diff options
| author | Mason S | 2010-05-10 14:31:44 +0000 |
|---|---|---|
| committer | Pavan Deolasee | 2011-05-19 16:38:45 +0000 |
| commit | b95b8c84c1f2b5cf5f9c55c5b4cb3448771660e2 (patch) | |
| tree | 21c78916e960c643ec58abccf43c171b851e33ba /src/include/postgres.h | |
| parent | 09b657378bf282096f9b690b8d7155377aedff44 (diff) | |
Modified pgbench for Postgres-XC.
By default, it will use branch id to distribute by, and also include it in the queries.
This allows for better multi-node scalability. For this configuration, we do not
have to worry about implicit two phase commits occurring.
If running in either initialization or execution mode, alternatively a new option
-k may be used (-k = "key" = no branch id).
If included with -i, tables are distributed according to their primary key.
This means we will see a lot of two phase commit activity over multiple data nodes.
Similarly, if -k is included when running tests, branchid (bid) will NOT
be used as part of the DML statements. This helps us measure the impact of 2PC.
Diffstat (limited to 'src/include/postgres.h')
0 files changed, 0 insertions, 0 deletions