasting a tremendous
- amount of time dealing with small block sizes.
-
- So what can be done to PC-NFS to get it to use the LARGE block sizes?
- Is there some parameter I have missed? The SMC8013 cards have
- buffers of 16K on the board and certainly could handle 4K blocks if
- not full 8K blocks. The 3C509 cards will stream the data off the
- card as it comes in so the block size is irrelavent. Obviously the
- 3C501 and 3C503 cards have small buffers and need to continue to
- operate in their current configuration.
-
- -Alan Arndt
- aga@Comtech.com
-