how many records a dataset can hold

Pavan_Yelugula · Post by **Pavan_Yelugula** » Thu Mar 03, 2005 12:00 am

hi
i have huge tables in my database records around 15millions. instead of always going and opening my table to get the required information. i am writing all the data to a Dataset(for one country at a time the dataset might be holding randomly around 6 millions records) and using it.

will this be a performance bottle neck.
how many records can i make my dataset to hold. where can i take a trade off between the dataset and a keeping a table and using it.

Can u also please tell me how many number of records i can let my files have to consider using them in the look-up stage.

Any inputs will be really helpful
Thanks and Regards
Pavan

ArndW · Post by **ArndW** » Thu Mar 03, 2005 2:00 am

Pavan,

15 Million records can easily be held in a dataset. If you still have a file system limited to 2Gb then your partitioning algorithm needs to ensure that no single file will exceed that size. This depends not only on the number of records but their data size as well.

nkreddy · Post by **nkreddy** » Thu Mar 03, 2005 4:10 pm

Arnd,

Would you please explain more on this..

Does that mean that if we have three partitions, we will have 2Gb times 3 or the 2Gb file is split evenly in all the three partitions.

Thanks,
NK

ArndW wrote:Pavan,

15 Million records can easily be held in a dataset. If you still have a file system limited to 2Gb then your partitioning algorithm needs to ensure that no single file will exceed that size. This depends not only on the number of records but their data size as well.

ArndW · Post by **ArndW** » Fri Mar 04, 2005 1:56 am

Pavan,

if you have an even split you will get (approximately) 2Gb for each partitioned file; assuming your OS and/or file system is still limited to 2Gb maximum file sizes; so with 3 partitions you can get up 6Gb of data. It would make sense to change your partition number to make sure that you won't come close to hitting that limit. In addition, the more files you have the more you can parallelize (is that a valid English verb?) intrinsically.

ray.wurlod · Post by **ray.wurlod** » Sat Mar 05, 2005 9:53 pm

No, it's an Americanism: "nouning" verbs and "verbing" nouns is a trait primarily encountered between the North Atlantic and Pacific oceans.
Ostensibly this is in the interests of efficient language. I disagree, and feel that they abase the language.

ArndW · Post by **ArndW** » Sun Mar 06, 2005 10:01 am

Ray, I agree with you that it is an abuse of English, but at least it is a non-cardinal language sin. Using telephone-shorthand (Cuz,Pls,u) and intentional mis-spellings in plain language English text is a crime punishable by either several hundreds years on the 9th plane of hell or having to watch a non-stop shopping TV channel while tied & bound to a chair like Alex was in Anthony Burgess' "A Clockwork Orange"