How do the various workstations compare in speed on the various servers available today when running Unibase by DMAC? About nine months ago we had planned to release the data in this article. At that time a prospective client complained that since we only had a 16 bit Windows client of Unibase by DMAC our results were meaningless.

Ok, so now we have a POSIX based ANSI standard C++ implementation 32 bit client for Windows in addition to the 16 bit POSIX based ANSI standard C implementation for Microsoft everything and POSIX based ANSI standard C 32 bit implementation for UNIX. We at DMAC know that speed may or may not have anything to do with a solution to a problem. We know that there are trends in the market which force certain decisions. We know that one walks on eggs when talking about this subject.

Fools Rush In

Even so, we have configured a single workstation (AMD586 with 133 mhz cpu) with "System Commander" so that we can run many different client operating systems with various servers on our network to give users the following numbers. Users can interpret the data as they wish; we think more data like this needs to be presented. We walk on eggs; don't push.

Three Test Comparision

For our comparisons we used three tests: mtest, dtest, and rawdata. Mtest (M) runs with drun and mp. It exercises all features of file and batch i/o including indexed files. The time to execute is recorded in minutes.

Dtest (D) runs with de and mp. It also exercises all features of file and batch i/o including indexed files. In addition it uses keystroke macros. The time to execute is recorded in minutes.

Rawdata (R) imports data into the Unibase by DMAC environment. It uses very little Unibase by DMAC unique code -- mainly standard POSIX ANSI C code. The time to execute is recorded in minutes.

Standalone Workstation Results

Client:                              Time in Minutes:
DOS 6.22                        M-6 D-7 R-4
Win 3.11                          M-9 D-10 R-4
Win 95                             M- D-13 R-4
Win NT4.0                       M-27 D-28 R-4

NOVELL 3.12 Server Results
Client:                               Time in Minutes:
DOS 6.22 (netx)               M-24 D-22 R-15
DOS 6.22 (client 32)         M-54 D-49 R-34
Win 3.11 (netx)                 M-38 D-35 R-20
Win 3.11 (client 32)          M-55 D-57 R-40
Win 95 (client 32)             M-62 D-60 R-41
Win NT4.0 (client 32)       M-41 D-35 R-30

NOVELL 4.0 Server Results
Client:                                Time in Minutes:
DOS 6.22 (netx)                M-51 D-45 R-21
DOS 6.22 (client 32)          M-42 D- ?  R-33
Win 3.11 (netx)                  M-63 D-37 R-27
Win 3.11 (client 32)            M-82 D-63 R-54
Win 95 (client 32)               M-73 D-95 R-45
WinNT4 (client 32)             M-40 D-42 R-34
Win 95 (TCP/IP) (32 bit)    M-80 D-? R-?

Windows NT 4.0 Server Results
Client:                                Time in Minutes:
DOS 6.22 (TCP/IP)           M-     D-  ? R-63
Win 95 (TCP/IP)                M-48 D-  ? R-68
WinNT4 (TCP/IP)              M-33 D-40 R-6

UNIX Server Results
Client:                                Time in Minutes:
DOS 6.22 (Kermit)            M-4 D-5 R-?
Win 95 (TCP/IP)                M-4 D-? R-?
WinNT4 (TCP/IP)              M-4 D-? R-?

Make Own Conclusions

Each person planning to capture data in a competitive environment must decide what is best for the organization.  Speed may be of little importance; and then again nothing impacts productivity like the overall throughput of a given segment of activity.