Hi All ,
I am a bit confused regarding how to increase this UDP_HTABLE_SIZE. I mean
where can i find this file to increase it size? I am using UBUNTU Linux 9.10
Karmic. I have increased the kernel timer frequency to 1000 Hz already but
how to do this change regarding to increase the UDP_HTABLE_SIZE?
Can anyone kindly help me in this issue , how to resolve this?
Regards ,
Sougata
On Mon, Oct 25, 2010 at 5:11 PM, Simonart, Xavier <xavier.simon...@intel.com
> wrote:
> Hi,
>
>
>
> I have not tried running IMS Bench SIPP with many IP per SIPP instance.
>
> But the following scenario works for us: one IP per SIPP instance, 4 SIPP
> instances, and 25K ports per SIPP instance. Hence you would need 4 IPs (or
> virtual IPs). This way you can simulate 100K users per system, each users
> being represented with its own IP/port combination.
>
> As mentioned in the doc, as you use many users per system, you might have
> to change UDP_HTABLE_SIZE in include/net/udp.h and rebuild the kernel of the
> system running the IMS Bench SIPP, for performance reasons.
>
>
>
> -Xavier
>
>
> ------------------------------
>
> *From:* Peter Higginson [mailto:plh...@hotmail.com]
> *Sent:* lundi 25 octobre 2010 14:34
> *To:* sougata....@gmail.com
> *Cc:* sipp_users
>
> *Subject:* Re: [Sipp-users] Fwd: SIPp IMS Bench : Query on Virtual IP
> Config on Test System
>
>
>
> Sougata,
>
> It is a couple of years since I worked on this. If you want to generate
> traffic apparently from 100K real users, then you need to have that traffic
> appear to come from 100K different IP addresses. If you are not bothered
> about emulating all the IP addresses then there is no problem - SIPp will
> use a small number of IP addresses and ports and distinguish the calls using
> the Call-ID. You can measure the space used per Call-ID and work out how big
> a machine you need to avoid disk swapping. (We needed to go to 1G for 100K
> calls.)
>
> The problems only come with large numbers of IP addresses. I mentioned some
> below and also many devices have a limit of 64K on the size of the ARP
> table. We ended up writing our own code to emulate being a router with lots
> of IP addresses behind it. That way we only needed a few ARP entries (both
> for the test device and the DUT) and we could keep a small indexed table for
> the IP addresses.
>
> Hope that helps,
> Peter
>
> ------------------------------
>
> Date: Mon, 25 Oct 2010 12:48:39 +0200
> Subject: Re: [Sipp-users] Fwd: SIPp IMS Bench : Query on Virtual IP Config
> on Test System
> From: sougata....@gmail.com
> To: plh...@hotmail.com
> CC: sipp-users@lists.sourceforge.net
>
> Hi Peter ,
>
> Can you elaborate a bit more regarding , why i need to be lucky to make
> this set-up work?
> I have decided to go for 1 IP address per SIPp instance. So as i want to
> have 2 SIPp instances , i want to have 2 virtual IP's configure on the
> network interface card. This 2 SIPp instances needs to generate 100,000
> users each.
>
> As stated in my first mail , i have connected both the PC's with Cat 5
> cable. I hope this set-up works.
>
> Btw , how did you solve this issue in your case , as you said being outside
> of SIPp.?
>
> Regards ,
>
> Sougata
>
> On Mon, Oct 25, 2010 at 11:52 AM, Peter Higginson <plh...@hotmail.com>
> wrote:
>
>
> My personal experience with the base SIPp (rather than IMS bench) is that
> you will be very lucky to make this work. If you are thinking of multiple
> instances, then it is fairly easy to measure the data space per instance (or
> per call) and multiply it by the number of instances and see if you have
> enough real memory to do that. In either case, each IP address and each UDP
> port has an overhead in the OS and they will mount up. We found that Linux
> just slowed to a crawl with more than 20K IP addresses and UDP ports in use.
> There is also a C library limit (10K from memory) that limits the number of
> streams per process (you can recompile the library but we never went there).
>
> We did solve the problem, but outside of SIPp.
>
> Peter
>
> ------------------------------
>
> Date: Mon, 25 Oct 2010 11:18:00 +0200
> From: sougata....@gmail.com
>
>
> To: sipp-users@lists.sourceforge.net
>
> Subject: [Sipp-users] Fwd: SIPp IMS Bench : Query on Virtual IP Config on
> Test System
>
>
>
> HI All ,
>
> Any SUGGESTION on my this query??
> ITS A BIT URGENT..... !!!
>
> Regards ,
>
> Sougata
>
> ---------- Forwarded message ----------
> From: *sougata pal* <sougata....@gmail.com>
> Date: Fri, Oct 22, 2010 at 7:23 PM
> Subject: [Sipp-users] SIPp IMS Bench : Query on Virtual IP Config on Test
> System
> To: sipp-users@lists.sourceforge.net
>
>
> Hi All ,
>
>
> I want to configure multiple IP address on my Test System which will
> generate traffic for near about 200,000 users for my SUT where i have
> installed OpenIMSCore. Now as i see the pre-requisites in the IMS Bench SIPp
> installation procedure there its written " The actual number of IP addresses
> to configure will depend on the transport option you select: a single IP
> address per SIPp instance, in which case you need many IP addresses as
> you'll run SIPp instances on a same physical system, or multiple IP
> addresses per SIPp instance in which case you will want plenty of IP
> addresses."
>
> Now in my Test System i want to have 2 SIPp instances , each of whom will
> be generating 100,000 users traffic.
>
> So , now my question is which transport option will be suitable for me?
> Shall i go for 1 IP address per SIPp instance , or i keep only one single
> SIPp instance having 100,000 IP addresses?
>
> Secondly , connecting both the Pc's , my TEST SYSTEM and my Desktop running
> OPenIMSCore ( SUT) , i am using CAT 5 cable connector. Will that be suitable
> for conencting? I have installed 2 separate network adapters for this case
> in both the PC's.
>
> Any suggestion on this scenario will be very helpful ,
>
> Regards ,
>
> Sougata
>
>
>
>
>
> ------------------------------------------------------------------------------
> Nokia and AT&T present the 2010 Calling All Innovators-North America contest
> Create new apps & games for the Nokia N8 for consumers in U.S. and Canada
> $10 million total in prizes - $4M cash, 500 devices, nearly $6M in marketing
> Develop with Nokia Qt SDK, Web Runtime, or Java and Publish to Ovi Store
> http://p.sf.net/sfu/nokia-dev2dev
> _______________________________________________ Sipp-users mailing list
> Sipp-users@lists.sourceforge.net
> https://lists.sourceforge.net/lists/listinfo/sipp-users
>
>
> Intel Corporation NV/SA
> Kings Square, Veldkant 31
> 2550 Kontich
> RPM (Bruxelles) 0415.497.718.
> Citibank, Brussels, account 570/1031255/09
>
> This e-mail and any attachments may contain confidential material for the
> sole use of the intended recipient(s). Any review or distribution by others
> is strictly prohibited. If you are not the intended recipient, please
> contact the sender and delete all copies.
>
------------------------------------------------------------------------------
Nokia and AT&T present the 2010 Calling All Innovators-North America contest
Create new apps & games for the Nokia N8 for consumers in U.S. and Canada
$10 million total in prizes - $4M cash, 500 devices, nearly $6M in marketing
Develop with Nokia Qt SDK, Web Runtime, or Java and Publish to Ovi Store
http://p.sf.net/sfu/nokia-dev2dev
_______________________________________________
Sipp-users mailing list
Sipp-users@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/sipp-users
