EA response time degradation ( >= 10 sec.) cause by ModelExpert

[Marc Lampron]

HI,
We are currently using ModelExpert (eaTeamWorks 64-bit) with Enterprise Architect v16.1 (64-bit) and we are seeing a degradation in the performance (user perceived response time) of Enterprise Architect (EA) when we create new elements in EA.
The degradation of the response time is at least 10 sec. and after 5 sec. the message “EA is no longer responding” appears in the EA window header.
Our configuration of the Sparx Systems platform is relatively standard in our local installations with the database on a SQL Server 2022 server and Pro Cloud Server and Prolaborate on a Windows server that we access via a cloud connection.
If we only disable the ModelExpert64 extension of the eaTeamWorks suite, the response time is now less than 5 sec., which is quite acceptable.
Our team of technical experts who investigated the issue with monitoring tools (network, server, anti-virus, workstation, etc.) noticed abnormal activity on the user workstation but everything was normal and optimal in all other aspects. Everything goes back to normal if you deactivate the ModelExpert64 add-in in EA.
I understand that some uses of ModelExpert (modeling assistance) must monitor certain internal events inside EA to offer guidance to modeling in real-time. Do you think the behavior of ModelExpert and the degradation of EA’s response time are normal ?

[Ian Mitchell]

If your normal response time to create a new element is 5 sec, then if you have Model Expert enabled, I would expect the response time to be of the order of 2-3 times as long. This is because, as you correctly said, MX is doing lots of checks when an element is created, even if at the end it takes no action. For users who are not using the full Guided Modelling function, I suggest you switch off MX. Then just enable it if they want to use any of the post-modelling functions, like a meta-model check.
We don’t see this in our testing, as we use only local databases, with fast response times.
Perhaps what we need is some kind of ‘fail fast’ feature, where you can specify, on a per-user basis, whether they are using MX Guided Modelling ( probably with default=no) and cache this setting, so that in general, MX can exit from handling the ‘new element’ event without any more database calls.
I am back in the office in around 10 days, so I will look into this as a priority.
Thanks
Ian

[Ian Mitchell]

Update:
I have added a new function into the next version of Model Expert – to be released in eaTeamworks 1.4 along with lots of other improvements.
This will change the way that Guided Modelling works. It will now be default OFF, so that most users, who probably don’t use MX-GM, will not see a performance loss.
Enabling GM will now be added to the list of things to do when starting to to use MX-GM.

User will still be able to use the analysis and dashboard functions of MX as today.

The setting is per-user, along with all the other user-based settings of eaTeamWorks, so should work ok in virtualised environments.
Thanks for the sugestion.

[Author]

Posts