garys@houxm.UUCP (#G.SEUBERT) (10/06/84)
Several sources have come up with varying opinions about the optimal Gap Size and Blocks/cylinder for DECs RA81 drives. We're under the assumption of 728 Blocks per cylinder (14 R/W heads x 52 sectors/track). If 728 is correct, what should the proper gap size be considering 3600 RPM? Thanks in advance. Gary Seubert - HOCC [..ihnp4!]houxm!garys (running System V if it makes a difference)
lou@AEROSPACE.ARPA (10/08/84)
From: Lou Nelson <lou@AEROSPACE.ARPA> >Return-Path: <ihnp4!houxm!garys@UCB-VAX.ARPA> >Date: 6 Oct 84 5:13:30-PDT (Sat) >To: Unix-Wizards@BRL-VGR.ARPA >From: ihnp4!houxm!garys@UCB-VAX.ARPA >Subject: 'mkfs(1m)' for RA81s - gap? > >Article-I.D.: houxm.926 > >Several sources have come up with varying opinions about the optimal >Gap Size and Blocks/cylinder for DECs RA81 drives. We're under the >assumption of 728 Blocks per cylinder (14 R/W heads x 52 sectors/track). > >If 728 is correct, what should the proper gap size be considering >3600 RPM? > >Thanks in advance. > >Gary Seubert - HOCC >[..ihnp4!]houxm!garys > >(running System V if it makes a difference) I'm not really a wizard, but here goes anyway. It does make a difference that you are running System V, since unlike 4BSD I think only 1 block is read in at a time. So your 728 is probably correct. The "gap size" is the important number however and, using a program posted by mh3bs!agp on 11/19/83 (who in turn got it from a USENIX paper), I measured 9 as the value that gave the best performance on my RA81s before I kicked them out the door in favor of Eagles. If you wish, assuming this reaches you, I can send you my modification of that program that includes RA81s and Eagles and you can run some experiments to determine which numbers are best for your configuration. Regards, Lou
smb@ulysses.UUCP (Steven Bellovin) (10/25/84)
One important point to remember -- a gap that is slightly below optimum is *much* worse than a gap that's somewhat above optimum. Round up if there's any doubt.