Hello Friends,

I apologies about I am asking questions here not related OVS. I am
sure I will get an answer of my question here.

Recently I am doing an experiment with my friend on KVM live
migration. While doing this we encountered some network speed issue. I
thought I should share this with you all and get your input on this.
It will help us to move further and to do deep
study.

We have two servers Host 1 and 2 both installed KVM, directly attached
to TP-Link TL-SG1016 16-port unmanaged Gigabyte switch

Host 1 --->TP-Link Giga Switch<------ Host 2

We run the command:  "tcpdump -i eth0   -w blabla.pcap" inside both
Host 1 and Host 2. We are trying to migrate a VM (copy-storage-all)
from Host 1 to Host 2. The network speed is around 600Mbps.

We found below issues,

First, it turned out that tcpdump slows down the network speed.
Second, Host 1 captured 4.8G data while Host 2 captured 5.3G data.
Third, just FYI, ifconfig or kernel's stats will rotate the network
transferred data every 4G.

We tried IPtraf, got the same issue. The source counts less than the
destination. But if we switch a 100M switch, the source outgoing data
will be exactly the same as the destination incoming data.

Here we pretty much need to handle two issues: high speed, and support
large data counts. Like migrating a 40G image, KVM will migrate at
least more than 40G data over the network. How can we capture the
total byte count of the data flow, that's our problem?

If you could propose a better solution for this, I can try it. Also I
will try with tcpstats.

Thank you all in advance for your time and consideration

After this I am going to try this with OVS :)

Regards,
Hitesh Wadekar
_______________________________________________
dev mailing list
[email protected]
http://openvswitch.org/mailman/listinfo/dev

Reply via email to