Has anyone experience of the 'packet size' used by Crystal when operating on ISAM files through a Novell network?
When Crystal is used to access native (ISAM style) DataFlex files, response is determined by whether caching on the Novell Client is turned on or off. The difference is usually about a factor of 10 but it can be much more - 45 times faster in one of our tests!
Running a PC with caching on gets files well screwed-up because implicit in the 'caching on' scenario is opportunistic locking which neither Microsoft nor Novell seem to have fully sorted. So, bottom line, caching is off and Crystal is very slow.
There is no reason why Crystal should be slower when reading records from a server. We put a 'sniffer' on the network and found was that the packets were tiny with caching off and large with it on. On one test we had 90,000 packets with caching off and 2,000 with it on.
This looks like a problem in the first Crystal 7 release where 'read' instead of 'burst read' was used. So, my question is, why does Crystal 10 read tiny packets of data when on a network with caching off?
The classic answer is to move to SQL but this is heavy going.
Any thoughts appreciated!
When Crystal is used to access native (ISAM style) DataFlex files, response is determined by whether caching on the Novell Client is turned on or off. The difference is usually about a factor of 10 but it can be much more - 45 times faster in one of our tests!
Running a PC with caching on gets files well screwed-up because implicit in the 'caching on' scenario is opportunistic locking which neither Microsoft nor Novell seem to have fully sorted. So, bottom line, caching is off and Crystal is very slow.
There is no reason why Crystal should be slower when reading records from a server. We put a 'sniffer' on the network and found was that the packets were tiny with caching off and large with it on. On one test we had 90,000 packets with caching off and 2,000 with it on.
This looks like a problem in the first Crystal 7 release where 'read' instead of 'burst read' was used. So, my question is, why does Crystal 10 read tiny packets of data when on a network with caching off?
The classic answer is to move to SQL but this is heavy going.
Any thoughts appreciated!