[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Xen-users] Xen & netperf


  • To: Tommy Wang <subscription@xxxxxxxxxxx>
  • From: Fasiha Ashraf <feehapk@xxxxxxxxxxx>
  • Date: Sat, 5 Sep 2009 12:23:33 +0530 (IST)
  • Cc: xen-users@xxxxxxxxxxxxxxxxxxx
  • Delivery-date: Fri, 04 Sep 2009 23:54:33 -0700
  • Domainkey-signature: a=rsa-sha1; q=dns; c=nofws; s=s1024; d=yahoo.co.in; h=Message-ID:X-YMail-OSG:Received:X-Mailer:Date:From:Subject:To:Cc:In-Reply-To:MIME-Version:Content-Type; b=KAQPTri21rhKZSpIt3iSdDJPczU0YlJCKWYrtNIzXcgbS4mnuMqGjh4z2W6i7XPvIa6hw+k80VrEjciBIrixasnK4OeWC7IosndbzpYgyjdD15nNkfcif5iKtsuCnudmqlvhN4uMI9519bEUjqk7pZTuVHUZGR1wRPHkTYKih9s=;
  • List-id: Xen user discussion <xen-users.lists.xensource.com>

I have set-up xen on my Intel quad core server. Now running different experiment to measure network throughput in virtualized environment. these are some of the results:

Netperf-4.5 results for inter-domain communication.

Sr.No.              Client               Server               Time(sec)        Throughput(Mbps)

   1                   Guest-1              Dom0                    180                            2398 - 2500
   2                     Dom0             Guest-1                   180                            1636 - 1883    3                     Dom0               Dom0                    180                                 2200
  4                     Guest-1           Guest-1                   180                                 1921
  5                     Guest-1           Guest-2                   180                            0.14 - 0.29

You can see throughput is ok in all cases else Guest to Guest communication.
If anyone has some idea, please share. What could be the reason?
Some more details about my set-up:
Guests are PV domUs of Fedora11(32bit). They are on the same host on same physical machine.
Sr.No.              Client               Server               Time(sec)        Throughput(Mbps)

   1                   Guest-1              Dom0                    180                            2398 - 2500
   2                     Dom0             Guest-1                   180                            1636 - 1883   
OS/Xen version you use on dom0?
Dom0 OS is also Fedora11.
[root@HPCNL-SR-2 ~]# xm info
host                   : HPCNL-SR-2
release                : 2.6.30-rc6-tip
version                : #1 SMP Tue Aug 25 12:26:16 PKST 2009
machine                : i686
nr_cpus                : 8
nr_nodes               : 1
cores_per_socket       : 4
threads_per_core       : 1
cpu_mhz                : 1995
hw_caps                : bfebfbff:20100000:00000000:00000140:000ce33d:00000000:00000001:00000000
virt_caps              : hvm
total_memory           : 8186
free_memory            : 2
node_to_cpu            : node0:0-7
node_to_memory         : node0:2
xen_major              : 3
xen_minor              : 4
xen_extra              : .2-rc1-pre
xen_caps               : xen-3.0-x86_32p hvm-3.0-x86_32 hvm-3.0-x86_32p
xen_scheduler          : credit
xen_pagesize           : 4096
platform_params        : virt_start=0xf5800000
xen_changeset          : Thu Aug 20 08:30:24 2009 +0100 19740:7a57911ff09e
cc_compiler            : gcc version 4.4.0 20090506 (Red Hat 4.4.0-4) (GCC)
cc_compile_by          : root
cc_compile_domain      : (none)
cc_compile_date        : Tue Aug 25 09:55:27 PKST 2009
xend_config_format     : 4

Please guide what could be the problem? Why is the guest to guest throughput is that low.

Regards
Fasiha Ashraf

--- On Fri, 4/9/09, Tommy Wang <subscription@xxxxxxxxxxx> wrote:

From: Tommy Wang <subscription@xxxxxxxxxxx>
Subject: [Xen-users] Xen & netperf
To: xen-users@xxxxxxxxxxxxxxxxxxx
Date: Friday, 4 September, 2009, 10:35 PM

First, I apologize if this message has been received multiple times.
I'm having problems subscribing to this mailing list:

Hi xen-users,

I am trying to decide whether I should run a game server inside a Xen
domain.  My primary reason for wanting to virtualize is because I want
to isolate this environment from the rest of my server.  I really like
the idea of isolating the game server from the rest of the service on
the box.  There will only be a single virtual machine, and I am for
all practical purposes, the only person with access to the box.

Anyway, the game server is more network intensive than CPU intensive,
and that will be my primary criteria for decided whether I virtualize.

I ran some naive benchmarks with netperf on my Dom0 (debian lenny w/
xen 3.2.1), DomU, and my Linux box at home.  Dom0 and DomU are
connected by a public network (100 Mbps link) and a private network (1
Gbps link).  All netperf tests were run with defaults (w/o any extra
options).

Dom0 to Dom0 (local)
TCP STREAM TEST from 0.0.0.0 (0.0.0.0) port 0 AF_INET to () port 0
AF_INET : demo
Recv   Send    Send
Socket Socket  Message  Elapsed
Size   Size    Size     Time     Throughput
bytes  bytes   bytes    secs.    10^6bits/sec

87380  16384  16384    10.00    5859.30

DomU to DomU (local)
TCP STREAM TEST from 0.0.0.0 (0.0.0.0) port 0 AF_INET to () port 0
AF_INET : demo
Recv   Send    Send
Socket Socket  Message  Elapsed
Size   Size    Size     Time     Throughput
bytes  bytes   bytes    secs.    10^6bits/sec

87380  16384  16384    10.00    3482.77

Dom0 to DomU (via a 100 Mbps uplink)
TCP STREAM TEST from 0.0.0.0 (0.0.0.0) port 0 AF_INET to () port 0
AF_INET : demo
Recv   Send    Send
Socket Socket  Message  Elapsed
Size   Size    Size     Time     Throughput
bytes  bytes   bytes    secs.    10^6bits/sec

87380  16384  16384    10.09      91.37

DomU to Dom0 (via a 100 Mbps uplink)
TCP STREAM TEST from 0.0.0.0 (0.0.0.0) port 0 AF_INET to () port 0
AF_INET : demo
Recv   Send    Send
Socket Socket  Message  Elapsed
Size   Size    Size     Time     Throughput
bytes  bytes   bytes    secs.    10^6bits/sec

87380  16384  16384    10.10      91.11

Dom0 to DomU (via a 1 Gbps uplink - private network)
TCP STREAM TEST from 0.0.0.0 (0.0.0.0) port 0 AF_INET to () port 0
AF_INET : demo
Recv   Send    Send
Socket Socket  Message  Elapsed
Size   Size    Size     Time     Throughput
bytes  bytes   bytes    secs.    10^6bits/sec

87380  16384  16384    10.00    1621.90

DomU to Dom0 (via a 1 Gbps uplink - private network)
TCP STREAM TEST from 0.0.0.0 (0.0.0.0) port 0 AF_INET to () port 0
AF_INET : demo
Recv   Send    Send
Socket Socket  Message  Elapsed
Size   Size    Size     Time     Throughput
bytes  bytes   bytes    secs.    10^6bits/sec

87380  16384  16384    10.00    8655.93

Home to Dom0 (Home network is 1 Mbps up)
TCP STREAM TEST from 0.0.0.0 (0.0.0.0) port 0 AF_INET to () port 0
AF_INET : demo
Recv   Send    Send
Socket Socket  Message  Elapsed
Size   Size    Size     Time     Throughput
bytes  bytes   bytes    secs.    10^6bits/sec

87380  16384  16384    13.81       0.96

Home to DomU (Home network is 1 Mbps up)
TCP STREAM TEST from 0.0.0.0 (0.0.0.0) port 0 AF_INET to () port 0
AF_INET : demo
Recv   Send    Send
Socket Socket  Message  Elapsed
Size   Size    Size     Time     Throughput
bytes  bytes   bytes    secs.    10^6bits/sec

87380  16384  16384    12.26       0.96

Dom0 to Home (Home network is 10 Mbps down)
TCP STREAM TEST from 0.0.0.0 (0.0.0.0) port 0 AF_INET to () port 0
AF_INET : demo
Recv   Send    Send
Socket Socket  Message  Elapsed
Size   Size    Size     Time     Throughput
bytes  bytes   bytes    secs.    10^6bits/sec

87380  16384  16384    10.85      16.44

DomU to Home (Home network is 10 Mbps down)
TCP STREAM TEST from 0.0.0.0 (0.0.0.0) port 0 AF_INET to () port 0
AF_INET : demo
Recv   Send    Send
Socket Socket  Message  Elapsed
Size   Size    Size     Time     Throughput
bytes  bytes   bytes    secs.    10^6bits/sec

87380  16384  16384    10.85      11.03

So, what do these results mean?  Of course, my home network connection
is asymmetrical, so I can guess why any connection with my home
computer was limited at 10 or 1 Mbps speeds.  And, for the 100 Mbps
link between Dom0 to DomU, I can understand a pretty consistent 91
Mbps.  For testing, I included the 1 Gbps private link between Dom0
and DomU, but I will not be using that link for the server.

Does this mean that despite any performance from running netperf from
within a Xen virtual machine, that since my uplink to the Internet is
only 100 Mbps it doesn't really matter (since, both Dom0 and DomU were
pushing the 100 Mbps throughput)?  Are there any other benchmarks that
I should run before making my decision?

Hope someone can help me interpret these results!

Thanks,
Tommy

_______________________________________________
Xen-users mailing list
Xen-users@xxxxxxxxxxxxxxxxxxx
http://lists.xensource.com/xen-users


See the Web's breaking stories, chosen by people like you. Check out Yahoo! Buzz.
_______________________________________________
Xen-users mailing list
Xen-users@xxxxxxxxxxxxxxxxxxx
http://lists.xensource.com/xen-users

 


Rackspace

Lists.xenproject.org is hosted with RackSpace, monitoring our
servers 24x7x365 and backed by RackSpace's Fanatical Support®.