[Dev] Chandler 0.7 perf goals
Katie Capps Parlante
capps at osafoundation.org
Sat Feb 4 15:30:36 PST 2006
Hi Heikki and Sheila,
I think we can identify 3 different ways to frame the performance tests
that we ought to be focusing on in 0.7...
1) A set of use cases and target times that represent "acceptable"
performance for our 0.7 target user. This means being clear about our
target user and the set of 0.7 use cases. It means identifying a target
data size (target calendar size, target # of tasks, that sort of thing.)
I don't think we're quite ready to do that, as Sheila mentioned, but
we're close and should put this on the list of things we need from design.
2) Use cases/tests that are consistent with the previous release so that
we can compare progress across releases. As Heikki mentioned, we want to
avoid regressions (unless there is some good reason) and improve somewhat.
3) Tests that push the size requirement up an order of magnitude, so
that we feel confident that we are making progress to a 1.0 goal of
large amounts of email and other data. (It would be great to formalize
this 1.0 goal, that should be part of working out the 1.0 roadmap, which
we can focus on after 0.7 planning has been flushed out).
I think we should try pick a suite of tests and targets for 0.7 that
cover all 3 of these needs.
Heikki, your list covers #2. Perhaps we could trim a few tests:
- New event creation (file menu) and (in place) could become one test
for both empty repository and 3000 item calendar
- Cut the test that creates a new calendar on empty repository
I don't think we need to keep thinking of that set of 9 tests as our
focus tests -- the tests we write to cover (1) should have that role.
We could also proceed with adding a test or two that exercises a bigger
repository, some 10,000 item test. I'm not sure that a 10,000 item
calendar test is appropriate, though. Perhaps we could add the email
test that Andi wrote (or a variant), or revive the old RSS test.
I'm not sure that I agree with your target time changes, but want to
mull this over before giving you a counter proposal.
Heikki Toivonen wrote:
> Sheila Mooney wrote:
>> Heikki, is this meant to be only calendar related workflows or all the
>> performance goals for 0.7?
> As I mentioned in the first message on Tuesday, we will be creating new
> tests as well. So no, it is not the complete list of performance goals.
> My purpose was to finalize the numbers for the existing tests, but I see
> my wording was a little misleading.
>> A major tenet for 0.7 is now around the table and task management and we
>> haven't really discussed what (if any) new use cases we should be
>> handling. Should we be trying to optimize for more than 3000 items? We
>> had talked about that back in the fall when we put together the 1.0
>> stickie plan. Also, I know we aren't intending on having a "usable"
>> dashboard for 0.7 but it might be worthwhile to at least start tracking
>> some table related use cases such as searching and sorting.
> Good, this is the kind of feedback I was hoping for.
> And since we haven't discussed that much at all, I'm calling off the
> Monday deadline.
> I would prefer to keep the existing 3000 events tests we have because
> they give us historical data on how we are doing, and create new tests
> for other cases. However, there is a thing about too many tests as well
> (takes too long to run the suite) so we need to strike a balance.
> _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _
> Open Source Applications Foundation "Dev" mailing list
More information about the Dev