Although there has always been quite a difference between the performance of batch tasks against filebased memories vs server based (Groupshare) memories, since the move from Groupshare 2015 to Groupshare 2017 this difference has started to become a major problem for production. I was wondering what sort of variation other users are seeing to help give an idea of what can be achieved. The batch tasks which are most singificantly impacted are analysis, pretranslate and TM Update.  Just as an example an analysis task for 36k words for 1 target languages take 2 minutes to run vs a filebased memory, if I import the same memory into groupshare server memory this same task takes a minimum of 15 x longer so takes over 30 minutes, the same ratios hold true for pretranslate tasks as well. For these two batch tasks, although highly undesireable it is possible to use filebase memory exports from the server to process files much more quickly.

The area which is most problematic and that seems to have been most severly impacted since the upgrade from GS2015 to GS2017 are the TM update batch tasks, these sometimes have to be run overnight for tasks that with GS 2015 used to only take a couple of files and often the tasks fail so it is a major problem and is almost not useable anymore.

Are other users seeing this level of performance drop between file based memory and server memory batch tasks? If not what sort of differences do you see?

Secondly are other users having problems with TM update batch task performance for server based memories and has there been a huge change since the move from GS 2015 to GS 2017?

We are using hardware well above the GS suggetsed specs and have worked with SDL to try and improve performance with no major progress so far, we have already upgraded several times to various Groupshare 2017 patches (currently at CU5) but are not keen to be an early adopter for GS 2017 SR1 as we don't want to be the ones to find all the bugs.

Any feedback about general groupshare performance from other users and also and feedback specifically about the TM Update batch task for server memories would be greatly recieved, many thanks!