Search Mailing List Archives


Limit search to: Subject & Body Subject Author
Sort by: Reverse Sort
Limit to: All This Week Last Week This Month Last Month
Select Date Range     through    

[mininet-discuss] Throughput of user switch very less than ovs with mininet

Bob Lantz rlantz at cs.stanford.edu
Fri Oct 16 16:13:23 PDT 2015


1. Kernel vs. user switch means fewer data copies, and OVS uses other techniques to improve performance as well (its performance has increased 10x over the past few years).

2. The user switch seems to have dropped in performance on recent Linux, kernels - I’d really be interested in finding out why!

> On Oct 16, 2015, at 4:05 PM, Seyyed Hesamoddin Ghasemi <s.hesam.ghasemi at gmail.com> wrote:
> 
> I have same problem with OVS throughput, it exceeds from my real host eth0 speed (1Gbps). Could somebody plz clarify what is happening?
> 
> On Fri, Sep 18, 2015 at 8:05 AM, Purnima Mohan <purnimamohan50 at gmail.com <mailto:purnimamohan50 at gmail.com>> wrote:
> Dear all,
> 
> The mininet topology with ovs switches gives a maximum throughput (measured using iperf) of around 30Gbps for OVS switch. Whereas the same with a user switch is much lesser than that (around 300kbps). Could you please clarify the reason for such a vast difference in throughput between ovs and user switch on the same system?
> 
> OVS:
> #sudo mn --topo linear --controller remote --switch ovsk,protocols=OpenFlow13
> *** Creating network
> *** Adding controller
> *** Adding hosts:
> h1 h2 
> *** Adding switches:
> s1 s2 
> *** Adding links:
> (h1, s1) (h2, s2) (s2, s1) 
> *** Configuring hosts
> h1 h2 
> *** Starting controller
> c0 
> *** Starting 2 switches
> s1 s2 ...
> *** Starting CLI:
> mininet> h1 ping h2
> PING 10.0.0.2 (10.0.0.2) 56(84) bytes of data.
> 64 bytes from 10.0.0.2 <http://10.0.0.2/>: icmp_req=1 ttl=64 time=0.731 ms
> 64 bytes from 10.0.0.2 <http://10.0.0.2/>: icmp_req=2 ttl=64 time=0.049 ms
> 64 bytes from 10.0.0.2 <http://10.0.0.2/>: icmp_req=3 ttl=64 time=0.053 ms
> ^C
> Interrupt
> stopping h1 
> mininet> 
> mininet> iperf h1 h2
> *** Iperf: testing TCP bandwidth between h1 and h2 
> *** Results: ['31.3 Gbits/sec', '31.3 Gbits/sec']
> 
> 
> 
> User switch:
> #sudo mn --topo linear --controller remote --switch user,protocols=OpenFlow13*** Creating network
> *** Adding controller
> *** Adding hosts:
> h1 h2 
> *** Adding switches:
> s1 s2 
> *** Adding links:
> (h1, s1) (h2, s2) (s2, s1) 
> *** Configuring hosts
> h1 h2 
> *** Starting controller
> c0 
> *** Starting 2 switches
> s1 s2 
> *** Starting CLI:
> mininet> 
> mininet> h1 ping h2
> PING 10.0.0.2 (10.0.0.2) 56(84) bytes of data.
> 64 bytes from 10.0.0.2 <http://10.0.0.2/>: icmp_req=1 ttl=64 time=0.832 ms
> 64 bytes from 10.0.0.2 <http://10.0.0.2/>: icmp_req=2 ttl=64 time=0.557 ms
> 64 bytes from 10.0.0.2 <http://10.0.0.2/>: icmp_req=3 ttl=64 time=0.563 ms
> ^C
> Interrupt
> stopping h1 
> mininet> iperf h1 h2
> *** Iperf: testing TCP bandwidth between h1 and h2 
> *** Results: ['309 Kbits/sec', '319 Kbits/sec']
> 
> 
> -- 
> Thanks & Regards,
> Purnima 
> National University of Singapore
> 
> 
> _______________________________________________
> mininet-discuss mailing list
> mininet-discuss at lists.stanford.edu <mailto:mininet-discuss at lists.stanford.edu>
> https://mailman.stanford.edu/mailman/listinfo/mininet-discuss <https://mailman.stanford.edu/mailman/listinfo/mininet-discuss>
> 
> 
> 
> 
> -- 
> Seyyed Hesamoddin Ghasemi
> _______________________________________________
> mininet-discuss mailing list
> mininet-discuss at lists.stanford.edu
> https://mailman.stanford.edu/mailman/listinfo/mininet-discuss

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.stanford.edu/pipermail/mininet-discuss/attachments/20151016/4eb340fc/attachment.html>


More information about the mininet-discuss mailing list