It would be good to get the breakdown of those figures - the UnixBench
index is a blend of CPU + other benchmarks, including disk IO, so its
possible that the choke point is not actually the CPU, but some other
resource - for example ran the benchmark with 5 and 10 cpu constrained
LXD instances and on the host itself:
------------------------------------------------------------------------
Benchmark Run: Tue Nov 08 2016 12:19:39 - 12:47:35
5 CPUs in system; running 5 parallel copies of tests
Dhrystone 2 using register variables 165738484.4 lps (10.0 s, 7 samples)
Double-Precision Whetstone 12007.5 MWIPS (9.7 s, 7 samples)
Execl Throughput 19225.7 lps (29.9 s, 2 samples)
File Copy 1024 bufsize 2000 maxblocks 1197112.8 KBps (30.0 s, 2 samples)
File Copy 256 bufsize 500 maxblocks 289235.2 KBps (30.0 s, 2 samples)
File Copy 4096 bufsize 8000 maxblocks 3294183.5 KBps (30.0 s, 2 samples)
Pipe Throughput 9264864.7 lps (10.0 s, 7 samples)
Pipe-based Context Switching 1461401.2 lps (10.0 s, 7 samples)
Process Creation 68332.4 lps (30.0 s, 2 samples)
Shell Scripts (1 concurrent) 48816.8 lpm (60.0 s, 2 samples)
Shell Scripts (8 concurrent) 4978.0 lpm (60.0 s, 2 samples)
System Call Overhead 6466142.5 lps (10.0 s, 7 samples)
System Benchmarks Index Values BASELINE RESULT INDEX
Dhrystone 2 using register variables 116700.0 165738484.4 14202.1
Double-Precision Whetstone 55.0 12007.5 2183.2
Execl Throughput 43.0 19225.7 4471.1
File Copy 1024 bufsize 2000 maxblocks 3960.0 1197112.8 3023.0
File Copy 256 bufsize 500 maxblocks 1655.0 289235.2 1747.6
File Copy 4096 bufsize 8000 maxblocks 5800.0 3294183.5 5679.6
Pipe Throughput 12440.0 9264864.7 7447.6
Pipe-based Context Switching 4000.0 1461401.2 3653.5
Process Creation 126.0 68332.4 5423.2
Shell Scripts (1 concurrent) 42.4 48816.8 11513.4
Shell Scripts (8 concurrent) 6.0 4978.0 8296.7
System Call Overhead 15000.0 6466142.5 4310.8
========
System Benchmarks Index Score 5014.5
------------------------------------------------------------------------
Benchmark Run: Tue Nov 08 2016 10:31:29 - 10:59:29
10 CPUs in system; running 10 parallel copies of tests
Dhrystone 2 using register variables 328990072.0 lps (10.0 s, 7 samples)
Double-Precision Whetstone 24217.4 MWIPS (9.7 s, 7 samples)
Execl Throughput 29086.5 lps (29.9 s, 2 samples)
File Copy 1024 bufsize 2000 maxblocks 876587.0 KBps (30.0 s, 2 samples)
File Copy 256 bufsize 500 maxblocks 228999.4 KBps (30.0 s, 2 samples)
File Copy 4096 bufsize 8000 maxblocks 2615620.6 KBps (30.0 s, 2 samples)
Pipe Throughput 18575730.1 lps (10.0 s, 7 samples)
Pipe-based Context Switching 2800666.2 lps (10.0 s, 7 samples)
Process Creation 92266.1 lps (30.0 s, 2 samples)
Shell Scripts (1 concurrent) 62788.9 lpm (60.0 s, 2 samples)
Shell Scripts (8 concurrent) 9062.0 lpm (60.0 s, 2 samples)
System Call Overhead 7114922.6 lps (10.0 s, 7 samples)
System Benchmarks Index Values BASELINE RESULT INDEX
Dhrystone 2 using register variables 116700.0 328990072.0 28191.1
Double-Precision Whetstone 55.0 24217.4 4403.2
Execl Throughput 43.0 29086.5 6764.3
File Copy 1024 bufsize 2000 maxblocks 3960.0 876587.0 2213.6
File Copy 256 bufsize 500 maxblocks 1655.0 228999.4 1383.7
File Copy 4096 bufsize 8000 maxblocks 5800.0 2615620.6 4509.7
Pipe Throughput 12440.0 18575730.1 14932.3
Pipe-based Context Switching 4000.0 2800666.2 7001.7
Process Creation 126.0 92266.1 7322.7
Shell Scripts (1 concurrent) 42.4 62788.9 14808.7
Shell Scripts (8 concurrent) 6.0 9062.0 15103.4
System Call Overhead 15000.0 7114922.6 4743.3
========
System Benchmarks Index Score 6779.1
------------------------------------------------------------------------
Benchmark Run: Mon Nov 07 2016 23:12:39 - 23:40:36
10 CPUs in system; running 10 parallel copies of tests
Dhrystone 2 using register variables 328659187.7 lps (10.0 s, 7 samples)
Double-Precision Whetstone 24220.9 MWIPS (9.7 s, 7 samples)
Execl Throughput 29037.4 lps (29.9 s, 2 samples)
File Copy 1024 bufsize 2000 maxblocks 872910.1 KBps (30.0 s, 2 samples)
File Copy 256 bufsize 500 maxblocks 238079.9 KBps (30.0 s, 2 samples)
File Copy 4096 bufsize 8000 maxblocks 2694390.3 KBps (30.0 s, 2 samples)
Pipe Throughput 24258200.6 lps (10.0 s, 7 samples)
Pipe-based Context Switching 3361177.8 lps (10.0 s, 7 samples)
Process Creation 82523.0 lps (30.0 s, 2 samples)
Shell Scripts (1 concurrent) 61048.5 lpm (60.0 s, 2 samples)
Shell Scripts (8 concurrent) 8772.3 lpm (60.0 s, 2 samples)
System Call Overhead 7604993.0 lps (10.0 s, 7 samples)
System Benchmarks Index Values BASELINE RESULT INDEX
Dhrystone 2 using register variables 116700.0 328659187.7 28162.7
Double-Precision Whetstone 55.0 24220.9 4403.8
Execl Throughput 43.0 29037.4 6752.9
File Copy 1024 bufsize 2000 maxblocks 3960.0 872910.1 2204.3
File Copy 256 bufsize 500 maxblocks 1655.0 238079.9 1438.5
File Copy 4096 bufsize 8000 maxblocks 5800.0 2694390.3 4645.5
Pipe Throughput 12440.0 24258200.6 19500.2
Pipe-based Context Switching 4000.0 3361177.8 8402.9
Process Creation 126.0 82523.0 6549.4
Shell Scripts (1 concurrent) 42.4 61048.5 14398.2
Shell Scripts (8 concurrent) 6.0 8772.3 14620.5
System Call Overhead 15000.0 7604993.0 5070.0
========
System Benchmarks Index Score 7012.1
I need to repeat the test in the 5 core LXD container with 10 parallel
and see if that makes a difference.
--
You received this bug notification because you are a member of Ubuntu
Bugs, which is subscribed to Ubuntu.
https://bugs.launchpad.net/bugs/1639932
Title:
cpu constraints not being applied
To manage notifications about this bug go to:
https://bugs.launchpad.net/ubuntu/+source/nova-lxd/+bug/1639932/+subscriptions
--
ubuntu-bugs mailing list
ubuntu-bugs@lists.ubuntu.com
https://lists.ubuntu.com/mailman/listinfo/ubuntu-bugs
