-
Notifications
You must be signed in to change notification settings - Fork 0
Netty 3 vs Netty 4 Perfromance Throughput tests
after years of happy production usage of netty-3.x we have decided to migrate to netty-4.x.
in general we are using netty for two main usecases:
- managing large amount of client connections coming-in from the interment to our data-center.
- managing internal cluster communication between our service to service interactions.
this leads us to be able to accept high volumes of TPS with low latency.
the benchmark test the ability of netty server to accepts Transactions Per Second (TPS) while we measure how effective it does so with resource utilization (CPU, Memory, GC activity)
the goal of the test is to bring each of the servers to 90% CPU and measure how much work they can process
the test is conducted in a lab of 5 linux CentOS with 4vCPU and 4GB Ram and over 1GB network since this is our default configuration in most of our production sites.
4 of the linux machines acts as netty clients generating high volumes of TPS 1 machines acts as server machine accepting TPS and calculating a 10 sec moving avg of TPS.
since we wanted to be sure we are using netty in the right way we created this project so everyone can review and contribute.