References

Parsec Performance Test Result

Server Machine Spec:

Summary:  Dell R410, 2 x Xeon E5620 2.40GHz, 23.6GB / 24GB 1333MHz DDR3, 1 x 500GB SATA
System:       Dell PowerEdge R410 (Dell 01V648), C-2N/24/500, ySPEC 25.0
Processors:   2 x Xeon E5620 2.40GHz, 5.86GT QPI (HT enabled, 8 cores, 16 threads) - Gulftown B1, 64-bit, quad-core, 32nm, L3: 12MB
Memory:       23.6GB / 24GB 1333MHz DDR3 == 6 x 4GB - 4GB PC3-10600 Samsung DDR3-1333 ECC Registered CL9 2Rx8

Jetty setting:

java_opts: -Dparsec.conf.env.context=dev.conf
maxThreads: 500

Test Scenario:

validation filter enable
config file size: 2.6k

average response time

average successful qps

CPU idle

However, after turn on the cookie filter, the performance drop dramatically.

average response time

average successful qps

CPU usage

Git Repositories

Core

Supporting Library

Parsec Client Performance Test Result

Client Machine Spec:

Summary:  Dell R410, 2 x Xeon E5620 2.40GHz, 23.6GB / 24GB 1333MHz DDR3, 1 x 500GB SATA
System:   Dell PowerEdge R410 (Dell 01V648), C-2N/24/500, ySPEC 25.0
Processors:   2 x Xeon E5620 2.40GHz, 5.86GT QPI (HT enabled, 8 cores, 16 threads) - Gulftown B1, 64-bit, quad-core, 32nm, L3: 12MB
Memory:   23.6GB / 24GB 1333MHz DDR3 == 6 x 4GB - 4GB PC3-10600 Samsung DDR3-1333 ECC Registered CL9 2Rx8

Server Machine Spec:

Summary:      Dell R410, 2 x Xeon E5530 2.40GHz, 23.1GB / 24GB 1333MHz DDR3, 1 x 500GB SATA
System:       Dell PowerEdge R410 (Dell 0N051F), C-2N/24/500, ySPEC 25.0
Processors:   2 x Xeon E5530 2.40GHz, 5.86GT QPI (HT enabled, 8 cores, 16 threads) - Gainestown D0, 64-bit, quad-core, 45nm, L3: 8MB
Memory:       23.1GB / 24GB 1333MHz DDR3 == 6 x 4GB - 4GB PC3-10600 Hynix DDR3-1333 ECC Registered CL9 2Rx4

Test Scenario:

Number of client: 1
Number of server: 1
Response size: 1k

Ning vs Parsec vs Parsec Status Retry Enabled

YhdrRequestFilter

HeaderBackpostRequestFilter

Sequential