C
ChUIMonster
Guest
I disagree about the discussion being moot. There is much to be learned even if you cannot act on things immediately. For instance -- using FIELDS reduces the size of the data but it doesn't really speak to how the marshaling and de-marhsalling of the data is managed. There are many ways to do that and some of them have performance implications. Saying that you are using FIELDS and that, therefore, the transfer between the app server and the client cannot be a problem is nonsense. Likewise -- deciding to try a dump & load based on fragmentatiion in your test database rather than the metrics from production would be ill-advised. BTW -- ProTop can easily show you CRUD stats per table and index by user. That would go a long ways toward eliminating the noise. Don't take this wrong but, from a performance analysis POV, you seem to be consistently focusing on minutiae and silver bullets. That's not unusual but it is also not going to solve your problem (except accidentally).
Continue reading...
Continue reading...