home *** CD-ROM | disk | FTP | other *** search
- Path: sparky!uunet!dtix!darwin.sura.net!zaphod.mps.ohio-state.edu!menudo.uh.edu!rosie.uh.edu!st1my
- From: st1my@rosie.uh.edu (Stich, Christian E.)
- Newsgroups: comp.sys.amiga.hardware
- Subject: Re: Hard disk throughput speed ?
- Message-ID: <9SEP199209193155@rosie.uh.edu>
- Date: 9 Sep 92 14:19:00 GMT
- References: <1992Sep8.171338.2492@Rapnet.Sanders.Lockheed.Com> <copes.716015341@marsh>
- Sender: usenet@menudo.uh.edu (USENET News System)
- Organization: University of Houston
- Lines: 53
- News-Software: VAX/VMS VNEWS 1.41
- Nntp-Posting-Host: rosie.uh.edu
-
- In article <copes.716015341@marsh>, copes@cs.curtin.edu.au (Simon Cope) writes...
- >msnider@Rapnet.Sanders.Lockheed.Com (Marc A. Snider x0193) writes:
- >
- >>What kind of throughput can I expect from a stock A3000? How about the
- >>numbers for a GVP combo accelerator (do they make an '040 based combo)
- >>using their scsi interface and an aftermarket drive? Finally, what is
- >>the fastest throughput possible today on an A3000 and by what means?
- >
- >>Thanks,
- >>M. Snider
- >>msnider@rapnet.sanders.lockheed.com
- >
- >My A3000 25/50 with 6 megs fastram runs around 850K/sec to 1M/sec depending
- >on disk usage and fragmentation. Thats set up with 100 buffers and the stock
- >Quantum 52LPS drive. Sling in a Quantum 240LPS and you can expect around
- >1.3-1.6 M/sec, sustainable. GVP's 040 for the A3000 doesn't have a scsi
- >controller onboard, even if it did it would be slower than the standard one
- >unless it was a FAST scsi-II controller.
- >Dave Haynie said they have a Maxtor Magic drive in the C= software department
- >that runs 3.9M/sec through the filesystem and 4.2M/sec direct to the driver.
- >I don't know what particular drive this is. Perhaps Dave could post the
- >model no. for us :-)
- >The A3000 scsi controller is probably the fastest scsi controller you'll find
- >on anything under $10000.
- >
- >--
- >copes@cs.curtin.edu.au
-
- I think that you should expect around 1.4 to 2.3 MBytes/s (depending on which
- track the r/w head is on the harddisk since the Quantum LPS120/240 use variable
- "bytes per track", the outer tracks contain more bytes than the inner ones).
- This is the throughput that scsispeed (4.2?) reports on my machine (A2000 w/
- 2630 and HardFrame, Quantum LPS120S with 250 buffers). The diskspeed results
- vary of course since it depends on the fragmentation of your harddisk etc.
- After a backup and restore (-> not fragmented) with approx. 3800 files and
- 66% full (-> due to the way the filesystem writes on the harddisk -> first empty
- space somewhere in the outer tracks with more bytes), I get up to 1.9 MB/s
- create, 1.7 MB/s write (why do I get better performance during creation?), and
- 2.1 MB/s read.
- Since an internal verify takes exactly as long as reading the entire harddisk
- with readblocks (69 seconds) I assume that the controller keeps up with the
- harddisk, therefore, the A3000 controller should also keep up (however with more
- available cpu time due to its wider and faster bus system (A2000 -> zorroII bus,
- A3000 scsi adapter "controller" -> internal A3000 bus)
-
- 238766 blocks * 512 bytes/block = 116.6 MBytes
- 116.6 MB / 69s = 1.7 MB/s average (over entire disk)
-
- Christian
-
- P.S.: I assume that an internal verify does not take any longer than
- a raw read since the LPS120/240 always checks the checksum of the
- blocks they read.
-