pgbench on 6c 2013 mac pro, internal SSD: 35,000+ iops, 12430 transactions per second.

iops

# make a new DB, in this case at ~/tmpdb, which is on the mac pro's internal SSD (disk0)

% ./bin/initdb -D ~/tmpdb


# set checkpoint_segments to 32 in postgres.conf, then start DB.

% ./bin/postgres -D /Users/andre/tmpdb
LOG:  database system was shut down at 2014-07-03 11:49:38 PDT
LOG:  database system is ready to accept connections
LOG:  autovacuum launcher started

# initialize the DB for use by pgbench, with a scale factor of 200. pgbench is probably not installed, but it's in the PostgreSQL source directory under contrib/pgbench.

% cd contrib/pgbench ; ./pgbench -i -s 200
creating tables...
100000 of 20000000 tuples (0%) done (elapsed 0.07 s, remaining 14.68 s).
200000 of 20000000 tuples (1%) done (elapsed 0.15 s, remaining 15.20 s).
300000 of 20000000 tuples (1%) done (elapsed 0.23 s, remaining 14.99 s).
...
19800000 of 20000000 tuples (99%) done (elapsed 25.81 s, remaining 0.26 s).
19900000 of 20000000 tuples (99%) done (elapsed 26.06 s, remaining 0.13 s).
20000000 of 20000000 tuples (100%) done (elapsed 26.31 s, remaining 0.00 s).
vacuum...
set primary keys...
done.


# This results in a DB of about 4.3 GB

% du -sh ~/tmpdb
4.3G	/Users/andre/tmpdb


# Run pgbench for 10 seconds with 14 clients (processes), using two threads each. Vacuum all tables prior to starting the benchmark.

% ./pgbench -T 10 -c 14 -j 2 -v
starting vacuum...end.
starting vacuum pgbench_accounts...end.
transaction type: TPC-B (sort of)
scaling factor: 200
query mode: simple
number of clients: 14
number of threads: 2
duration: 10 s
number of transactions actually processed: 124927
tps = 12430.767431 (including connections establishing)
tps = 12449.323445 (excluding connections establishing)

# During the above run, iostat -w1 disk0:
          disk0       cpu       load average
    KB/t tps   MB/s    us sy id   1m   5m   15m
    7.83  23   0.18     0  0 100  3.70 2.85 2.49
# vacuum starts
   24.85 2103  51.04    2  2 96   3.70 2.85 2.49
   22.37 8219  179.54   4  5 91   3.56 2.83 2.49
   26.29 7066  181.41   5  5 90   3.56 2.83 2.49
   29.14 5294  150.63   4  4 92   3.56 2.83 2.49
   25.04 6700  163.83   5  5 91   3.56 2.83 2.49
   20.10 9888  194.14   4  6 90   3.56 2.83 2.49
   23.17 7880  178.28   5  5 90   3.44 2.82 2.49
   14.23 33987 472.15   4 11 86   3.44 2.82 2.49
   67.12 402   26.32    8  2 90   3.44 2.82 2.49
   17.34 8536  144.53  52 31 17   3.44 2.82 2.49
# vacuum finishes, benchmark starts
   14.31 14173 198.02  58 36  6   3.44 2.82 2.49
   13.18 22618 291.23  54 36 10   3.72 2.89 2.51
   14.13 17489 241.35  56 35  9   3.72 2.89 2.51
   14.95 14119 206.07  58 36  6   3.72 2.89 2.51
   14.66 13962 199.90  58 35  7   3.72 2.89 2.51
   14.41 14194 199.79  58 35  6   3.72 2.89 2.51
   15.02 13946 204.61  58 35  7   3.66 2.89 2.52
   12.08 35218 415.34  53 38  9   3.66 2.89 2.52
   15.28 13411 200.08  61 33  6   3.66 2.89 2.52
   13.76 1906  25.61   7  10 83   3.66 2.89 2.52
# benchmark finishes