|
Andy Brummer wrote: inserting 35,000 records a second for 22 hours straight
Inserting nearly 3 billion records is going to take a while no matter what you do.
Andy Brummer wrote: it makes the difference between inserting 5 million records in a few minutes or
in a few days.
I don't think it would cause a 1,000x slowdown. 10x, maybe... but 1,000x? That doesn't sound right.
|
|
|
|
|
I am only speaking from experience here: Inserting large numbers of data at a high rate of speed into the middle of a table with seemingly random clustered index can really hurt performance. I've done it in the past, rate: ~10-30k rows per second on a table ~ 100m to 1b rows in size. Using a clustered index, and the pushing the data to the table in the order of the clustered index made queries nearly unusable or in the very least lengthy, however when doing a time-series based search, on a table with a time series clustered index was fine and allowed for a realtime monitoring solution.
"Life should not be a journey to the grave with the intention of arriving safely in a pretty and well preserved body, but rather to skid in broadside in a cloud of smoke, thoroughly used up, totally worn out, and loudly proclaiming "Wow! What a Ride!"
— Hunter S. Thompson
|
|
|
|
|
Well, the clustered index you talked about is certainly not ideal and is less performant than it should be. I supposed the magnitude of the problem depends on the exact amount of data you are using and the frequency of inserts/updates/reads/deletes. I guess the "ultimate" fail here would be that the others you work with didn't find the problem sooner.
|
|
|
|
|
A lot of the time has to do with speculative reads missing and the random read write access on the disks pushing them from < 10ms access times to over 100ms. SQL server optimizes it's accesses for overall throughput and that can drive latency to ridiculous levels. As much as it doesn't sound like it should be a big deal. I can tell you from real world experience it is.
|
|
|
|
|
I don't know why this was down voted, hopefully I corrected it enough.
"Life should not be a journey to the grave with the intention of arriving safely in a pretty and well preserved body, but rather to skid in broadside in a cloud of smoke, thoroughly used up, totally worn out, and loudly proclaiming "Wow! What a Ride!"
— Hunter S. Thompson
|
|
|
|
|
I'm not saying exactly where this came from (to protect the guilty), but it was a question on this very site. (I have also changed the field and variable names)
A code fragment you may appreciate:
float result = 0F;
SqlCommand cmd = new SqlCommand("Select SUM(myField) From myTable Where myOtherField = 'Value'", con);
if (cmd.ExecuteScalar().ToString() != "0" && cmd.ExecuteScalar().ToString() != "")
result= float.Parse(cmd.ExecuteScalar().ToString().Trim());
Three lines of code; How many don't-do-its can you spot?
Two extra database accesses
Three unnecessary int-to-string conversions
One unnecessary Trim operation (with the output guaranteed to equal the input)
One unnecessary Parse operation
Six unnecessary string creations
Real men don't use instructions. They are only the manufacturers opinion on how to put the thing together.
Digital man: "You are, in short, an idiot with the IQ of an ant and the intellectual capacity of a hose pipe."
|
|
|
|
|
I think I work with this guy.... Just kidding, but I do know a developer who insisted on always doing this sort of this. He was like a little kitten. He was always compelled to use them.
I wasn't, now I am, then I won't be anymore.
|
|
|
|
|
Marcus Kramer wrote: I think I work with this guy
If you do, could you give him a slap round the back of the head, and tell him it's from me?
Real men don't use instructions. They are only the manufacturers opinion on how to put the thing together.
Digital man: "You are, in short, an idiot with the IQ of an ant and the intellectual capacity of a hose pipe."
|
|
|
|
|
You have no idea how much I would like to accommodate you on this one, but I'm not ready to retire quite yet and I don't think I could stop at just one, so I can't make it look like an accident.
I wasn't, now I am, then I won't be anymore.
|
|
|
|
|
You know you want to... 5!
Real men don't use instructions. They are only the manufacturers opinion on how to put the thing together.
Digital man: "You are, in short, an idiot with the IQ of an ant and the intellectual capacity of a hose pipe."
|
|
|
|
|
...so how would you write that bit of code?
OriginalGriff wrote: Three unnecessary int-to-string conversions
There are 0 int-to-string conversions in the code you posted
OriginalGriff wrote: One unnecessary Parse operation
Need to get it to a float somehow, and parse will only take a string. You could use ...
(float)Convert.ToDouble(cmd.ExecuteScalar());
...but not sure how good that is for performance, or perhaps take the chance of a straight cast (but could be risky)...
float result = (float)cmd.ExecuteScalar();
EDIT:
What deserves a vote of 1? the fact that somebody failed to note that ExecuteScalar return an object not an int? or the fact that people don't like other people not being in complete agreement?
Illogical thoughts make me ill
|
|
|
|
|
More like this, I didn't do any tests, so I'm sure it can still be improved
var cmd = new SqlCommand("Select SUM(myField) From myTable Where myOtherField = 'Value'", con);
var resultString = cmd.ExecuteScalar().ToString();
var result = (!string.IsNullOrEmpty(result) & !result.Equals("0"))
? (float)System.Convert.ToDouble(resultString)
: 0F;
|
|
|
|
|
get rid of the var's and then we can talk
certainly an improvement thou you would be wasting your time checking for result.Equals("0") to end up setting it to 0 anyway, also Convert.ToDouble() will return 0 for empty string anyway (meant that for a null object). Problem is if the result is non-numeric. That's why I would lean more to a (try)parse
SqlCommand cmd = new SqlCommand("...", con);
float result = 0;
float.TryParse(cmd.ExecuteScalar().ToString(), out result);
return result;
I am not saying this is the best best way, but it covers null, empty and non-numeric values
EDIT: as I have correctly been corrected, I should test for null before using ToString() a shameful mistake...
SqlCommand cmd = new SqlCommand("...", con);
float result = 0;
float.TryParse(cmd.ExecuteScalar() AS string, out result);
return result;
GETTING THERE...
SqlCommand cmd = new SqlCommand("...", con);
float result = 0;
object cmdResult = cmd.ExecuteScalar();
if(cmdResult != null)
float.TryParse(cmdResult.ToString(), out result);
return result;
...of course _Erik_ has the better (and smaller) amount of code
Don't vote my posts down just because you don't understand them - if you lack the superior intelligence that I possess then simply walk away
modified on Wednesday, March 2, 2011 7:10 AM
|
|
|
|
|
I was copying the logic of the OP, like I said, I hadn't tested anything. But reducing the DB access from 3 to 1, will make a HUGE performance gain.
ps. 'var' is gold. I use it everywhere since my variable names are clear, and I refactor a lot. And not having to change types saves a lot of time.
|
|
|
|
|
Yep I agree reducing the DB access is definitely the best performance optimisation.
GibbleCH wrote: I use it everywhere since my variable names are clear
That's fair enough, I just prefer having the type declared at the start as it makes easier scanning IMO
Don't vote my posts down just because you don't understand them - if you lack the superior intelligence that I possess then simply walk away
|
|
|
|
|
musefan wrote: float.TryParse(cmd.ExecuteScalar().ToString(), out result);
That will throw a null reference exception if ExecuteScalar returns null.
|
|
|
|
|
just testing
...perhaps you should put it in a new hall of shame post
Don't vote my posts down just because you don't understand them - if you lack the superior intelligence that I possess then simply walk away
|
|
|
|
|
Oh, no... We all make this kind of mistakes from time to time. However, the way you have corrected it would not work either. In this case ExecuteScalar would return null or a boxed value type, so checking its result with "as string" would always return null. I you don't use a nullabe type here you would have to use an object and check for null before trying to make any conversion, I mean:
object obj = cmd.ExecuteScalar();
if (obj != null)
I my opinion, nullable types are much cleaner for this.
|
|
|
|
|
musefan wrote: ...so how would you write that bit of code?
musefan wrote: Need to get it to a float somehow
int? result = (int?)cmd.ExecuteScalar();
return (float)result.GetValueOrDefault();
There is no need for esoteric conversions, Parsing or ToString at all. Actually, there is no need for an if in the original code posted, and if ExecuteScalar returns null, the original code would throw a null reference exception.
musefan wrote: What deserves a vote of 1?
I don't know, I did not downvote you.
Edit: Where you see "int?" I put "float?" before, and that was a mistake.
modified on Wednesday, March 2, 2011 6:18 AM
|
|
|
|
|
I think you where better of with float, what if SUM returns a double value? Plus int to float can be implicitly cast
I get what you are saying thou. I am not sure on the exact ins and outs of the performances of casting to nullable and using GetValueOrDefault() over an if null statement but I doubt it would be much either way so preference will get the job done
Don't vote my posts down just because you don't understand them - if you lack the superior intelligence that I possess then simply walk away
|
|
|
|
|
Yup, you are right, I was better with "float?". I did not read the query (as usual, I am soooo absentminded). The nullable type to use must be the one which fits with the one defined into the database, I mean, if the query is defined to return a float you cannot use "double?", becouse the unboxing operation would fail. Anyway, GetValueOrDefault is a constant complexity operation, so in these cases I usually choose the solution which gives me a cleaner implementation.
|
|
|
|
|
_Erik_ wrote: The nullable type to use must be the one which fits with the one defined into the database
I think a nullable float would still do the trick, even if the DB defined the value as decimal or numeric or int or float then the explicit case to float would still work, and as that is your retuning value anyway you need to cast to float at some point
Don't vote my posts down just because you don't understand them - if you lack the superior intelligence that I possess then simply walk away
|
|
|
|
|
No, that is not right. Just try this:
object obj = 1;
float? f = (float?)obj;
This would throw an exception. The value type boxed into the object in this case is an int, so you need an (int) or (int?) casting to unbox it. Any other type would throw the same exception. Remember we first have to unbox the value type, and the compatibility rules among numeric types are not applied for boxing-unboxing operations.
|
|
|
|
|
oh yes, correct again sir... lesson learnt
Don't vote my posts down just because you don't understand them - if you lack the superior intelligence that I possess then simply walk away
|
|
|
|
|
Straight out of a C# Unified Communications code sample from MS:
if (string.IsNullOrEmpty(customMessage.Trim()))
...
|
|
|
|