Performance on HFCS.v27 questions

Greetings to all WD HFCS users:

We have a school districts with 18K+ students in 35 schools and up to 3 machines per school. Each school had a separate database on the same network HFCS. They were able to process meals fine and then we consolidated all the information in a Summary database. All of thesehwere using the same network HFCS using v23 in 32bit.

Now we are trying to put all students in one database shared by all 35 schools and up to 3 machines per school and performance has become too slow and unusable. We tried HFilter to select the students for a school which helped but still not fast enough. We have caches set to max recommended and 32GB of memory assigned to this HFCS server.

We are going to try using a Query containing all students attending a school and use that instead of the HFilter.

Will HFCS v27 installed as s64bit give better performance?

Will HFCS v27 take advantage of multiple cores and multiple threads?

What configuration options would improve performance?

What performance tips can you suggest?

We are under a tight deadline to fix this problem before the next school day.

Thanks in advance for any help!

Jeff Graham
Cascade Consulting



Edited 1 time(s). Last edit at 09/10/2022 06:40PM by Jeff Graham.

In order to get good performances in your situation, you need ALL of the following:

1. HFSQL in 64 bits (32 bits can manage only 2 GB of memory, as all 32 bits program do, so having 32 GB of memory means nothing)
2. MODIFY YOUR ANALYSIS: All keys must now become composite keys with the school ID as the first component and whatever was in the eky after that
3. Modify your program to use hfilter ON THE composite keys from point 2 before any access
4. If you were using queries, modify them to add the school condition

I may be forgetting other things, but that is the MINIMUM

Thanks Argus. those are good points.

Hi Jeff,

It also depends on the hardware and the speed of the network.
If the data transfer to the clients is slow then optimizing your software will not help much.
I inherited a database that is very badly designed, containing millions of records, but even then performance with queries is fine without optimizing analysis as long as the hardware is ok.

Regards,
Piet

Thanks Piet,

We did modify our application to use a GLOBAL modifiable query for the 18K+ file and made adjustments for concurrent updates to the balance from multiple machines. It was tested yesterday and the speed was very good.

Jeff Graham
Cascade Consulting