On Apr 4, 2005 7:11 PM, David <thump@(protected):
> What are the largest pros and cons involved in discussing mysql versus
MySQL doesn't have the recoverability features of Oracle, you lose a
disk and you've lost everything since the last backup. WikiPedia
recently lost 2 days of updates when they lost a 'disk' (actually a
RAID arrary presented as a virtualdisk) with their MySQL data on, it
took them a day and a half to get up and running again. LiveJournal,
in January, didn't lose any data when their CoLo company lost power
but did take 3 days to get a working system back. They didn't lose
data because their middleware had features to make up for the lack in
MySQL. Still, a 3 day outage for a power loss of less than 1 hour is
really not good.
MySQL scaling is a joke. High volume sites (such as WikiPedia and
LiveJournal) scale by running many servers in paralell then
implementing bespoke replication and multiple master-slave (typically
reads go to the slaves and writes to the masters, periodically the
slaves pull updates from the Masters) clusters. This works well where
reads massively out number writes but quickly gets bottlenecked on
writes (as can be seen when trying to post a journal entry on
LiveJournal at peak times, it presents as a server busy error).
MySQL was originally developed by a supermarket (as I recall) for use
in daily reporting, essentially a datamart. The day's transactions
would be loaded in and summaries generated then the reports would be
run. If the database got fried then they still had the original files
so could rebuild it easily. All releases have carried that original
assumption forward. Not very suitable for OLTP use.
It's better to ask a silly question than to make a silly assumption.