EMG benchmark results Print
Sunday, 26 November 2006

EMG is available in a benchmark edition which can be used to measure EMG performance on different hardware platforms.

The actual benchmark is performed by injecting a message (140 octets) into EMG which is then "looped" between an incoming and outgoing connector of the same protocol type. After a runtime of 60 seconds the server is stopped and the throughput calculated. Since each message is disassembled when received by EMG and then re-assembled when sent out the throughput seen can be used as an indicator of maximum possible throughput.

Only one EMG instance on used and network connections are performed via "localhost" so it is an ideal environment with no network latency etc which would of course be present in a real-life deployment of EMG.

Other factors in the EMG configuration that affects performance is "message persistence" and database logging.

Performance figures (in mps, messages per second).

Dell PowerEdge 1850 (2 CPUs)

CPU: 2 x Intel Xeon 3.0 GHz
RAM: 1 GB
OS: RedHat Enterprise Linux 4

EMG build: EMG 3.0.8.8716/20060207
Protocol: SMPP

Throughput: 4331 mps

Sun Fire V210 (1 CPU)

CPU: 1 x UltraSparc IIIi 1.0 GHz
RAM: 2 GB
OS: Solaris 10

EMG build: EMG 3.0.8.8716/20060207
Protocol: SMPP

Throughput: 1927 mps

Sun Fire V210 (2 CPUs)

CPU: 2 x UltraSparc IIIi 1.0 GHz
RAM: 2 GB
OS: Solaris 10

EMG build: EMG 3.0.8.8716/20060207
Protocol: SMPP

Throughput: 2188 mps

Sun Fire V20z (1 CPU)

CPU: 1 x AMD Opteron 2.2 GHz
RAM: 2 GB
OS: Solaris 10 (x86)

EMG build: EMG 3.0.8.8716/20060207
Protocol: SMPP

Throughput: 5544 mps