Very Computer

Hello there.

We've been doing a lot of installations of the Btrieve 6.15 lately, due to
upgrading our application (Scala) at numerous clients (both small and large
volume sites). We have had our share of troubleshooting and
performance-tuning to do, but 2 sites still run slowly, regardless of the
hardware. Both sites run Novell Netware 4.11 on Token-Ring (4Mbit), it's the
only 2 sites we have that run Token-Ring.

We've upgraded Btrieve to 6.15.451 and solved the -94 problems, and set
the -r parameter for BSPXCOM to -r=4096.

Still, both installations run noticeably slower than other equivalent
networks running on Ethernet.

So, if anyone has some hints or could sketch out what is going on/wrong for
us on Token-Ring networks, it would be greatly appreciated.

Best regards,
Helge Tengstedt
Technical Manager
Strongline Software Ltd.

By 'noticeable slower', do you mean something like half the speed?

You say the token ring networks are 4Mb/s. This is less than half
of ethernets 10Mb/s. I've seen production ethernets happly run
with data transfer rates of 7 or 8 Mb/s. This could be enough to
account for the difference.

Guy
-- ---------------------------------------------------------------------

4.4>5.4 4.4>5.4 4.4>5.4 The Reality Check's in the Post! 4.4>5.4 4.4>5.4

The most conclusive way to solve this by capturing packet traces with a
protocol analyzer and comparing the two.  This is the primary way that I
support my Btrieve customers...it gives you ultimate visability into
whats happening at all layers (network through application).

There are analyzers available that will at least let you capture packets
for < $1000 (e.g. run on windows).  Then you need only find someone with
higher end analyzer (e.g. Network General Sniffer) to decode and analyze
your traces.  I might be of more help, but I see you're not in the U.S.

Since the maximum packet size for SPX (Novell) is about 500 bytes
anyway...there shouldn't be any difference between Token Ring and
Ethernet due to the larger frame allowed frame sizes.  Obviously 4Mbps
Token ring vs. 10Mbps Ethernet could be substantially slower depending
on what other traffic is on these segments.