home *** CD-ROM | disk | FTP | other *** search
- Comments: Gated by NETNEWS@AUVM.AMERICAN.EDU
- Path: sparky!uunet!gatech!paladin.american.edu!auvm!MSUS1!LAVENDA
- X-Envelope-to: notabene@taunivm.BITNET
- X-VMS-To: IN%"notabene@taunivm"
- Message-ID: <01GNJ2NPHMCW0008ND@MSUS1.MSUS.EDU>
- Date: Thu, 13 Aug 92 10:21:00 CST
- Sender: Nota Bene List <NOTABENE@TAUNIVM>
- From: LAVENDA@MSUS1
- Subject: Cache tests
- Newsgroups: bit.listserv.notabene
- Lines: 47
-
- As part of my general approach to serious writing projects is finding other
- projects to keep me from having to write, I've just spent a few minutes
- (really! only a few minutes. Not more than an hour....) testing disk caches.
- I tested Hyperdisk, PC-Cache, Ncache (from Norton), and Combi (which, as
- many of you know is a combination RAM-disk and disk cache). I used both the
- Coretest program (v. 2.92) and Cache Test v4.2 (v4.3 available on SIMTEL
- and its mirrors is defective as downloaded). The latter is a very thorough
- test for disk caches. The former is a hard disk speed test that also is
- differentially affected by different speed caches.
-
- I set up each cache in its default configuration, with 1MB as the cache
- size (this is the minimum Hyperdisk size). The Coretest scores gave a
- significant edge to Hyperdisk, with a performance index of over 800. This
- compares to a PI of over 80 for PC-Cache, 79 for Ncache, and 67 for Combi.
- But I think Cache Test is a better measure of how caches work.
-
- Cache test measures sequential disk access (which is more common in wordprocessi
- ng, where information is sequential in the file) and rando disk access (more
- common in databases, where records accessed may be anywhere in a file--that
- should be "random", not "rando"). It creates a files of 320,000 characters,
- writes it, and then reads it sequentially, and randomly in several different
- ways. A summary of results follows:
-
- PC-Cache Hyperdisk Ncache Combi
- Seq. write 15.94 sec 23.94 15.72 24.22
- Seq. read 16.00 sec 16.44 16.00 16.67
- Total sequential 31.94 40.39 31.72 40.89
- Random write 15.5 23.83 15.39 23.94
- Read x 10 1.00 1.00 1.00 1.00
- Writ x 10 1.11 11.50 1.17 11.50 [not typos]
- Random R/W 70.39 216.11 70.22 244.22
-
- [Note: the Read and Write x 10 reads and updates every 10th record in
- the file]
-
- It seems that the major differences between PC-Cache and Ncache on the one
- hand, and Hyperdisk and Combi on the other are in the Write speeds. It
- may be possible to adjust the writes on Hyperdisk, but as noted above, it
- was my intention to test default setups. Machine specifics: 386DX-33 clone,
- IDE controller, Seagate 107MB drive with 15 millisecond average access time.
-
- This may be of some value to someone. And now, back to Minnesota festivals
- and the carnivalesque....
-
- Cheers,
- Robert Lavenda
- lavenda@msus1.bitnet
-