The Following 2 Users Say Thank You to mr_jrt For This Useful Post: | ||
![]() |
2011-11-17
, 18:51
|
Posts: 839 |
Thanked: 3,386 times |
Joined on Mar 2009
|
#122
|
The Following 4 Users Say Thank You to AapoRantalainen For This Useful Post: | ||
![]() |
2011-11-21
, 14:55
|
Posts: 85 |
Thanked: 97 times |
Joined on May 2011
|
#124
|
I was musing...
Having done quite a lot of work with the eventlogging system, I was wondering if extending it to make an easy optimisation might be a good idea, namely, splitting the database. Have all events older than a week stored in a second database, and merge the two via the API.
Would make all the common use cases (most recent events, showing last few events, etc) constant time, and mean corruption is likely only to affect the last week or so's events as the archive db would only be being updated once a week or so. Main downside would be software that accesses the DB directly using SQL would get very upset (naughty, naughty!)
Thoughts?
The Following User Says Thank You to computerinfo21 For This Useful Post: | ||
![]() |
2011-11-21
, 15:42
|
Posts: 249 |
Thanked: 277 times |
Joined on May 2010
@ Brighton, UK
|
#125
|
Makes sense, but wonder if the problems it would cause for software not aware of it would be worth it.
Also, seems that the conversations window only accesses a small subset of the data at a time anyway. ~30 messages at a time I think. If you try to scroll quickly you will notice a definite lag as it loads the next set when you get to the top of each screen.
I think you were on to something with your post here about changing the way it updates the group cache. Perhaps threading & updating the cache with a different thread would avoid locking things and causing the corruption in the first place? I am no expert in this area, just musing out loud for those who are to consider.
The Following User Says Thank You to mr_jrt For This Useful Post: | ||
![]() |
2011-11-21
, 15:44
|
Posts: 132 |
Thanked: 27 times |
Joined on Jan 2011
|
#126
|
The Following 2 Users Say Thank You to rishabsp For This Useful Post: | ||
![]() |
2011-11-21
, 18:57
|
Posts: 12 |
Thanked: 2 times |
Joined on Nov 2011
@ Bangladesh
|
#127
|
![]() |
2011-11-25
, 05:40
|
Posts: 569 |
Thanked: 462 times |
Joined on Jul 2010
@ USA
|
#128
|
![]() |
2011-11-25
, 11:53
|
Posts: 569 |
Thanked: 462 times |
Joined on Jul 2010
@ USA
|
#129
|
FOR GODS SAKE FIx THE ALARM!!! I NEVER get to work at time after last update :P
![]() |
2011-11-29
, 13:48
|
Posts: 839 |
Thanked: 3,386 times |
Joined on Mar 2009
|
#130
|
The Following User Says Thank You to AapoRantalainen For This Useful Post: | ||
![]() |
Tags |
cssu stable, nokia n900 |
Thread Tools | |
|
Having done quite a lot of work with the eventlogging system, I was wondering if extending it to make an easy optimisation might be a good idea, namely, splitting the database. Have all events older than a week stored in a second database, and merge the two via the API.
Would make all the common use cases (most recent events, showing last few events, etc) constant time, and mean corruption is likely only to affect the last week or so's events as the archive db would only be being updated once a week or so. Main downside would be software that accesses the DB directly using SQL would get very upset (naughty, naughty!)
Thoughts?