dillon@CORY.BERKELEY.EDU (Matt Dillon) (07/28/88)
I solved the problems I had with my Adaptec 4000A (st506/412->SCSI). (I accidently had specified a 3ms non-buffered seek instead of a 17uS buffered seek). I am now running a hard disk system using that and a StarDrive SCSI controller slapped on the side of my StarBoard on the side of my A1000. Results are as follows: System: Duel ST4096 (80MB*2 = 160MB) w/Adaptec 4000A all in an enclosure, microbotics Starboard II & StarDrive. Completely Custom Setup. Max Throughput (massive read on trackdisk.device): 200K/sec (this is what I expect with the fast filesystem) DOS Throughput (DiskPerf program): (SLOW FILESYSTEM) File Creations, files/sec: 7 File Deletions, files/sec: 25 Directory Scan, entries/sec: 42 Seek+read/sec: 68 READ SPEEDS w/ varying buffer sizes, bytes/sec: 512 33608 bytes/sec 4096 43690 8192 44431 32768 43690 WRITE SPEEDS w/ varying buffer sizes, bytes/sec: 512 13582 bytes/sec 4096 14018 8192 13943 32768 14018 -Matt
scott@applix.UUCP (Scott Evernden) (07/30/88)
In article <8807281433.AA02114@cory.Berkeley.EDU> dillon@CORY.BERKELEY.EDU (Matt Dillon) writes: > > I am now running a hard disk system using that and a StarDrive SCSI >controller slapped on the side of my StarBoard on the side of my A1000. >Results are as follows: > >System: Duel ST4096 (80MB*2 = 160MB) w/Adaptec 4000A all in an enclosure, > microbotics Starboard II & StarDrive. Completely Custom Setup. Congrats. I think July was Hard Disk Month, since I and _several_ other Amigoids I know got disks this month (also StarDrives). Your performance figures are pretty good for the SFS; were you using the StarDrive's FASTMODE to achieve these Diskperfa results? I performed these experiments myself, and discovered that performance improved noticeably if I _decreased_ the number of buffers- I currently use 8. I also noticed that the WRITE speeds could almost be doubled by adjusting the interleave to 2 or 3. -scott
dillon@CORY.BERKELEY.EDU (Matt Dillon) (07/31/88)
>Your performance figures are pretty good for the SFS; were you using the >StarDrive's FASTMODE to achieve these Diskperfa results? Yes. >I performed these experiments myself, and discovered that performance >improved noticeably if I _decreased_ the number of buffers- I currently >use 8. I also noticed that the WRITE speeds could almost be doubled >by adjusting the interleave to 2 or 3. > >-scott Ditto, except I haven't tried reducing the # of buffers (30 now). Also, if you are using a Mountlist, make sure you've got 'Priority = 0' on all the StarDrive entries (I know I've said this before, but it makes *all* the difference in percieved smooth functioning). No more jerkyness to the Amiga... -Matt
page@swan.ulowell.edu (Bob Page) (08/01/88)
>performance improved noticeably if I _decreased_ the number of buffers
The slow file system in most cases takes longer to search through its
buffers than to just re-fetch the block from the drive (depending on
how many buffers you have) so in many cases the fewer the better, and
in ALL cases, there's a point where you slow yourself down as you add
buffers.
Fixed in the rumored Fast Filesystem.
..Bob
--
Bob Page, U of Lowell CS Dept. page@swan.ulowell.edu ulowell!page
"What a wonder is USENET; such wholesale production of conjecture from
such a trifling investment in fact." -- Carl S. Gutekunst
jdow@gryphon.CTS.COM (J. Dow) (08/02/88)
In article <8807281433.AA02114@cory.Berkeley.EDU> dillon@CORY.BERKELEY.EDU (Matt Dillon) writes: > > I solved the problems I had with my Adaptec 4000A (st506/412->SCSI). >(I accidently had specified a 3ms non-buffered seek instead of a 17uS >buffered seek). > > I am now running a hard disk system using that and a StarDrive SCSI >controller slapped on the side of my StarBoard on the side of my A1000. >Results are as follows: > >System: Duel ST4096 (80MB*2 = 160MB) w/Adaptec 4000A all in an enclosure, > microbotics Starboard II & StarDrive. Completely Custom Setup. > >Max Throughput (massive read on trackdisk.device): 200K/sec > (this is what I expect with the fast filesystem) > More likely you'll see figures running from 163k/sec to 187k/sec with the fast filesystem. (How'd you like the read speeds for small reads - ie old_fs reads? That was my major contribution to the code... There is a 2k read ahead buffer in there. It made a huge difference is speed. And it turns out the choice of 2k was fairly straight forward as well. It sounds like you went to a lot of unnecessary trouble to build yourself an adaptec formatter utility. There IS one available from MicroBotics that I developed (well, threw together) and Mike fine tuned. It handles 4000A's (I have one) and 4707's (I'm running one on an ST238. I like the extra storage. Not much speed difference, though.) There is a new driver available that handles the Adaptec 5500 as well. (That booger assumes you support SCSI disconnect unles syou tell it otherwise. Well, not SD tells it otherwise all the time as the non-interrupt driven SD controller cannot feasibly support reconnect. <sigh>) The real speedup will be from the HardFrame controller that I am testing here. I get at least 330k/sec ffs speeds. I suspect it should be higher but the controllers aren't cooperating right with 1:1 interleave. <sigh> That's testing with an ST277N on the HF. Alas it won't work on the A1000 or A500 without a new ZORRO adaptec. (And an aside t0 all youse guys with 4.0 rev motherboards - get new GARY and U808 chips for starters. Until i did that DMA on my board was hopeless.) (Oh - on the 4070 I also have my "chuckles" drive - an old ST604. Well - it IS another 8.3 megs. And it works just peachy if a tad slower due to the slow seek. - and re the 3ms seeks - I made that booboo first cut as well. Stupid board numbers faster seeks with higher numbers rather than lower which is more intuitive. Dumb!) {@_@} -- Sometimes a bird in the hand leaves a sticky deposit. Perhaps it were best it remain there in the bush with the other one. {@_@} jdow@bix (where else?) Sometimes the dragon wins. Sometimes jdow@gryphon.CTS.COM the knight. Does the fair maiden ever {backbone}!gryphon!jdow win? Surely both the knight and dragon stink both stink to high heaven.
jdow@gryphon.CTS.COM (J. Dow) (08/02/88)
In article <751@applix.UUCP> scott@applix.UUCP (Scott Evernden) writes: >In article <8807281433.AA02114@cory.Berkeley.EDU> dillon@CORY.BERKELEY.EDU (Matt Dillon) writes: >> >> I am now running a hard disk system using that and a StarDrive SCSI >>controller slapped on the side of my StarBoard on the side of my A1000. >>Results are as follows: >> >>System: Duel ST4096 (80MB*2 = 160MB) w/Adaptec 4000A all in an enclosure, >> microbotics Starboard II & StarDrive. Completely Custom Setup. > >Congrats. I think July was Hard Disk Month, since I and _several_ other >Amigoids I know got disks this month (also StarDrives). > >Your performance figures are pretty good for the SFS; were you using the >StarDrive's FASTMODE to achieve these Diskperfa results? > >I performed these experiments myself, and discovered that performance >improved noticeably if I _decreased_ the number of buffers- I currently >use 8. I also noticed that the WRITE speeds could almost be doubled >by adjusting the interleave to 2 or 3. > >-scott Based on his figures he is running with a 2:1 interleave with fastmode. Those are about as good as it can do. I have not tried decreasing the number of buffers. Thanks for the tip. (Moot for me now as I use FFS exclusively these days. Love it! - Oh - for 4070's use 3:1 for best speed...) {@_@} -- Sometimes a bird in the hand leaves a sticky deposit. Perhaps it were best it remain there in the bush with the other one. {@_@} jdow@bix (where else?) Sometimes the dragon wins. Sometimes jdow@gryphon.CTS.COM the knight. Does the fair maiden ever {backbone}!gryphon!jdow win? Surely both the knight and dragon stink both stink to high heaven.
peter@sugar.uu.net (Peter da Silva) (08/03/88)
> >Max Throughput (massive read on trackdisk.device): 200K/sec
M-m-m-max. Max Th-throughput?
Hello, Max.
Hello, H-h-hello...
--
Peter da Silva `-_-' peter@sugar.uu.net
Have you hugged U your wolf today?
scott@applix.UUCP (Scott Evernden) (08/03/88)
In article <5102@gryphon.CTS.COM> jdow@gryphon.CTS.COM (J. Dow) writes: >Based on his figures he is running with a 2:1 interleave with fastmode. Those >are about as good as it can do. I have not tried decreasing the number of >buffers. Thanks for the tip. (Moot for me now as I use FFS exclusively these >days. Love it! - Oh - for 4070's use 3:1 for best speed...) >{@_@} This was my best run (StarDrive to ST277N): $ cat disk.i0.b8.fast File create/delete: create 10 files/sec, delete 26 files/sec Directory scan: 48 entries/sec Seek/read test: 78 seek/reads per second r/w speed: buf 512 bytes, rd 42281 byte/sec, wr 13582 byte/sec r/w speed: buf 4096 bytes, rd 52428 byte/sec, wr 13653 byte/sec r/w speed: buf 8192 bytes, rd 53498 byte/sec, wr 13582 byte/sec r/w speed: buf 32768 bytes, rd 53498 byte/sec, wr 13582 byte/sec -scott