Skip to main content

[archive] Vison file formant performance

  • June 9, 2008
  • 1 reply
  • 0 views

[Migrated content. Thread originally posted on 29 May 2008]

Does anyone have any experience with performance testing differences between Vision 3 and Vision 4?

From a past conversation with technical support, I was under the impression that while vision 4 allows a data file size larger than 2 gb, I was potentially expecting to see a decrease in performance due to the requirement that all I-O would be forced to be a minimum of two prcesses, one to find the index, then another to load the data into memory.

After a customer was forced to convert a file to vision 4 due to its size and ran into performance issues with the file rebuild program (deletes it and rebuilds all the information in it from other files)

I did some testing in house and don't see much of a difference, but I don't have 2gb datafiles to test with unfortunately. I talked to support and they said they've seen some differences both better and worse between the two, completely dependant on the situation.

Does anyone else have experience with this?

brad

1 reply

[Migrated content. Thread originally posted on 29 May 2008]

Does anyone have any experience with performance testing differences between Vision 3 and Vision 4?

From a past conversation with technical support, I was under the impression that while vision 4 allows a data file size larger than 2 gb, I was potentially expecting to see a decrease in performance due to the requirement that all I-O would be forced to be a minimum of two prcesses, one to find the index, then another to load the data into memory.

After a customer was forced to convert a file to vision 4 due to its size and ran into performance issues with the file rebuild program (deletes it and rebuilds all the information in it from other files)

I did some testing in house and don't see much of a difference, but I don't have 2gb datafiles to test with unfortunately. I talked to support and they said they've seen some differences both better and worse between the two, completely dependant on the situation.

Does anyone else have experience with this?

brad
V4 versus V3 performance differences should be minimal. While there is more disk I-0 going on in V4 versus V3, having the indices in a separate file has many advantages. Rebuilds (even with very larger files) should not take very long especially when using a current version (since bulf addition was made available). Rebuilds can go slow if the disk or memory do not have a lot of room to create temp files before the final file is written.