Article 39272 of comp.sys.amiga.programmer:
Newsgroups: comp.sys.amiga.programmer
Path: news.csos.orst.edu!flop.ENGR.ORST.EDU!gaia.ucs.orst.edu!umn.edu!spool.mu.edu!agate!library.ucla.edu!europa.eng.gtefsd.com!uunet!majipoor.cygnus.com!fnf
From: fnf@cygnus.com (Fred Fish)
Subject: CD-ROM filesystem performance
Message-ID: <CFHAHL.8KK@cygnus.com>
Organization: Cygnus Support, Mountain View, CA
Date: Tue, 26 Oct 1993 00:24:55 GMT
Lines: 57


I'm curious to know if anyone has done any CD-ROM benchmarking.  One
of the goals of my CD-ROM distribution is to provide a set of tools
(typically large things like gcc, TeX, etc) which can be run directly
off the CD-ROM, thus freeing up a significant amount of hard drive
space (potentially hundreds of megabytes).  In order for this to be
practical, it has to be possible to run the tools without excessive
slowdown relative to running them off a hard drive.

Now that I've got test CD-ROM's to play with while I wait for the
production CD-ROM's to arrive, I've tried some quickie performance
tests to see how practical that goal is.  One of these tests includes
reading a subset of the CD-ROM file tree and verifying the checksums
with brik.  This is essentially a test of how quickly we can read
data off the source tree and process it.  Since I'm on a 4000 I
suspect (though haven't yet tested it) that the I/O time should
dominate for this test.  The source tree for this test is the
entire GNU source tree on the CD-ROM, some 3871 files of various
size, totalling 72Mb.

The results of this test are interesting:

	CD-ROM Software				Elapsed Time
	-------------------------------		------------

	"Brand -X" (under NDA, can't identify)	0 hr 15 min
	Asimware Innovations AsimCDFS V2.0	1 hr  2 min
	Xetec CDxFileSystem V90			1 hr  6 min
	AmiCDROM V1.2 (freely redistributable)	7 hr 36 min

By contrast, the time to process the same tree when the tree is
resident on a Fujitsu 1Gb, 10ms drive, on the same bus, with the same
controller, is 9 minutes.  So the "Brand X" at least is *very*
comparable.  Note also that I haven't yet spent any time trying to
"tune" the various drivers, so the above comparisons may not
accurately reflect the relative performances of each driver under
optimal conditions.

Other hardware details are:

	A4000/040
	Fastlane SCSI-II controller
	Toshiba 3401B CD-ROM drive

One last interesting note.  The CD-ROM contains a crclist of all of
the files that are on the CD-ROM except for the crclist itself of
course.  The CD-ROM was mastered using Commodores software to generate
an ISO-9660 image.  Using this crclist to verify the contents of the
test CD-ROM turned up problems in everyone's CD-ROM filesystem code
except for Xetec.  The Xetec code was the only software that was able
to correctly read every file on the CD-ROM.  Typical problems were
failure to find files that had international characters in their
filenames, or that had semicolon characters in their filenames.  Some
brands worked fine on semicolons as long as it wasn't the first
character in the filename.  Etc, etc.

-Fred