home *** CD-ROM | disk | FTP | other *** search
Text File | 1986-04-04 | 10.8 KB | 209 lines | [TEXT/EDIT] |
- DISKBENCH_11.BIN (MAUG/DL2) 25-Mar-86 (1258 bytes)
- DiskBench does simple performance testing using the disk volume from
- which it is run. Results are meaningful only relative to those from
- competing systems. MDS source code is available (BROwse DBENCH.*).
- Please report results to me. Don't move the mouse during the test. On
- HyperDrive, run from Startup, which must be more than 1MB in size and not
- fragmented (e.g., 1MB+ Startup restored to clean disk). This new version
- fixes serious bug in previous one. --Steve Brecher 70001,1011
-
- [Editorial note: what follows is a digest of messages relating to DiskBench.]
-
- -----------------------------
- Sb: HD Benchmark
- Fm: Steve Brecher 70001,1011
- To: All
-
- I wrote a simple disk performance benchmark program which,
- along with its MDS assembler source code, is in MAUG/DL2.
-
- The test is designed to measure hardware and disk driver performance, with no
- influence from the file system (HFS/MFS) or volume/file/Finder configuration.
- The program issues I/O requests directly to the disk driver.
-
- The benchmark consists of three parts:
-
- (1) 100 reads of 32KB of data from the start of the volume;
- (2) 100 writes of 32KB of data to the start of the volume;
- --the above two tests measure data transfer speed; 32KB was chosen to be a
- reasonably large chunk, but not so large that it would cross a cylinder
- boundary (thus not requiring any head movement).
-
- (3) 40 iterations of: read one 512-byte block from an offset of 1MB,
- followed by read of one 512-byte block from start of volume;
- --the above test measures access time, i.e., seek or head movement speed.
-
- The test is non-destructive --
- Test (2) writes the data that was read in test (1).
- Test (3) is bypassed if the volume size is less than 1MB+512bytes.
- The volume tested is that from which the program is run.
-
- On a Mac Plus, make sure disk caching is disabled in the Control
- Panel. Also, DO NOT move the mouse while the test is in progress.
-
- The data transfer read/write test results are the time, in sixtieths of a
- second, it takes to read/write 32KB, 100 times, from/to the start of the
- volume from which DiskBench is run.
-
- The access time test is the time, in sixtieths of a second, it takes to do
- forty repetitions of: read 512 bytes from an offset of 1MB into the volume;
- read 512 bytes from the start of the volume.
-
- The data transfer test is designed to indicate relative speed of transferring
- data between the disk and Mac RAM, with no influence from the software
- environment, data/directory/file structure on the disk, or movement of the
- disk heads. (The latter is not true with floppies; they have less than 32KB
- in a cylinder, so they have to move their heads back and forth during the
- data transfer tests.)
-
- The access time test is designed to indicate relative speed of moving the
- disk heads to a desired position.
-
- The DiskBench program issues requests directly to the disk driver, bypassing
- the file system (bypassing HFS/MFS). It provides a relative measure of the
- performance of the disk hardware and disk driver combination. Any disk which
- has substantially better results on both tests than another will be faster in
- actual use, given the same surrounding hardware and software. I believe (but
- cannot prove) that on the Mac (not necessarily on all computers) the data
- transfer time is relatively more important than the access time; this is not
- to say the latter should be ignored.
-
- My belief is based on impressions and hearsay. Impressions -- listening to
- disks in operation; not many long seeks on the Mac. Hearsay -- a report that
- MicahDrive is "35% faster" than AST's fast-seeker, and "10-15% faster" than
- HyperDrive (which, best I can tell so far, seeks somewhat faster than
- MicahDrive). No details supplied with the hearsay, so they could be hogwash.
-
- MicahDrive has the fastest data transfer, but not particularly fast access
- time; and I haven't heard of a case where a given application loads faster
- from another drive. Excel loads in about 6.5 seconds on MicahDrive (give or
- take -- measured by eyeball on adjacent Mac's alarm clock, from Open to
- appearance of worksheet).
-
- Files tend not to be too fragmented in my experience, so seeks involved in
- resource loading to launch an application are typically over a short distance
- -- a cylinder or two. (And many if not most of the small pieces will be in
- the same cylinder as the previous.) Differences among drives in
- track-to-track times are not as great as differences in longer movements
- (e.g., as in the average access times usually quoted). If a disk does become
- unduly fragmented, as can happen over time, it can be repaired by
- backup/restore.
-
- Suppose it took 50 seeks to load a 100K application. Suppose there was a
- difference of 20ms per seek between two drives, but that the faster seeker
- transferred each 2K piece at 200KBits/sec while the slower seeker did
- 450KBits/sec. The fast seeker gains 1.0 second on seeks, but loses 2.2
- seconds on data transfer. In this example, the data transfer ratio is
- typical of, say, that between MicahDrive and a decent external SCSI; but the
- difference in access time is, I think, exaggerated for the typical distances
- involved even if the external SCSI's average access time spec is 40ms faster.
- (That spec measures time to seek across a big chunk -- a third or half or so
- of the surface; I forget exactly how it's computed.)
-
- For the majority of the market, the question is moot. All of the under-$2000
- drives have average access times in the same ballpark (65-85ms).
-
- (Current DiskBench results list follows.)
-
-
- Data transfer Access Tester
- ---- time ---- time
- Reads Writes
-
-
- 400K floppy drive, Apple 8756 11816 N/A S. Brecher
- 400K floppy drive, Apple 2984(?) 12392 N/A G. Frascadore
- 400K floppy drive, Apple 2984(?) 12351 N/A R. Perez
- 800K floppy drive, SS, Apple 8758 11407 N/A S. Brecher
- 800K floppy drive, DS, Apple 7701 10874 N/A S. Brecher
- 800K floppy drive, DS, Apple 7523 10913 N/A N. Fong
- AST 4000, AST Research 1495 1533 159 KATZ, Mousehole BBS
- AST 4000, AST Research 1495 1549 160 KATZ (second drive)
- AST 4000, AST Research 1495 1537 169 KATZ (third drive)
- DataFrame 20, SuperMac 1319 2233 488 J. Bean
- DataFrame 20, SuperMac 1344 2233 487 S. Brecher
- Hard Disk 20, Apple 7074 7871 368 N. Fong
- Hard Disk 20, Apple 7054 7944 370 KATZ, Mousehole BBS
- Hard Disk 20, Apple 9883 6948 368 R. Wiggins
- HyperDrive 10, obsolete model 1591 1616 401 S. Brecher
- HyperDrive 10, GCC (V2R1) 8000(?) 7982(?) 648(?) H. Conover
- HyperDrive 10, GCC (V2R1) 7985(?) 6892(?) 485 R. Perez
- HyperDrive 20, GCC (V2R1) 1703 1506 640(?) R. Ford
- HyperDrive 20, GCC 1704 1506 241(?) W. Luckie
- MacBottom 10, v2.1, PCPC 4159 6897 686 M. O'Connor
- MacBottom 10, v2.6, PCPC 4159 6897 608 S. Aronian
- MacBottom 20, v2.1, PCPC 4110 6817 601 L. Becker
- MacDrive, 10MB Fixed, Tecmar 6017 6719 401 C. Nicholais
- MicahDrive 20 AT, Micah 508 507 488 S. Brecher
- Quark QC-20 6476 6488 82(?) R. Thacker
- QuickDrive external RAMdisk 2411 2479 52 R. Bates
- QuickDrive external RAMdisk 2466 2535 33 J. Eugenides
- RamStart RAMdisk/Beck-Tech RAM 186 186 N/A G. Frascadore
- Warp 20, Warp Nine Engin'rng 14537 14537 321 G. Frascadore
- To: Andy Hertzfeld 70167,3430 (X)
-
- Latency is generally included in access time.
-
- I think the benchmarks don't confuse people as much as advertising hype and
- vague claims ("blinding speed") posted on networks. And I don't think it's
- easy to just measure how long it takes to do things -- your parenthetical
- remark about different [contextual factors] on different disks is the big
- problem there. Many of the "disk comparisons" I've seen were really
- comparisons of the file system (MFS vs HFS).
-
- DiskBench is a cheap (easy to write, easy to run) measurement of the relative
- performance of what the consumer is buying when he buys a hard disk --
- *provided* that it is understood that the relative numbers do not translate
- directly into erceived speed of overall Mac operation, or even overall disk
- performance. The only reason they don't translate into the latter is the
- existence of the issue we were discussing -- the relative importance of data
- transfer vs. access time.
-
- I think the DiskBench results will surprise a lot of people. By and large, I
- think those people will be better informed about relative disk performance
- than if the results weren't available. Further, I think it will surprise
- some vendors, and will lead them to take another look at their driver
- software in an attempt to improve it. I know of at least one vendor who
- plans to do that as a result of seeing the benchmark. If he succeeds, his
- future customers will benefit.
-
- In sum, I don't see how the widespread pre-existing confusion about disk
- performance could be made any worse, and I think DiskBench, while very much a
- "poor man's benchmark," is a lot better than nothing.
-
- A better benchmark -- one that measured performance in typical use -- is
- either much harder to implement or so hard to use (requiring initializing the
- disk and comparing only when running the same Apple system software and
- identical disk contents ) that it's not feasible without a lot of time and
- effort. MacInTouch has designed a reasonably good benchmark of that sort,
- but to my knowledge they haven't yet published a results listing including
- recently-available disks all running under the same file system, etc. Those
- kind of results are expensive to generate.
-
- It *would* be possible to do a vastly improved DiskBench by instrumenting a
- driver to record the I/O patterns of one or more end users and then using
- that pattern in the benchmark. But that would take a lot longer than the few
- hours that DiskBench took to put together...
-
- 400K floppy drive, Apple, Sys 3.1.1/Find 5.2:
-
- Test#1 Test#2 Test#3
- Internal drive:
- Reads 2987 2989 2994
- Writes 12359 12444 12466
-
- External Drive:
- Reads 2981 2985 2979
- Writes 12325 12333 12349
-
- RAMdisk "old" system/Finder:
- Reads 184 184 185
- Writes 184 185 184
-
- RamStart Sys 3.1.1/Find 5.2
- Reads 184 185 184
- Writes 185 186 184
-
- The 29xx times for the 400K read tests are a mystery. Floppies cannot go
- that fast (the external drive port cannot go that fast). My own test was
- done with 128K ROMs.