Please find attached a HTML file with a few bullet points and a few
graphs on what exactly was done on the performance front, in the
Evolution 2.7 development cycle.

Hope you find this useful. Let me know if you would like to have
more/different information.


Performance and Memory optimizations in Evolution 2.8 :



    Do not fetch all HEADERS. Only fetch a minimal functional subset of Headers. This resulted in a reduction of initial header loading time by about 48%, against Evolution 2.6. See Fig. IMAP Headers Load Time.

GroupWise :

Comparison of item_ids during server synchronization operations reduced to O(n) from O(n^2) , thereby eliminating frequent CPU spikes during periodic refresh operations.

Evolution Data Server

Reduction of memory footprint :

  • Multiple copies of libical not statically linked to the providers anymore. This cuts down

    primary memory requirements by around 600 KB.

  • Allow dynamic linking to system db

  • GroupWise – split the provider cache by components (Calendar/Tasks/Memos) ensuring

    only one copy of the item exists in cache. Reduces memory used by the GW provider by 66%.

GroupWise backend :

  • Loading of Tasks into the cache during account creation reduced by about 80% .

    See Graph – Tasks Loading.

And to let you know that this is not optimization just for the sake of itself but a trade-off for better user experience...

  • With Smarter sequencing of GroupWise Calendar pre-load in Evolution 2.8, it takes longer to fetch all items into the cache overall (See Graph : Calendar Loading Time) but the overall user experience (measured by the time taken to load/display events closer to current date) improved by a whopping 94% in average.

Attachment: evoperformanceformarketing_html_71baac1e.gif
Description: GIF image

Attachment: evoperformanceformarketing_html_m3d90fa27.gif
Description: GIF image

Attachment: evoperformanceformarketing_html_m22de1aa3.gif
Description: GIF image

Attachment: evoperformanceformarketing_html_m5559c5ad.gif
Description: GIF image

marketing-list mailing list

Reply via email to