Author Topic: Is EA 16 noticeably faster than previous versions?  (Read 643 times)

Richard Freggi

  • EA User
  • **
  • Posts: 390
  • Karma: +14/-7
    • View Profile
Is EA 16 noticeably faster than previous versions?
« on: April 27, 2022, 03:28:43 pm »
I'm using v 1310 and my *.feap (Firebird) model is about 1GB.  The PC is Win10 Intel i7 16GB RAM,  500M SSD.
It takes me about 15 hours to import total 500K elements (classes, attributes) into my model using Geert's importer.
It takes about 20 seconds to delete a single element from the project explorer.
It's possible that EA is slowed down by some antivirus thingy from my work PC, but I don't think so.

One hope of V16 is that using the lighter footprint SQLite (and maybe Sparx can find a way of letting us use SQLite 'memory' mode, with database loaded fully in RAM?) , I can reduce these times significantly.  Can anyone comment on performance improvement with *.QEA files compared to *.eap or *.feap?

Geert Bellekens

  • EA Guru
  • *****
  • Posts: 11849
  • Karma: +460/-33
  • Make EA work for YOU!
    • View Profile
    • Enterprise Architect Consultant and Value Added Reseller
Re: Is EA 16 noticeably faster than previous versions?
« Reply #1 on: April 27, 2022, 04:55:59 pm »
Hi Richard,

I don't have an answer about the performance improvement of v16, but did you consider using a "real" database as backend?
I think that will improve the situation greatly.

You don't even need a server. I'm running EA on a local MS SQL Express (=free) installation

Geert

qwerty

  • EA Guru
  • *****
  • Posts: 12814
  • Karma: +366/-295
  • I'm no guru at all
    • View Profile
Re: Is EA 16 noticeably faster than previous versions?
« Reply #2 on: April 27, 2022, 05:22:26 pm »
As long as databases are at "decent" sizes you can go on with any of the cheap ones. Probably you will find benchmarks for any that proves them the fastest. But databases are getting tricky at certain sizes. And the assume or don't how they are queried. Oracle for example is optimized for large bulk queries. And that works for warehouses and the like. But it kills EA. As a general user experince MS SQL, Maria, PostGreSQL (and others) are best for large repositories. EAP (and the other tiny ones) are good for a few 100 MB.

It would be nice if Sparx had some ressource where they recommend which DB to use when (there might alread be something but I don't kow it).

q.

wivel

  • EA User
  • **
  • Posts: 140
  • Karma: +9/-1
  • Driven by Models
    • View Profile
Re: Is EA 16 noticeably faster than previous versions?
« Reply #3 on: April 27, 2022, 08:29:14 pm »
I have been using EA 16 for a few months now in different scenarios, mainly on an SQL Server somewhere in the organisation, a local SQL Server Express installation and the new local SQLite based format, QEAX. Same scenarios as I had using EA 15.2, except for the QEAX file and I see no change in performance/speed.

What I do notice is, as Geert pointed out, that the choice of database technology is what matters most when it comes to EA performance in general, and yes, it would be interesting if Sparx could provide comparisons.

To me the upside of the 64bit version of EA 16 is not performance, but its ability to handle larger models when importing and exporting and also migrating data. I had an issue at a client, where EA 15.2 crashed and/or hang when trying to migrate a large model to a new MS SQL Database. When using EA 16 64bit all went smoothly. Slow but smoothly :)

Henrik

Richard Freggi

  • EA User
  • **
  • Posts: 390
  • Karma: +14/-7
    • View Profile
Re: Is EA 16 noticeably faster than previous versions?
« Reply #4 on: April 28, 2022, 12:24:25 am »
Thanks all.  So I should not expecting performance improvements, that's OK.  I need to keep my work machine as lean as possible so *.feap is good enough, no motivation to go to Oracle or other.  I do have a thing for open source solutions though after being burned by proprietary license changes a couple of times, so Firefox and SQLite and all related ecosystem solutions are definitely my favorites.