Just somethings a few of us have talked about in the past few days:

1. createrepo --noxml - generates ONLY the sqlite dbs, no xml at all

2. yum --break-up-transaction: Takes a resolved yum transaction and
attempts to break it up into as many discrete, dependency-closed sets as
possible. Think about them like microtransactions which ultimately
create the overall desired transaction. One potential benefit of this is
it is done in a smaller chunk to minimize memory usage

3. Making sure, after we have depsolved, that our package objects drop
access to data like filelists, additional prcos and changelogs. Again to
minimize memory use

4. This ties in with rpm: working on a true sql-based schema for the
rpmdb to be stored in sqlite. The first step of this would be to write a
converter from the rpmdb to the schema devised above. The second step
would be writing a yum plugin to fill in the transaction data into the
sqlite db during a transaction, to see how it would impact performance
during the transaction.

5. outputting more stats information about how long certain processes
take in a little more obvious way. IE: setup repos, setup rpmdb, setup
package sacks, setup update lists, depsolve, transaction test,
transaction run w/ simple -d 3 outputs so we can see where people are
seeing sluggishness.

Anyone hate/love some/all of these ideas?
-sv

-- 
I only speak for me.

_______________________________________________
Yum-devel mailing list
[email protected]
https://lists.dulug.duke.edu/mailman/listinfo/yum-devel

Reply via email to