SCSI Throughput vs. Network Throughput, Part 2

1998 – I’ve looked at the theory of using an older, slower Mac as a server on a 10Base-T ethernet network in SCSI Throughput vs. Network Throughput – now on to the testing.

Ethernet vs. SCSI on a Mac IIsi

Using a Macintosh IIsi with 17 MB RAM running System 7.5.5 on an Apple-branded Quantum hard drive, I benchmarked performance using Speedometer 3.06. It’s not the best benchmark program, but it runs on almost everything (even 68000-based Macs), and it runs fairly quickly.

System Drive Tested Disk Rating
1 MB cache
Disk Rating
64 KB cache
IIsi Quantum 80 MB 5.2  2.4
IIsi, Speed Doubler Quantum 80 MB 2.0 not tested
IIsi Quantum 40 MB 6.3  1.8
IIsi Zip drive 5.1  1.6
IIsi ramBunctious RAM disk 5.3 14.9
IIsi, OpenTransport ethernet, IIfx RAM* 0.38  0.38
IIsi, classic network ethernet, IIfx RAM* 0.52  0.49
IIsi, classic network ethernet, IIfx HD 0.31  0.31
IIsi, OpenTransport ethernet, 7600 HD** 0.50  0.50
IIsi, classic network ethernet, 7600 HD** 1.1  1.1
IIsi, classic network LocalTalk, 7600 HD 0.16 not tested
* Macintosh IIfx, 32 MB RAM, System 7.5.5, OpenTransport, AppleShare 3.0, 7.5 MB disk cache, 8 MB RAM disk created with ramBunctious.
** Power Mac 7600/132, 48 MB RAM, virtual memory enabled, System 7.6.1, AppleShare 5.0, 4 MB disk cache.

The Mac IIsi accessed the hard drives in the Mac IIfx and Power Mac 7600 over ethernet, and I also tested LocalTalk performance on the 7600 – just because. Its performance was about 15% of ethernet in this test. The size of the disk cache on the IIsi made almost no difference over a network.

It’s interesting to note that Speedometer reports a lower hard drive score with Speed Doubler than with Apple’s disk cache. (Speedometer 3.06 tests the drive using a 1 MB file, so a larger cache would not produce a higher score.) Part of the reason is that Speedometer uses random access while Speed Doubler is trying to intelligently predict access.

It’s also interesting that classic networking is faster than Open Transport on the Mac IIsi. Accessing the 7600 server with classic networking was over twice as fast as the newer Open Transport protocol. There was only a 30% improvement with the IIfx.

It’s no great surprise that cache size on the client computer makes no significant difference.

Bear in mind that all these systems are using different hard drives, which will explain some of the difference in performance.

Macintosh II Series as AppleShare Server

In my previous article, I claimed that any Mac II series computer would make a decent small group server. It’s time to test that claim. My first testbed is a Macintosh IIsi with 17 MB RAM, an 80 MB Apple-branded Quantum hard drive, 10Base-T ethernet, System 7.5.5, and AppleShare 3. I also have a Mac IIfx with 32 MB RAM, a 160 MB IBM hard drive, ethernet, System 7.6.1, and AppleShare 3. These are compared with a corporate server at my workplace – a Power Mac 7600 with 48 MB RAM, RAID drives, and AppleShare IP 5.0.

As above, tests are conducted with Speedometer 3.06. Tests are run from a Power Mac 7600/180 with 96 MB RAM, a 4 MB disk cache, RAM Doubler, and Speed Doubler. Both computers are on a switched ethernet hub, so network traffic should have no bearing.

Variables include cache size, the server (7600, IIsi, IIfx), personal file sharing. AppleShare, classic v. Open Transport networking, and the drive mechanism.

System Drive Tested Disk Rating
1 MB cache
Disk Rating
64 KB cache
Mac SE, AppleShare 3 via classic networking n/a 0.19
Mac II, AppleShare 3 via classic networking 0.50 0.45
IIsi, file sharing via classic networking 0.45 0.42
IIsi, file sharing via Open Transport 0.30 0.28
IIsi, AppleShare 3 via classic networking 0.41 not tested
IIfx, AppleShare 3 via Open Transport 0.38 not tested
IIfx, AppleShare 3 RAM disk via OT 0.57 not tested
7600/132, AS IP 5 via Open Transport 1.75 n/a
7600/180 internal drive 8.4 n/a

Based on these results, I plan to move the Mac IIfx back from Mac OS 7.6.1, which only supports Open Transport networking, to System 7.5.5 so I can use classic networking, which is less demanding of resources and faster on 680×0 Macs, instead of the more processor intensive Open Transport.

Update: Part of the reason AppleShare 3 is slower with Open Transport (OT) is that it was designed before OT and doesn’t take advantage of OT calls. The current version of AppleShare, which won’t run on pre-68040 Macs, takes full advantage of OT.

According to Speedometer, the ancient IIsi with an equally old hard drive offers one-fourth the performance of a 7600 with RAID drives and AppleShare 5. Moving the IIfx from OT to classic networking should put performance at about 0.6, one-third the performance of the Power Mac.

Using a modern hard drive could boost performance even further, perhaps into the area of 0.6 on the IIsi and 0.9 on the IIfx.

Keywords: #opentransport #appleshare

Short link: