Joinphobia
-
@Steve_The_Cynic said in Joinphobia:
@izzion said in Joinphobia:
@Steve_The_Cynic said in Joinphobia:
@Mason_Wheeler said in Joinphobia:
@dkf said in Joinphobia:
@LaoC said in Joinphobia:
Without an index to boot
Ooooh, where have I seen that before? Only in too many damn places "because making an index is expensive!!!".
*sigh*
If I had a nickel for every time I've heard "we don't want to index that because it will make INSERTs run slower"...
If the most common (by a factor of ten or more) is an insert, that attitude becomes slightly less stupid.
And before you say it, yes, I've worked with a system that operated on that basis. It was ... interesting.
Never underestimate the value of WMRN datasets
The thing I was specifically thinking of was a financial market activity data set that took high-frequency inserts and (relatively) low-frequency reads. Write performance was key...
Then you absolutely should not use any standard SQL database.
-
@Kamil-Podlesak said in Joinphobia:
@Steve_The_Cynic said in Joinphobia:
@izzion said in Joinphobia:
@Steve_The_Cynic said in Joinphobia:
@Mason_Wheeler said in Joinphobia:
@dkf said in Joinphobia:
@LaoC said in Joinphobia:
Without an index to boot
Ooooh, where have I seen that before? Only in too many damn places "because making an index is expensive!!!".
*sigh*
If I had a nickel for every time I've heard "we don't want to index that because it will make INSERTs run slower"...
If the most common (by a factor of ten or more) is an insert, that attitude becomes slightly less stupid.
And before you say it, yes, I've worked with a system that operated on that basis. It was ... interesting.
Never underestimate the value of WMRN datasets
The thing I was specifically thinking of was a financial market activity data set that took high-frequency inserts and (relatively) low-frequency reads. Write performance was key...
Then you absolutely should not use any standard SQL database.
It was a couple of steps down from the speed that would imply (and yes, the actual market data didn't use SQL). It was an alert-generating system, to handle things like "tell me when the price of IBM's shares goes above X dollars", and could generate bursts of activity, but wouldn't normally sustain really high write frequencies.
-
@Steve_The_Cynic That sounds like an ideal use case for something like Redis. Of course, it would then have trouble with any sort of correlation with different stocks or the price a year ago, but those would be much rarer queries for you.
-
@boomzilla said in Joinphobia:
Does anyone really use anything other than the dbo schema in SQL Server?
YES.. In fact our DRC [Design Rule Checks] flag any usage of the dbo schema! Schema should be specific to a business case (large scale) so that associated tables, procs, et. al. are grouped in meaningful ways.
-
@dkf Could be, except that I was working on that before the first release of Redis...