Snapshot Isoation

Question

Hi Everyone,
In the documentation related to snapshot isolation level it is mentioned that for better performance with MVCC
1) larger cache size should be used
2) shorter transacitons should be used.
This question is focused on second aspect.
In my application, There are two types of transactions,
a) first one is normal transaction started without any type of argument. This, we use for all write activities to DB.
b) other type of transaction is opened as snapshot transaction. This is used for all read activities.
I need to implement a couple of use cases, where I need to read full database in a single snapshot transaction. Therefore my question is that when we talk about keeping transactions short, is it write transactions or snapshot transactions. I know that ideally I should keep all the transactions short but this would help me in implementing the abovementioned functionality.
Thanks,
Shishir

Answer 1

Thanks a lot Ashok for this really nice explanation.
This precisely answers my question.
I have another question which is on snapshot isolation again. It will be great if you can answer that.
My application is basically user driven so I don't have much control on number of concurrent operations going on. I am using snapshot and default transactions both.
Few days back, on two different occasions, I had freezer files accumulated and following that database environment became really slow. I increased the cache and restarted environment and its behaving fine now.
My questions is that both the times db_stat -e looked fine (98% cache hit ratio etc). So what is the best way to find that this situtaion is about to happen ?
I know that its possible that there is no simple answer to this question but this will also help me with application footprint sizing. I can send you db_stat -e output for one of these incidents. Am pasting -m part of it here.
Thanks,
Shishir
===============================
512MB     Total cache size
1     Number of caches
1     Maximum number of caches
512MB     Pool individual cache size
0     Maximum memory-mapped file size
0     Maximum open file descriptors
0     Maximum sequential buffer writes
0     Sleep after writing maximum sequential buffers
0     Requested pages mapped into the process' address space
366M     Requested pages found in the cache (98%)
5662872     Requested pages not found in the cache
28885     Pages created in the cache
5662887     Pages read into the cache
212793     Pages written from the cache to the backing file
6115927     Clean pages forced from the cache
1631     Dirty pages forced from the cache
0     Dirty pages written by trickle-sync thread
126503     Current total page count
126314     Current clean page count
189     Current dirty page count
65537     Number of hash buckets used for page location
377M     Total number of times hash chains searched for a page (377476452)
10     The longest hash chain searched for a page
787M     Total number of hash chain entries checked for page (787470619)
420893     The number of hash bucket locks that required waiting (0%)
6973     The maximum number of times any hash bucket lock was waited for (5%)
90M     The number of region locks that required waiting (52%)
9962     The number of buffers frozen
1502     The number of buffers thawed
0     The number of frozen buffers freed
6387030     The number of page allocations
2856M     The number of hash buckets examined during allocations (2856267125)
65873     The maximum number of hash buckets examined for an allocation
6117798     The number of pages examined during allocations
13     The max number of pages examined for an allocation
928725     Threads waited on page I/O
Pool File: MetadataDB
4096     Page size
0     Requested pages mapped into the process' address space
366M     Requested pages found in the cache (98%)
5662872     Requested pages not found in the cache
28885     Pages created in the cache
5662887     Pages read into the cache
212793     Pages written from the cache to the backing file
=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=
22049/8823442     File/offset for last checkpoint LSN
Fri Sep 11 13:00:36 2009     Checkpoint timestamp
0x804ed335     Last transaction ID allocated
500000     Maximum number of active transactions configured
295     Active transactions
675     Maximum active transactions
5165877     Number of transactions begun
336     Number of transactions aborted
5165246     Number of transactions committed
4784     Snapshot transactions
34220     Maximum snapshot transactions
0     Number of transactions restored
185MB 24KB     Transaction region size
215222     The number of region locks that required waiting (1%)