Simple database software for sysV?

[Leslie Mikesell]

In article <6332 at spdcc.SPDCC.COM> rbraun at spdcc.COM (Rich Braun) writes:

>Insofar as VP/ix is concerned, I'm concerned about its performance.  Can
>you tell me what to expect in terms of speed when I instruct the freebie
>DOS (read: low-performance even with PCKWIK, a disk cache) software to
>sort 10,000 records under VP/ix?  Let's say an operation now takes 8
>minutes.  How long will it take with VP/ix, given identical hardware and
>given the existence of a disk cache under DOS?

I haven't done any formal timing but the general feeling I get is that
VP/ix disk operations tend to be faster than native DOS due to the
better performance of the unix disk cache.  I would expect this to
vary wildly depending on (a) the match between the i/o requests and
the cache management (note that a cache may not help much with a huge
sort or random i/o), and (b) the size of the individual requests, with
smaller requests generating more system call overhead.
On a personal machine, though, you have the option of simply re-booting
under dos if it really makes a difference.  All you have to do is reserve
a partition on the HD and keep a boot floppy around. 

>Minor note on VP/ix performance (note that I have this on my office
>computer):  it causes the serial port to drop characters.  One can
>observe this either by using Procomm to dial out, or by dialing into
>the Unix system and running VP/ix.  (Yes, this does give you COMMAND.COM
>remotely at 2400 baud.  You've got to see it to believe it...!)

Like everything else on unix, it depends on the machine speed and the
load of other tasks.  On a 33Mhz 386 w/5Megs memory and no network, I
have no trouble at all with 2400 baud operations under VP/ix even with
a compile going on at the same time.  However, if you have enough operations
going on that you are paging out the VP/ix process, I would expect trouble.   

Les Mikesell
 les at chinet.chi.il.us