Commit | Line | Data |
---|---|---|
c557cacd C |
1 | .ds RH Motivation |
2 | .NH | |
3 | Motivation | |
4 | .PP | |
5 | These benchmarks were performed for several reasons. | |
6 | Foremost was our desire to obtain guideline to aid | |
7 | in choosing one the most expensive components of any | |
8 | VAX UNIX configuration, the disk storage system. | |
9 | The range of choices in this area has increased dramatically | |
10 | in the last year. | |
11 | DEC has become, with the introduction of the UDA50/RA81 system, | |
12 | cost competitive | |
13 | in the area of disk storage for the first time. | |
14 | Emulex's entry into the VAX 11/780 SBI controller | |
15 | field, the SC780, represented a important choice for us to examine, given | |
16 | our previous success with their VAX 11/750 SC750 controller and | |
17 | their UNIBUS controllers. | |
18 | The Fujitsu 2351A | |
19 | Winchester disk drive represents the lowest cost-per-byte disk storage | |
20 | known to us. | |
21 | In addition, Fujitsu's reputation for reliability was appealing. | |
22 | The many attractive aspects of these components justified a more | |
23 | careful examination of their performance aspects under UNIX. | |
24 | .PP | |
25 | In addition to the direct motivation of developing an effective | |
26 | choice of storage systems, we hoped to gain more insight into | |
27 | VAX UNIX file system and I/O performance in general. | |
28 | What generic characteristics of I/O subsystems are most | |
29 | important? | |
30 | How important is the location of the controller on the SBI/CMI versus | |
31 | the UNIBUS? | |
32 | Is extensive buffering in the controller essential or even important? | |
33 | How much can be gained by putting more of the storage system | |
34 | management and optimization function in the controller as | |
35 | DEC does with the UDA50? | |
36 | .PP | |
37 | We also wanted to resolve particular speculation about the value of | |
38 | storage system optimization by a controller in a UNIX | |
39 | environment. | |
40 | Is the access optimization as effective as that already provided | |
41 | by the existing 4.2BSD UNIX device handlers for traditional disks? | |
42 | VMS disk handlers do no seek optimization. | |
43 | This gives the UDA50 controller an advantage over other controllers | |
44 | under VMS which is not likely to be as important to UNIX. | |
45 | Are there penalties associated with greater intelligence in the controller? | |
46 | .PP | |
47 | A third and last reason for evaluating this equipment is comparable | |
48 | to the proverbial mountain climbers answer when asked why he climbs | |
49 | a particular mountain, | |
50 | ``It was there.'' | |
51 | In our case the equipment | |
52 | was there. | |
53 | We were lucky enough to assemble all the desired disks and controllers | |
54 | and get them installed on a temporarily idle VAX 11/780. | |
55 | This got us started collecting data. | |
56 | Although many of the tests were later rerun on a variety of other systems, | |
57 | this initial test bed was essential for working out the testing bugs | |
58 | and getting our feet wet. | |
59 | .ds RH Equipment | |
60 | .bp |