Benchmarking

May 12, 2017

Tools for Stress Testing your RIG

Some nice tools to stress-test your computer:

CPU: Prime95
GPU: FurMark
RAM: MemTest86+
HDD/SSD: S.M.A.R.T

Ref: http://www.pcworld.com/article/2028882/keep-it-stable-stupid-how-to-stress-test-your-pc-hardware.html

Networking

1. Connected my [laptop] and [desktop] via 5meter cat5e cable -> 1000mbit full duplex link
2. created ~1.5G ramdisks with tmpfs on each.
3. created a 1 giga file with data from /dev/urandom on [desktop]
4. copied file with scp from [desktop] to [laptop]
5. copied file with scp from [laptop] to [desktop]
6. repeated 4 & 5 10 times : average speed ~70+MB/s from desk & 80+MB/s to desk, link stays up all the time. ping around 0.2-0.3ms
7. repeated 1-6 with my other 3meter ethernet cable.
8. tried both direction simultanously which gave a total data throughput even over 100MB/s

Hard Disk

To get some information about the disk run:

Request identification info directly from the drive, which is displayed in a new expanded format with considerably more detail than with the older -i option.

$ hdparm -I /dev/sda

Timings

$ hdparm -tT /dev/sda

Clear Cache

$ echo 3 > /proc/sys/vm/drop_caches 

I am using this block-size and count which results in ~2GB:

$ dd [...] bs=1048576 count=2048

READ:

$ dd if=test of=/dev/null
$ dd if=/dev/sdX of=/dev/null

WRTIE:

$ dd if=/dev/zero of=test
$ dd if=/dev/zero of=/dev/sdX

smartmontools:

seatage: 
Raw_Read_Error_Rate
Seek_Error_Rate  

example:
% python
>>> 200009354607 & 0xFFFFFFFF
2440858991 <---- total number
>>> (200009354607 & 0xFFFF00000000) >> 32
46 <--- number of errors