VoIP QoS Tools

ccnxjr

Just wondering what tools you guys use to test QoS in your LAN environments.
Specifically I'm looking at QoS for VoIP, so the metrics I'm looking for are Throughput, Jitter, Delay and Packet Loss.


Right now i'm using iperf, but I think that just works one way, it provides the stats I need but not sure if it's a true test of how VoIP traffic will flow .
I've read the documentation on SIPp and find it a little bit of a challenge to get it to do what I want, so if anyone has any experience with it that would be awesome!

Of course i'm looking for free/opensource tools, i've seen the testkits that are in the thousands of dollars, not in my budget at all.

The take away from the testing process is supposed to be how many simultaneous calls our current setup can handle without appreciable loss of quality.

Thanks all,

it_consultant

ccnxjr wrote: »

Just wondering what tools you guys use to test QoS in your LAN environments.
Specifically I'm looking at QoS for VoIP, so the metrics I'm looking for are Throughput, Jitter, Delay and Packet Loss.


Right now i'm using iperf, but I think that just works one way, it provides the stats I need but not sure if it's a true test of how VoIP traffic will flow .
I've read the documentation on SIPp and find it a little bit of a challenge to get it to do what I want, so if anyone has any experience with it that would be awesome!

Of course i'm looking for free/opensource tools, i've seen the testkits that are in the thousands of dollars, not in my budget at all.

The take away from the testing process is supposed to be how many simultaneous calls our current setup can handle without appreciable loss of quality.

Thanks all,

We all want an F-250 Super Duty at the price of a Kia. You should be able to determine how much bandwidth one SIP call takes and multiply appropriately to determine how much your ethernet cabling can handle. SIP is a pretty efficient protocol and you should find that on a gigabit network you will really need to put a lot of calls on the network to degrade quality.

If you problem is trunking between sites or to the POTS network, this is a completely different animal.

Lizano

I just do extended ping with TOS value (ping -v) matching that of what I'm trying to test. I don't really trust those apps that do MOS tests and give you weird jitter stats and stuff. I have used them an gotten poor results at sites where I have 50+ phone users that never have an issue.

Are you looking to test where QoS is working on the LAN? Or do you have QoS links with the carrier and want to see if the carrier is passing along the tagged packets?

ccnxjr

We have an internal PBX system, that is running on a virtual machine, for reasons beyond my payscale .
Of course, we need to factor in some processing overhead and IP table rules.
I'm not wholly privy to the design details, but suffice it to say there is some additional processing overhead internally.
Therefore scaling observed wire traffic for one call over wire capacity won't give a very accurate measure of quality.
There are some formulas that we've factored in , but ultimately my boss would like to run a "stress test", just to guage the accuracy of our calculations.
Thats kinda why I was looking at SIPp in the first place, i'd like to simulate X-number of calls passing through our LAN, with some way of measuring the call quality at the end point.
I *can* manually place a call, get a packet capture at the end point and then review the jitter/delay/loss using Wireshark's telephony tools, hell, it will even replay the audio so i can hear the quality!
However, I'm sure for larger VoIP environments this kind of testing is automated.
I'm aware that SIPp can be configured as a UAC and UAS so you can send SIP messages from one box to another and gather stats on rec'd SIP messages.
I'm also aware that it can be used to simulate a call with audio, but i've had not much luck in getting that setup and running.

SIPp is just the one package I've found that promises these things, if there are others packages out there or a different methodology I'm open.

Thanks a bunch,

it_consultant

I see why this is tricky. With all the major phone vendors they tell US what their equipment is capable of handling with regards to processors, internal memory, etc. If you are running a soft switch in VMWARE your possible variables for poor phone call quality are many. There are other things too, like if an incoming call that hits a voicemail server for logic processing (it is 9 AM you get this menu, it is 9 PM you get another menu) where the voicemail server is experiencing performance issues. Additionally you may be in a situation where all your calls are recorded by the phone switch (QA purposes etc) and that process is causing delay and jitter.

Before you start troubleshooting the bit layer, I think it is important to get a really good idea of the whole design before you beat your head against the wall for too long.

Lizano

I'm with IT consultant on that, sometimes is better to go back and look at the design. If you are looking for something similar to SIPp, I think sipvicious has some features that might be useful, thought I'm not sure its any better than SIPp.

ChyrillStucker

I do not know about iperf but I am using VoIP services from The Real PBX. I think your issues of jitter, throughput, delay and packet loss can be easily solved with them. You may check with them.

ccnxjr

With all the potential ways to spin this I am bounded by a few limitations
-Going with a hosted provider is not an option, we're confident that WAN links are good and happy with our upstream provider, we also don't want hosted providers poking around our LAN. Most won't.
-Buying new hardware or migrating to a sytem is also not an option, we have to work with what was inherited
-The object is monitoring and testing for troubleshooting purposes, not build-out or re-engineering .

Within those constraints I'm going with PJSUA.
It's a command line SIP phone that I can configure to auto-answer, auto-play media and auto-record calls (one call at a time, still working on that).
Effectively I can have multiple calls with REAL rtp traffic traversing the network!
The real seller for this is the reports it generates at the end of a call has all the data I require, the call recording is a big plus!
I can actually dynamically generate these reports during a call and possibly record only the ones with low quality.
It's a beautifully light application, if i can strategically get this on different nodes on my network I can call a box on each subnet periodically. In concert with IPerf i can max out wire speed and see our QoS rules in action by placing a call on top of it using PJSUA

Sample report (I get goosebumps everytime i see it )
[DISCONNCTD] To: <sip:sipaccount@example.com>;tag=
Call time: 00h:02m:38s, 1st res in 1 ms, conn in 163ms
SRTP status: Not active Crypto-suite: (null)
#0 PCMU @8KHz, sendrecv, peer=192.168.1.40:4000
RX pt=0, stat last update: 00h:00m:00.149s ago
total 4.1Kpkt 652.7KB (816.9KB +IP hdr) @avg=32.9Kbps/41.2Kbps
pkt loss=0 (0.0%), discrd=0 (0.0%), dup=0 (0.0%), reord=0 (0.0%)
(msec) min avg max last dev
loss period: 0.000 0.000 0.000 0.000 0.000
jitter : 0.125 2.632 5.125 2.500 0.273
TX pt=0, ptime=20ms, stat last update: 00h:00m:03.268s ago
total 7.6Kpkt 1.23MB (1.53MB +IP hdr) @avg 62.0Kbps/77.5Kbps
pkt loss=144 (1.8%), dup=0 (0.0%), reorder=0 (0.0%)
(msec) min avg max last dev
loss period: 20.000 320.000 2340.000 2340.000 59.655
jitter : 0.000 0.266 0.875 0.125 0.157
RTT msec : 1.281 1.821 4.211 1.541 0.569