Question : Partition in Sybase for record insertion & select query

Hi, I am new to Sybase. We need to create Partition in Sybase for record insertion.I am guessing that, we can count number of records & if it exceeds the certain threshold limit, we can rename the current table to something like table_001 & we can create new table with the table name to continue the process of inserting the record. Please advise & correct me on this. Also Once the partition is done, how can we fetch the records using select statement without giving the partition table name? For example, if the record is in table_005, do i need to specify partition name table_005 in select query? Or do i need to give only table name in selectquery(Not Partition name). Does Sybase take care of itself in finding the correct parition where my record exists which is in where clause of the select query? Please give me more insight on this,Thanks in advance,Thanks,ram  

Answer : Partition in Sybase for record insertion & select query

Sybase IQ 12.7 does not support partitioning per se.  It does, however have special support and optimizations in views for combining tables as if they were partitions.

You can do what you want not by renaming the table but by creating your "partition" tables and then layering a view on top.  You probably need at least two views; one for the "current" parition you want to load data into and one that has the UNION of all the partitions.

There are a couple of issues to take into account here.  First, you have to decide why you are splitting up the table.  If you are doing it because your data volumes are very large and you are trying to minimize the load time caused by index reorganization in pass 2, then this technique works great.  For example, you can create a table for each month and keep the most recent 6 months in the Union view.  At the end of the 7th month, you change the view to include the next table and drop the oldest month table from the view.  You can also truncate it to purge old data or keep the table around as an archive of sorts so that you can go back to the old data yet it does not effect the performance of loads and queries on recent data.

If you are expecting it to reduce your query times, you may be dissapointed, particularly if you are running on a smaller machine with just a few CPUs/licenses.  Since the query engine and optimizer have no idea what records are in what tables, you are actually running your query on each one of the underlying tables and then corellating the results.  This can be counter productive unless you can create a view that refers to only some of the paritition tables; say 6 months worth out of two years total in the database.

IQ 15, currently in early beta, does support semantic paritioning.  The question remains as to whether Sybase will do with this feature what they did with it in ASE and charge an additional license fee for it.

Finally, I hope you are not trying to load the tables with Insert statements.  The performance of that technique leaves a lot to be desired given the column store nature of Sybase IQ.  By far the fastest and most flexible way to get data into IQ is with the LOAD TABLE statement.  Depending on what you are doing, it can be anywhere from 10 to 1000 times faster than straight SQL Insert statements.

Regards,
Bill