Mark Roberts wrote:
> - It'd be nice if the query planner was more "stable" - sometimes the
> queries run fast, and then sometimes they randomly take 2 hours for a
> delete that normally runs in a couple of minutes.
I was going to stay silent, because my pet peeves were already covered or had been fixed (btw, thanks to whomever fixed
sqlstandard "quote escaping a quote" all those years ago :-) ). But Mark's suggestion is excellent. Plan stability /
Storedplanner outlines / whatever you want to call it, is hugely valuable when data volumes change so frequently that
theplanner never knows the "good" stats from the "bad", and also when upgrading to lessen the "OMG, I have to add set
enable_nestloop=falseto 48 billion queries just to overcome new planner quirks" situations. $OTHER_BIG_RDBMS have had
thisto varying degrees for a while (stored outlines/plan stability in Oracle; bind in DB2; whatever crap name MS gave
theirhalf-arsed version), and when it's mature, the certainty around execution is a life-saver.
And just to chime in on the already mentioned things:
- in-place upgrades
- replication engine in the core
- true stored procedures
- job scheduler in the core
In all, a short list, which is an oblique way of saying thanks to everyone for the enormous strides that have been made
inthe last few years :-)
Ciao
Fuzzy
:-)
------------------------------------------------
Dazed and confused about technology for 20 years
http://fuzzydata.wordpress.com/