448073 - figure out the best cache values for Fennec on the n8x0 devices

Status: VERIFIED FIXED

(Firefox for Android Graveyard :: General, defect)

Description

[Doug Turner (:dougt)]

right now it is 1mb.  We will increase it to 4mb because loads "feel" snappier.  We should investigate and get real values here.

Comment 1

[Stuart Parmenter]

this will be a lot easier to test once we get things like talos and whatnot running fully

Comment 2

[John O'Duinn [:joduinn] (please use "needinfo?" flag)]

Is this still a problem or can we close this as WORKSFORME?

Comment 3

[Stuart Parmenter]

we can't close this -- we still don't have automated tools available to help automate finding the right set of numbers.

Comment 4

[Mark Finkle (:mfinkle) (use needinfo?)]

Stuart - how do you want to proceed with this? We have Talos up and running now right?

Comment 5

[Stuart Parmenter]

yeah.  I'd suggest us doing a binary search through different disk cache numbers until we find something we're mostly happy with.  Maybe start with 10mb as an upper bound?  We should take in to account that many people don't have a lot of room on their main partition on an n810...

Comment 6

[timeless]

um. the n810 has two partitions and you should really consider using the non root partition for cache.

if the user plugs in usb, they're probably not browsing, and a degraded experience is probably best.

the reason we didn't manage to do this for microb is not really technical. we were just distracted.

i'd suggest trying numbers closer to 50mb

Comment 7

[Mark Finkle (:mfinkle) (use needinfo?)]

* What prefs are we talking about here?
* How do we control the partition used for the cache?

Here are some current prefs:

/* cache prefs */
pref("browser.cache.disk.enable", false);
pref("browser.cache.disk.capacity", 0);
pref("browser.cache.memory.enable", true);
pref("browser.cache.memory.capacity", 1024);

/* http prefs */
pref("network.http.pipelining", true);
pref("network.http.pipelining.ssl", true);
pref("network.http.proxy.pipelining", true);
pref("network.http.pipelining.maxrequests" , 6);
pref("network.http.keep-alive.timeout", 600);
pref("network.http.max-connections", 6);
pref("network.http.max-connections-per-server", 4);
pref("network.http.max-persistent-connections-per-server", 4);
pref("network.http.max-persistent-connections-per-proxy", 4);

/* session history */
pref("browser.sessionhistory.max_total_viewers", 0);
pref("browser.sessionhistory.max_entries", 50);

Comment 8

[timeless]

pref("browser.cache.disk.parent_directory", "/media/mmc2/.mozilla/fennec");

Comment 9

[Mark Finkle (:mfinkle) (use needinfo?)]

Attachment: patch (10MB setting)

try 10MB and use the /media/mmc2 partition

Comment 10

[timeless]

Comment on attachment 380827 [details] [diff] [review]
patch (10MB setting)

please don't add the extra blank line before http prefs :)

Comment 11

[Mark Finkle (:mfinkle) (use needinfo?)]

Attachment: patch 2 (more caches)

This patch adds smaller settings for:
* image.cache.size (1MB instead of 5MB) bug 487061
* offline.cache.capacity (5MB instead of 500MB) bug 472341
* moves offline cache to mmc2 on maemo, like the first patch did for the disk.cache

Comment 12

[Mark Finkle (:mfinkle) (use needinfo?)]

Attachment: patch 3

This patch is the same as the previous, but is a bit more explicit for disk.cache on windows mobile. We don't need to turn off disk.cache for desktop, only winmo

Comment 13

[Mark Finkle (:mfinkle) (use needinfo?)]

https://hg.mozilla.org/mobile-browser/rev/5a33f5fa5f73

Comment 14

[Mark Finkle (:mfinkle) (use needinfo?)]

missed a semi-colon in the pref file, which breaks things (silently):
https://hg.mozilla.org/mobile-browser/rev/fcbe40a7894d

Comment 15

[Joel Maher ( :jmaher ) (UTC -8)]

these seem to allow for good browsing in beta3.  We still need to figure out how to not crash or lock up without swap enabled though