MonkeyDiscussion

MeatballWiki | RecentChanges | Random Page | Indices | Categories

General discussion about InfiniteMonkey.

Chris:

I am really impressed by what you are doing. As an analogy... On several occassions I've lived in a house while it was being renovated, a process that has always been a most disruptive mess. What you have done here, upgrading the system(s) to the extent you have, with such minimal disruption, is "Truly impressive. Thanks!" HansWobbe.

: Thanks, Hans. I got good at fast transitions when I was playing with PeriPeri/1. It's a lot harder with a house: you can't just build a new one and quickly switch it in the place of the old one ;) -- ChrisPurcell

Chris, regarding the switch from file system to mySQL: I would be very much interested in the performance comparison of this step. Redarding: (1) difference in memory usage (2) difference in average time of page delivery (3) difference in timing of full text search. Would you mind recording this on your transition? -- HelmutLeitner

: (1) I have absolutely no idea. I would expect flatfiles to win here, because of the overhead of running MySQL?.

: (2) Average time of page delivery is quite similar, with flatfiles winning by a bit, probably because determining whether a link is broken or not is much faster for flatfiles.

: (3) There's an order of magnitude difference in simple search times. Backlinks are more variable, as each page hit must be rechecked within Perl to verify there really are backlinks in the latest revision. MySQL? still beats out flatfiles, but not necessarily by an order of magnitude.

: The largest performance gain is on the maintenance run: formerly taking minutes to run, and blocking all use of the site for the duration, it now completes in about ten seconds concurrently with regular operations.

: Moving to a newer version of MySQL? (allowing subqueries), and storing backlinks in a separate table would probably speed up the engine, as the number of queries could be reduced. Equally, mod_perl would save time opening the initial database connection, which adds a noticeable and quite variable performance hit. Load's not enough of a problem to worry about right now, though.

: Does latency seem significantly increased to MB regulars? -- ChrisPurcell

I just want to note, that the mb wiki engine has become less useful to me during the last developments. There are many improvements on the other side, which partially make up for this, but:

I miss full text search, the real thing where "roadmap" will find "MeatballRoadmap?".
I miss searching of page titles (I can work around this by searching the text of action=index, maybe there is a better way?)
I miss the combined display of diff+page.
(I slightly dislike that rc and diff have become much slower)

-- HelmutLeitner

: Thanks for the criticisms, Helmut! I've just had an idea for fixing the interaction between full-text search and CamelCase, so will let you know when that's in place and you can see whether it works for you. Speeding up rc and diff is on the to-do list, but not as a priority. -- ChrisPurcell

: It's taken me a while (3 months?!), but I've fixed CamelCase and page-title searches. (Amusingly, your example user-story still doesn't work, but that's because MeatballRoadmap? doesn't exist.) -- ChrisPurcell

CategoryMonkey