mobo@sphinx.uchicago.edu (Samuel Wilson) (09/26/87)
I know that this has come up before, but before I trashed my hard disk completely, I thought I would ask some advice. I have a 6300 (with a V30) with a 20 meg HD. The disk and controller came in a box that says EVEREX, but it looks like a Western Digital WX2 controller and a Seagate 225 drive. I am using ATT's version of MS-DOS 3.2. For some reason, this is the slowest setup I have ever seen. The CORE disk performance test gives it a performance rating of .735 with an average seek time of 93.6 ms. I also use applications with lots of very small files, and want to try a sector size smaller than the present 8K. I am going to do some experimenting with different interleaves and sector sizes, but have a couple questions. 1) does anyone have any suggestions about which interleaves to try first, and 2) how do you change the sector size with this combination of machine/DOS/controller/drive? If it is of any use, here is what the beginning of the current boot record looks like. EB429050 53412031 2E303400 02100100 0200043B A2F80800 11000400 11000000 000000CF 02250209 2AFF50F6 19040D0A 424F4F54 2076312E 30342000 20627920 53475300 33C08ED0 ... thanks, Samuel Wilson ..ihnp4!gargoyle!sphinx!mobo FOTMOBO@UCHMVS1.Bitnet University of Chicago, Division of Social Sciences
wtm@neoucom.UUCP (09/28/87)
There doesn't really seem to be a pat answer on what interleave factor is best. A lot depends on the applications programs as well as the underlying DOS. Sometimes a lousy interleave factor that results in poor general performance will speed up applications programs that do a lot of processing in between reading sectors from the disk. I know it is a pain, but trying several different interleave settings is worthwhile. In general, I've found that with the WX series (non RLL) controllers and Seagate 225s, an interleaf factor of 6 seems to work well with the 6300. You might try 5 or 7 as well. For RLL controllers, an interleaf factor of 3 or, perhaps, 4 seems to do pretty well. I noticed right away that DOS 3.2 seems slower than 3.1. Remember that the internal processing is greater for reading a given size file under 3.2. This is a consequence of the fact that cluster size is smaller. No matter waht you do, this'll get you. I don't mind slowing down a bit in return for the massive improvement in storage capacity. On my own 6300 at home, I had 5.5 meg free on a 30 meg drive with 3.1. After a backup and restore to the disk reformatted with 3.2, I had 11.3 meg free! Also, Lightning Disk Cache from Computer Support Group can make a big improvement is disk performance. We've tried one from Mace and Vcache from Golden Bow. Lightning provided the biggest benefit with the least memory requirement. By the way, Mace does market an interleave testing program, but we don't have it here yet. Also there is a P/D interleave tester available from one of the P/D distributors in Indiana (forget the name at the moment, and the catalog is at home). I think you can get information by calling 1-800-IBM-DISK. What does irk me is the $$##%@&! clock bug introduced into DOS 3.2. Each time the system is warmbooted, the dag-gone IBMDOS.COM gets the T.O.D. from the battery-backed clock. In the process, it sets the seconds register of the battery-backed clock to zero! On average, your battery clock is destined to loose 30 sec. each time you reboot.-- the amount lost obviously depends on when in a given minute you press the reset button (or CTL-ALT-DEL). When I get a chance, I'll see if I can find the offending code and develop a patch. This was *not* a bug under previous DOS versionns. Bill Bill Mayhew, Electrical Engineering Division of Basic Medical Sciences Northeastern Ohio Universities' College of Medicine Rootstown, OH 44272-9989 USA phone: 216-325-2511 (wtm@neoucom.UUCP ...!cbsogd!neocom!wtm)
dpbaudra@ihlpe.ATT.COM (Braune) (10/01/87)
In article <710@neoucom.UUCP> wtm@neoucom.UUCP (Bill Mayhew) writes: > >There doesn't really seem to be a pat answer on what interleave >factor is best. A lot depends on the applications programs as well.... > >In general, I've found that with the WX series (non RLL) >controllers and Seagate 225s, an interleaf factor of 6 seems to >work well with the 6300........... > >For RLL controllers, an interleaf factor of 3 or, perhaps, 4 seems >to do pretty well.......... > >I noticed right away that DOS 3.2 seems slower than 3.1. Remember >that the internal processing is greater for reading a given size..... > >What does irk me is the $$##%@&! clock bug introduced into DOS 3.2. >Each time the system is warmbooted, the dag-gone IBMDOS.COM gets >the T.O.D. from the battery-backed clock. In the process, it sets....... > >Bill > >Bill Mayhew, Electrical Engineering >Division of Basic Medical Sciences >Northeastern Ohio Universities' College of Medicine >Rootstown, OH 44272-9989 USA phone: 216-325-2511 >(wtm@neoucom.UUCP ...!cbsogd!neocom!wtm) I've been meaning to post this info for some time particularly each time I see the question appear. I, too, wanted to know if my St225 and WD WX2 were performing at their optimum. A friend gave me some P/D software that first checked the number of revolutions to read a track and then the data xfer rate. Another pgm actually tried to determine the optimum interleave, trying factors from 1 to 16. At the end of this posting I've included the data I accumulated during the process; it includes stopwatch timing of some common pgm loads from hard disk as well as some of the analysis pgm results. For the ST225/WD WX2 combo, I found 6 to be the best interleave and that has been confirmed other places on the NET. I also have another 6300 with a WD RLL cntrlr and a Miniscribe 3425 that I formatted as a 30 MB drive (I fully recognize what I have done, understand some or all of the potential dangers, and use the machine in a totally non-critical application); there the best interleave was 7 and the optimizer pgm failed to give good results which meant a lot of tedious test-reformat-test cycles. End result as shown is a healthy increase in the speed of disk I/O operation and a noticeable speed of access to pgms such as LOTUS 1-2-3 (admittedly when you start in the subbasement even the 2nd floor looks high). I have since retrieved some good discussions on interleave, low level format, WD BIOS versions, and step rate suggestions from Ed Hopper's BBS in Houston, TX (1-713-782-5474). I also downloaded a pgm called CORETEST which sounds like a common piece of SW for these tasks. I'm sorry I have no before/after results for CORETEST. I will be trying an 18msec step rate on the ST225 as soon as I get some time. Info I retrieved says one needs the WD T.8 BIOS or later to change all the low level format params. I've included the results of reformatting with DOS 3.2 from 3.1; 15.3% space saving. Hopper's BBS has file containing a patch for the BIOS/clock problem. For what its worth, here's my data: TABLE 1 ------- SPINTEST and HOPTIMUM Results for ST225 hdisk(20MB) and WD Cntlr Interleave Read Revs to Data factor Time(sec) Read 1 trk Rate(B/sec) =========== ========== ========== =========== 1 14 2 15 3 16 20 21000 <=== Org lo lvl frmt 4 16 5 13 16 32640 6 6 6 87040 4.14 x Data rate 7 7 7 74596 8 9 10 10 52224 11 12 12 43520 13 14 15 16 16 32640 TABLE 2 ------- Load time tests for ST225 hdisk(20MB) and WD Cntlr for IL=3 and IL=6 Interleave=3 Interleave=6 SPINTEST=20 SPINTEST=6 Date rate=21000 Date rate=87040 Load Load Percent Program Time(sec) Time(sec) Decrease =========== ========== =========== ========= L123 12.5 7.8 37.6% dBASE III+ 10.0 6.1 39.0% WordPerfect 15.0 7.2 52.0% Multiplan 7.3 3.7 49.3% CHKDSK(exec time) 13.6 * 16.6 @ 4.2%Increase PC/VI 3.3 GEM 11.3 Spintest(exec time) 28.5 L123 file1(75490B) 11.7 L123 file2(122269B) 19.1 Size(Bytes) Exec rate ---------- ---------- * 16670720 1225788.2 Bytes/sec @ 21204992 1277409.1 Bytes/sec TABLE 3 ------- Results of ST-225 hdisk(20MB) Reformat MS-DOS 3.1 to 3.2 MS-DOS 3.1 MS-DOS 3.2 Virtual Single C/D drives C drive Space Percent (4096/clus) (2048/clus) Gain Gain ========== ========== =========== ========= Avail bytes C 3141632 9078784 Avail bytes D 4730880 ---------- ---------- ----------- --------- Total bytes 7872512 9078784 1206272 15.3% TABLE 4 ------- SPINTEST and HOPTIMUM Results for Miniscribe 3425 hdisk(30MB) and WD RLL Cntlr Interleave Read Revs to Data factor Time(sec) Read 1 trk Rate(B/sec) ================= ========= ========= ========= 1 14 2 24 3 25 4 15 19 27472 <=== Org lo lvl frmt 5 16 6 14 17 30719 1.12 x Data rate 7 15 5 104448 3.80 x Data rate 8 6 87040 3.17 x Data rate 9 6 87040 3.17 x Data rate 10 6 87040 3.17 x Data rate 11 12 8 65280 2.38 x Data rate 13 14 15 16 10 52224 1.90 x Data rate TABLE 5 ------- Load time tests for Miniscribe 3425 hdisk(30MB) and WD RLL Cntlr for IL=3 and IL=6 Interleave=3 Interleave=7 SPINTEST=19 SPINTEST=5 Date rate=27472 Date rate=104448 Load Load Percent Program Time(sec) Time(sec) Decrease ================= ========= ========= =============== L123 11.5 6.7 41.7% dBASE III+ 9.5 5.5 42.1% WordPerfect 13.2 5.8 56.1% Multiplan 6.8 3.2 52.9% CHKDSK(exec time) 16.2 * 14.4 @ 11.1% PC/VI 5.3 2.9 45.3% GEM 15.2 9.9 34.9% Spintest(exec time) 70.8 21.5 69.6% L123 file1(75490B) 11.2 11.2 0.0% L123 file2(122269B) 18.4 18.4 0.0% Size(Bytes)Exec rate --------- --------- * 32428032 2001730. Bytes/sec @ 32428032 2251946. Bytes/sec Dave Braune AT&T Naperville, IL 60566 ihnp4!ihlpe!dpbaudra