[QA][nova][Concurrent performance]
Hello everyone!
Why the performance of Stein concurrent creation of VM is not as good as that of Ocata? Create 250 VMs concurrently, the O version only needs 160s, but the S version needs 250s. Security_group and port_security functions are disabled. Among 250 VMs, there are single network card and multi-network card. Using neutron-openvswitch-agent.
Regards, Cuihx
Hi
on 2019/9/2 15:43, 崔恒香 wrote:
Why the performance of Stein concurrent creation of VM is not as good as that of Ocata? Create 250 VMs concurrently, the O version only needs 160s, but the S version needs 250s.
Openstack's execution environment is complicated. It's hard to say O is faster than S from the architecture viewpoint. Are you sure two systems' running environment are totally the same for comparision? including the software, hardware, network connection etc.
regards.
Hi
At 2019-09-02 15:58:00, "Wesley Peng" weslepeng@gmail.com wrote:
Hi
on 2019/9/2 15:43, 崔恒香 wrote:
Why the performance of Stein concurrent creation of VM is not as good as that of Ocata? Create 250 VMs concurrently, the O version only needs 160s, but the S version needs 250s.
Openstack's execution environment is complicated. It's hard to say O is faster than S from the architecture viewpoint. Are you sure two systems' running environment are totally the same for comparision? including the software, hardware, network connection etc.
The same set of servers, after testing version O, re-install the system, and then test version S. First deployed version O, the operating system is Ubuntu 16.04. Then deploy the version S, the operating system is ubuntu18.04.
The port_security function is disabled.But in the S version environment, adding a lot of flows on br-int. Does this slow down the creation of VM?
The flows is as follows:(for port qvo83b4285e-b5) cookie=0x9ca1d4c6ecdcb31f, duration=257335.826s, table=0, n_packets=0, n_bytes=0, priority=10,icmp6,in_port="qvo83b4285e-b5",icmp_type=136 actions=resubmit(,24) cookie=0x9ca1d4c6ecdcb31f, duration=257335.823s, table=0, n_packets=19, n_bytes=798, priority=10,arp,in_port="qvo83b4285e-b5" actions=resubmit(,24) cookie=0x9ca1d4c6ecdcb31f, duration=257335.831s, table=0, n_packets=95, n_bytes=10680, priority=9,in_port="qvo83b4285e-b5" actions=resubmit(,25) cookie=0x9ca1d4c6ecdcb31f, duration=257335.829s, table=24, n_packets=0, n_bytes=0, priority=2,icmp6,in_port="qvo83b4285e-b5",icmp_type=136,nd_target=fe80::f816:3eff:fe39:1601 actions=resubmit(,60) cookie=0x9ca1d4c6ecdcb31f, duration=257335.826s, table=24, n_packets=19, n_bytes=798, priority=2,arp,in_port="qvo83b4285e-b5",arp_spa=30.0.1.180 actions=resubmit(,25) cookie=0x9ca1d4c6ecdcb31f, duration=257335.841s, table=25, n_packets=114, n_bytes=11478, priority=2,in_port="qvo83b4285e-b5",dl_src=fa:16:3e:39:16:01 actions=resubmit(,60)
Regards, Cuihx
Hi, When we did our testing a few months ago we saw the same thing with the degradation of the performance. Keystone was a notable bottleneck. Thanks Gary
________________________________ From: Wesley Peng weslepeng@gmail.com Sent: Monday, September 2, 2019 10:58 AM To: openstack-discuss@lists.openstack.org openstack-discuss@lists.openstack.org Subject: Re: [QA][nova][Concurrent performance]
Hi
on 2019/9/2 15:43, 崔恒香 wrote:
Why the performance of Stein concurrent creation of VM is not as good as that of Ocata? Create 250 VMs concurrently, the O version only needs 160s, but the S version needs 250s.
Openstack's execution environment is complicated. It's hard to say O is faster than S from the architecture viewpoint. Are you sure two systems' running environment are totally the same for comparision? including the software, hardware, network connection etc.
regards.
participants (3)
-
Gary Kotton
-
Wesley Peng
-
崔恒香