Performance Benchmarking CPU

Tool Used: sysbench, Phoronix Test Suite

First Instance ECS on Cloud:

root@iZj6c5vmzbgujq3xi2m45aZ:~# phoronix-test-suite system-info

Phoronix Test Suite v5.2.1

System Information

Hardware:

Processor: Intel Xeon E5-2682 v4 @ 2.50GHz (1 Core), Motherboard: Alibaba Cloud ECS, Chipset: Intel 440FX- 82441FX PMC, Memory: 1 x 1024 MB RAM, Disk: 40GB, Graphics: Cirrus Logic GD 5446, Network: Red Hat Virtio device

Software:

OS: Ubuntu 18.04, Kernel: 4.15.0-45-generic (x86_64), File-System: ext4, Screen Resolution: 1024x768, System Layer: KVM

Second similar Instance on Azure Cloud:

root@testbenchmark:~# phoronix-test-suite system-info

Phoronix Test Suite v5.2.1

System Information

Hardware:

Processor: Intel Xeon Platinum 8171M @ 2.10GHz (1 Core), Motherboard: Microsoft Virtual Machine v7.0, Chipset: Intel 440BX/ZX/DX, Memory: 1 x 1024 MB Microsoft, Disk: 32GB Virtual Disk + 4GB Virtual Disk, Graphics: Microsoft Hyper-V virtual VGA

Software:

OS: Ubuntu 18.04, Kernel: 5.0.0-1035-azure (x86_64), File-System: ext4, Screen Resolution: 1152x864, System Layer: Microsoft Hyper-V Server

	ECS	Azure VM
Processor	Intel Xeon E5-2682 v4 @ 2.50GHz (1 Core)	Intel Xeon Platinum 8171M @ 2.10GHz (1 Core)
Motherboard	Alibaba Cloud ECS	Microsoft Virtual Machine v7.0
Chipset	Intel 440FX- 82441FX PMC	Intel 440BX/ZX/DX
Memory	1 x 1024 MB RAM	1 x 1024 MB Microsoft
Disk	40GB	32GB Virtual Disk + 4GB Virtual Disk
Graphics	Cirrus Logic GD 5446	Microsoft Hyper-V virtual VGA
Network	Red Hat Virtio device
OS	Ubuntu 18.04	Ubuntu 18.04
Kernel	4.15.0-45-generic (x86_64)	5.0.0-1035-azure (x86_64)
File-System	ext4	ext4
Screen Resolution	1024x768	1152x864
System Layer	KVM	Microsoft Hyper-V Server

NOTE: The test and comparison here is very high level and the purpose here is to show how I do usually benchmarking and reporting.

CPU workload

Test1:

When running with the CPU workload, sysbench will verify prime numbers by doing standard division of the number by all numbers between 2 and the square root of the number. If any number gives a remainder of 0, the next number is calculated. As we can imagine, this will put some stress on the CPU, but only on a very limited set of the CPUs features.

The benchmark can be configured with the number of simultaneous threads and the maximum number to verify if it is a prime. I have chosen 2 thread here.

On Alibaba Cloud:

root@iZj6c5vmzbgujq3xi2m45aZ:~# sysbench cpu --cpu-max-prime=20000 --threads=2 run

sysbench 1.0.11 (using system LuaJIT 2.1.0-beta3)

Running the test with following options:

Number of threads: 2

Initializing random number generator from current time

Prime numbers limit: 20000

Initializing worker threads...

Threads started!

CPU speed:

events per second: 320.26

General statistics:

total time: 10.0053s

total number of events: 3205

Latency (ms):

min: 2.65

avg: 6.24

max: 668.54

95th percentile: 6.79

sum: 19994.63

Threads fairness:

events (avg/stddev): 1602.5000/47.50

execution time (avg/stddev): 9.9973/0.00

The number to verify with other systems (given the same or similar parameters) is given by the General statistics:

total time: 10.0053s

total number of events: 3205

The number of ‘events per second’ should be used as a performance indicator instead of the total execution time.

CPU Metric during this test.

On Azure VM:

root@testbenchmark:~# sysbench cpu --cpu-max-prime=20000 --threads=2 run

sysbench 1.0.11 (using system LuaJIT 2.1.0-beta3)

Running the test with following options:

Number of threads: 2

Initializing random number generator from current time

Prime numbers limit: 20000

Initializing worker threads...

Threads started!

CPU speed:

events per second: 322.58

General statistics:

total time: 10.0049s

total number of events: 3228

Latency (ms):

min: 3.01

avg: 6.20

max: 22.58

95th percentile: 7.43

sum: 20005.16

Threads fairness:

events (avg/stddev): 1614.0000/3.00

execution time (avg/stddev): 10.0026/0.00

Result: total time is approximately same in both machine.

Test2:

Another CPU test using Phoronix Test Suite. PTS is the most comprehensive testing and benchmarking platform available for Linux.

# No of test available

hasnain@benchmarktest:~$ phoronix-test-suite list-available-tests | wc -l

376

# To get no of test for Processor

phoronix-test-suite list-available-tests | grep Processor

Smallpt is a C++ global illumination renderer written in less than 100 lines of code. Global illumination is done via unbiased Monte Carlo path tracing and there is multi-threading support via the OpenMP library.

On ECS: Test Results Uploaded To:
https://openbenchmarking.org/result/2004045-KH-ALIBABACP05

On Azure Cloud: Test Results Uploaded To:
https://openbenchmarking.org/result/2004041-KH-CPUAZURET20

Result: Score is approximately same in both machine.

https://openbenchmarking.org/result/2004045-KH-ALIBABACP05,%202004041-KH-CPUAZURET20