[Openstack-operators] Download speed limit on instances

Salman Toor salman.toor at it.uu.se
Tue Dec 11 13:30:06 UTC 2012


Hi, 

Recently I have deployed a test setup using one controller node and 3 compute nodes. I am using flat dhcp. 

I am not a network guy. 

on my compute node's nova.conf, the network section looks like following

# NETWORK
network_manager=nova.network.manager.FlatDHCPManager
force_dhcp_release=True
dhcpbridge_flagfile=/etc/nova/nova.conf
dhcpbridge = /usr/bin/nova-dhcpbridge
firewall_driver=nova.virt.libvirt.firewall.IptablesFirewallDriver
# Change my_ip to match each host
my_ip=130.238.17.162
public_interface=em1
flat_network_bridge=br100
flat_interface=em1
fixed_range=10.0.0.0/24
root_helper = sudo nova-rootwrap 

On the compute node (real machine) the max download speed that I can achieve is 24.6MB, whereas on openstack instances its never goes beyond 4MB and these the top speed. I am using curl to get an estimate. So its the average of last 5 seconds download speed. 

Instances are not routed from the controller node. 

i mean if I use tranceroute command 

[sztoor at sl-wn0 ~]$ traceroute google.com 
traceroute to google.com (173.194.32.4), 30 hops max, 60 byte packets
 1  10.0.0.3 (10.0.0.3)  0.265 ms  0.132 ms  0.177 ms
 2  r1.n.it.uu.se (130.238.17.1)  5.513 ms  5.583 ms  5.994 ms
 3  * * *
 4  * * *

10.0.0.3 is the ip on the br100 on compute node. 

[root at sal3 sztoor]# ifconfig br100
br100: flags=4163<UP,BROADCAST,RUNNING,MULTICAST>  mtu 1500
        inet 10.0.0.3  netmask 255.255.255.0  broadcast 10.0.0.255
        inet6 fe80::7ca7:4ff:fed0:a181  prefixlen 64  scopeid 0x20<link>
        ether 00:1a:64:62:41:03  txqueuelen 0  (Ethernet)
        RX packets 60311200  bytes 125300407382 (116.6 GiB)
        RX errors 0  dropped 22  overruns 0  frame 0
        TX packets 69367004  bytes 54110674615 (50.3 GiB)
        TX errors 0  dropped 0 overruns 0  carrier 0  collisions 0


Currently I am running 5 instances on this compute node. 

Can someone please help me to understand why instances are slow. One important thing, There is no network activite going on while these measurements. Is there some limit set somewhere? or I need to set some parameter? or something else? 

Or its a simple math that if max on real machine goes upto 24MB and there are 5 instances running on the machine than each of the instance should never go beyond 4.8MB .... 


Regards..
Salman. 
 
Salman Toor
salman.toor at it.uu.se



-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openstack.org/pipermail/openstack-operators/attachments/20121211/1cba2cd2/attachment.html>


More information about the OpenStack-operators mailing list