From R.Voermans at global-e.nl Mon Aug 4 11:45:32 2008 From: R.Voermans at global-e.nl (Ronald Voermans) Date: Mon, 4 Aug 2008 20:45:32 +0200 Subject: [RTPproxy Users] Problem with lost packets Message-ID: <1EBF014F0B26AF42B43FBC1BB5974F76AE9E6D9A@gi-glogil-ex1.gilze.global-e.nl> Hi RTPProxy users, Our company is still experiencing major upstream-quality issues (IP->PSTN) with our voice-service. We've OpenSer 1.3 in combination with RTPProxy to which all our customers connect. OpenSer forwards these calls to our ITSP, which terminates the calls to the PSTN. The PSTN-side of the call experiences jitter etc. on the call. The IP-side of the call hears the PSTN-side correct. I've placed a monitor (whireshark) on the exit point of our network. From then on it's our ITSP (direct leased-line). Next to that i've listened to several calls from which our customers say it's a bad call. These calls sound ok! So it must be our ITSP (or the leased line) which is causing the problems. But still the ITSP claims there's nothing wrong on their side of the network. They also have a monitor on their PSTN-gateway, and they're seeing a large jitter on our calls. Since it's impossible for me to listen to all the problem-calls I have the following questions: When analyzing the wireshark calls, it also displays the jitter, lost packets and max. delta. Are these values applicable for the complete call (so from customer to OpenSer/RTPProxy to the exitpoint), or only from OpenSer/RTPProxy to the exit-point? I ask this, because it is still possible that there are some packets lost before it enters OpenSer. Would Whireshark notice this? I've asking these kind of questions for several times now. But i'm going mad in analyzing for the cause of these problems. I still think it's our ITSP, but I need some definite proof of this. Simply saying: choose another ITSP isn't that simple. There aren't that many ITSP here around. Thank you for taking the time reading this message, and perhaps answering it!!! Regards, Ronald Voermans From R.Voermans at global-e.nl Sat Aug 9 13:36:18 2008 From: R.Voermans at global-e.nl (Ronald Voermans) Date: Sat, 9 Aug 2008 22:36:18 +0200 Subject: [RTPproxy Users] About Jitter Message-ID: <1EBF014F0B26AF42B43FBC1BB5974F76AE9E6DEA@gi-glogil-ex1.gilze.global-e.nl> Dear listusers, I've been running wireshark on the server which has installed OpenSer and RTPProxy. After doing a trace I find the following: Example: Call goes from 77.94.246.11 -> 10.254.254.1 -> 10.166.38.109 (NAT is in place). OpenSer and RTPProxy are installed on 10.254.254.1. RTPProxy is started via de UDP-socket. This was the only call going on! Analysis shows: 77.94.246.11 -> 10.254.254.1 : Max delta 34.24, Max Jitter 1.69, Mean Jitter 1.05 10.254.254.1 -> 10.166.38.109: Max delta 41.78, Max Jitter 5.50, Mean Jitter 3.83 10.254.254.1 is a HP DL140 (Xeon Processor) with 4 GB memory. CPU is around 0.2%. I've installed the latest stable RTPProxy. Even with one call I get a reasonably extra jitter by RTPProxy. What could be the cause of this? TIA, Ronald. From sobomax at sippysoft.com Sat Aug 9 16:38:01 2008 From: sobomax at sippysoft.com (Maxim Sobolev) Date: Sat, 09 Aug 2008 16:38:01 -0700 Subject: [RTPproxy Users] About Jitter In-Reply-To: <1EBF014F0B26AF42B43FBC1BB5974F76AE9E6DEA@gi-glogil-ex1.gilze.global-e.nl> References: <1EBF014F0B26AF42B43FBC1BB5974F76AE9E6DEA@gi-glogil-ex1.gilze.global-e.nl> Message-ID: <489E2A59.7040308@sippysoft.com> Ronald Voermans wrote: > Dear listusers, > > I've been running wireshark on the server which has installed OpenSer and RTPProxy. After doing a trace I find the following: > > Example: > Call goes from 77.94.246.11 -> 10.254.254.1 -> 10.166.38.109 (NAT is in place). OpenSer and RTPProxy are installed on 10.254.254.1. RTPProxy is started via de UDP-socket. > This was the only call going on! > > Analysis shows: > 77.94.246.11 -> 10.254.254.1 : Max delta 34.24, Max Jitter 1.69, Mean Jitter 1.05 > 10.254.254.1 -> 10.166.38.109: Max delta 41.78, Max Jitter 5.50, Mean Jitter 3.83 > > 10.254.254.1 is a HP DL140 (Xeon Processor) with 4 GB memory. CPU is around 0.2%. > > I've installed the latest stable RTPProxy. Even with one call I get a reasonably extra jitter by RTPProxy. What could be the cause of this? The RTPproxy is by design introduces some extra jitter - around 5ms with default settings. However in practice this doesn't cause any quality issues, since any UA should be able to do jitter compensation for much larger jitter up to tens if not hundred ms. You can decrease additional jitter (at the expense of some extra CPU time) by increasing POLL_LIMIT constant in the rtpp_defines.h and recompiling the software. Increasing it by 5 times should decrease extra jitter down to less than 1 ms. Regards, -- Maksym Sobolyev Sippy Software, Inc. Internet Telephony (VoIP) Experts T/F: +1-646-651-1110 Web: http://www.sippysoft.com From R.Voermans at global-e.nl Mon Aug 11 23:19:03 2008 From: R.Voermans at global-e.nl (Ronald Voermans) Date: Tue, 12 Aug 2008 08:19:03 +0200 Subject: [RTPproxy Users] About Jitter In-Reply-To: <489E2A59.7040308@sippysoft.com> References: <1EBF014F0B26AF42B43FBC1BB5974F76AE9E6DEA@gi-glogil-ex1.gilze.global-e.nl> <489E2A59.7040308@sippysoft.com> Message-ID: <1EBF014F0B26AF42B43FBC1BB5974F76AE9E6DF7@gi-glogil-ex1.gilze.global-e.nl> Maksym, Thank you for your answer. I'll go and try it. One other question about it: The call is exactly as described below. With every call I see the following: every third or fourth RTP packet has a arrival time (Delta) of around 11 ms instead of around 20-23 ms. When I capture the traffic *before* it enters RTPproxy, every RTP packet has a normal arrival time (Delta) of around 20-23ms. Why does this happen? Regards, Ronald Voermans -----Oorspronkelijk bericht----- Van: users-bounces at rtpproxy.org [mailto:users-bounces at rtpproxy.org] Namens Maxim Sobolev Verzonden: zondag 10 augustus 2008 1:38 Aan: RTPproxy Users Onderwerp: Re: [RTPproxy Users] About Jitter Ronald Voermans wrote: > Dear listusers, > > I've been running wireshark on the server which has installed OpenSer and RTPProxy. After doing a trace I find the following: > > Example: > Call goes from 77.94.246.11 -> 10.254.254.1 -> 10.166.38.109 (NAT is in place). OpenSer and RTPProxy are installed on 10.254.254.1. RTPProxy is started via de UDP-socket. > This was the only call going on! > > Analysis shows: > 77.94.246.11 -> 10.254.254.1 : Max delta 34.24, Max Jitter 1.69, Mean Jitter 1.05 > 10.254.254.1 -> 10.166.38.109: Max delta 41.78, Max Jitter 5.50, Mean Jitter 3.83 > > 10.254.254.1 is a HP DL140 (Xeon Processor) with 4 GB memory. CPU is around 0.2%. > > I've installed the latest stable RTPProxy. Even with one call I get a reasonably extra jitter by RTPProxy. What could be the cause of this? The RTPproxy is by design introduces some extra jitter - around 5ms with default settings. However in practice this doesn't cause any quality issues, since any UA should be able to do jitter compensation for much larger jitter up to tens if not hundred ms. You can decrease additional jitter (at the expense of some extra CPU time) by increasing POLL_LIMIT constant in the rtpp_defines.h and recompiling the software. Increasing it by 5 times should decrease extra jitter down to less than 1 ms. Regards, -- Maksym Sobolyev Sippy Software, Inc. Internet Telephony (VoIP) Experts T/F: +1-646-651-1110 Web: http://www.sippysoft.com _______________________________________________ Users mailing list Users at rtpproxy.org http://lists.rtpproxy.org/mailman/listinfo/users From sobomax at sippysoft.com Tue Aug 12 12:56:41 2008 From: sobomax at sippysoft.com (Maxim Sobolev) Date: Tue, 12 Aug 2008 12:56:41 -0700 Subject: [RTPproxy Users] About Jitter In-Reply-To: <1EBF014F0B26AF42B43FBC1BB5974F76AE9E6DF7@gi-glogil-ex1.gilze.global-e.nl> References: <1EBF014F0B26AF42B43FBC1BB5974F76AE9E6DEA@gi-glogil-ex1.gilze.global-e.nl> <489E2A59.7040308@sippysoft.com> <1EBF014F0B26AF42B43FBC1BB5974F76AE9E6DF7@gi-glogil-ex1.gilze.global-e.nl> Message-ID: <48A1EAF9.1090203@sippysoft.com> Ronald Voermans wrote: > Maksym, > > Thank you for your answer. I'll go and try it. One other question about it: > > The call is exactly as described below. With every call I see the following: > > every third or fourth RTP packet has a arrival time (Delta) of around 11 ms instead of around 20-23 ms. > > When I capture the traffic *before* it enters RTPproxy, every RTP packet has a normal arrival time (Delta) of around 20-23ms. > > Why does this happen? The RTPproxy processes incoming packets in "batches", with roughly 10ms interval between batches (by default). Let's say first packet arrives at 9th ms, while the second at 21st ms. As a result, the first will get to the first "batch" and will be sent out almost immediately at 10th ms with the 1ms of delay, while the second will only get to the third "batch", and therefore will go out around 30th ms with 9ms of additional delay. Interval between first and the second would be 12ms in and 20ms out. This is the mechanism that causes jitter to be slightly higher after rtpproxy processing. As I have said, you can decrease this effect by increasing the value of POLL_LIMIT. Regards, -- Maksym Sobolyev Sippy Software, Inc. Internet Telephony (VoIP) Experts T/F: +1-646-651-1110 Web: http://www.sippysoft.com From R.Voermans at global-e.nl Wed Aug 13 08:03:50 2008 From: R.Voermans at global-e.nl (Ronald Voermans) Date: Wed, 13 Aug 2008 17:03:50 +0200 Subject: [RTPproxy Users] About Jitter In-Reply-To: <48A1EAF9.1090203@sippysoft.com> References: <1EBF014F0B26AF42B43FBC1BB5974F76AE9E6DEA@gi-glogil-ex1.gilze.global-e.nl> <489E2A59.7040308@sippysoft.com> <1EBF014F0B26AF42B43FBC1BB5974F76AE9E6DF7@gi-glogil-ex1.gilze.global-e.nl> <48A1EAF9.1090203@sippysoft.com> Message-ID: <1EBF014F0B26AF42B43FBC1BB5974F76AE9E6E1A@gi-glogil-ex1.gilze.global-e.nl> Maksym, That seemed to do the trick. I increased the POLL_LIMIT to 500, and all the jitter I experienced is over. Looks like it still was a problem of our ITSP (who uses a Cirpack platform), because their jitterbuffer was set at 120ms, which should be more than enough for the jitter we were experiencing. But changing the POLL_LIMIT made the problems go away!!! What could be the maximum setting for the POLL_LIMIT? Thanks!!! Regards, Ronald -----Oorspronkelijk bericht----- Van: users-bounces at rtpproxy.org [mailto:users-bounces at rtpproxy.org] Namens Maxim Sobolev Verzonden: dinsdag 12 augustus 2008 21:57 Aan: RTPproxy Users Onderwerp: Re: [RTPproxy Users] About Jitter Ronald Voermans wrote: > Maksym, > > Thank you for your answer. I'll go and try it. One other question about it: > > The call is exactly as described below. With every call I see the following: > > every third or fourth RTP packet has a arrival time (Delta) of around 11 ms instead of around 20-23 ms. > > When I capture the traffic *before* it enters RTPproxy, every RTP packet has a normal arrival time (Delta) of around 20-23ms. > > Why does this happen? The RTPproxy processes incoming packets in "batches", with roughly 10ms interval between batches (by default). Let's say first packet arrives at 9th ms, while the second at 21st ms. As a result, the first will get to the first "batch" and will be sent out almost immediately at 10th ms with the 1ms of delay, while the second will only get to the third "batch", and therefore will go out around 30th ms with 9ms of additional delay. Interval between first and the second would be 12ms in and 20ms out. This is the mechanism that causes jitter to be slightly higher after rtpproxy processing. As I have said, you can decrease this effect by increasing the value of POLL_LIMIT. Regards, -- Maksym Sobolyev Sippy Software, Inc. Internet Telephony (VoIP) Experts T/F: +1-646-651-1110 Web: http://www.sippysoft.com _______________________________________________ Users mailing list Users at rtpproxy.org http://lists.rtpproxy.org/mailman/listinfo/users From sobomax at sippysoft.com Wed Aug 13 11:43:13 2008 From: sobomax at sippysoft.com (Maxim Sobolev) Date: Wed, 13 Aug 2008 11:43:13 -0700 Subject: [RTPproxy Users] About Jitter In-Reply-To: <1EBF014F0B26AF42B43FBC1BB5974F76AE9E6E1A@gi-glogil-ex1.gilze.global-e.nl> References: <1EBF014F0B26AF42B43FBC1BB5974F76AE9E6DEA@gi-glogil-ex1.gilze.global-e.nl> <489E2A59.7040308@sippysoft.com> <1EBF014F0B26AF42B43FBC1BB5974F76AE9E6DF7@gi-glogil-ex1.gilze.global-e.nl> <48A1EAF9.1090203@sippysoft.com> <1EBF014F0B26AF42B43FBC1BB5974F76AE9E6E1A@gi-glogil-ex1.gilze.global-e.nl> Message-ID: <48A32B41.1050505@sippysoft.com> Ronald Voermans wrote: > Maksym, > > That seemed to do the trick. I increased the POLL_LIMIT to 500, and all the jitter I experienced is over. Looks like it still was a problem of our ITSP (who uses a Cirpack platform), because their jitterbuffer was set at 120ms, which should be more than enough for the jitter we were experiencing. But changing the POLL_LIMIT made the problems go away!!! > > What could be the maximum setting for the POLL_LIMIT? Glad to hear that. The reasonable values are within range 100-1000. Above 1000 it should make no measurable difference on additional jitter introduced by the proxy. I would consider increasing default value 2-3 times in the next release, thank you for your input! Regards, -- Maksym Sobolyev Sippy Software, Inc. Internet Telephony (VoIP) Experts T/F: +1-646-651-1110 Web: http://www.sippysoft.com From R.Voermans at global-e.nl Wed Aug 13 13:25:14 2008 From: R.Voermans at global-e.nl (Ronald Voermans) Date: Wed, 13 Aug 2008 22:25:14 +0200 Subject: [RTPproxy Users] About Jitter In-Reply-To: <48A32B41.1050505@sippysoft.com> References: <1EBF014F0B26AF42B43FBC1BB5974F76AE9E6DEA@gi-glogil-ex1.gilze.global-e.nl> <489E2A59.7040308@sippysoft.com> <1EBF014F0B26AF42B43FBC1BB5974F76AE9E6DF7@gi-glogil-ex1.gilze.global-e.nl> <48A1EAF9.1090203@sippysoft.com> <1EBF014F0B26AF42B43FBC1BB5974F76AE9E6E1A@gi-glogil-ex1.gilze.global-e.nl> <48A32B41.1050505@sippysoft.com> Message-ID: <1EBF014F0B26AF42B43FBC1BB5974F76AE9E6E1C@gi-glogil-ex1.gilze.global-e.nl> OK. Glad to help! When we're discussing the variables in the rtpp_defines.h file, what does #define LBR_THRS mean? The other variables I understand... Regards, Ronald -----Oorspronkelijk bericht----- Van: users-bounces at rtpproxy.org [mailto:users-bounces at rtpproxy.org] Namens Maxim Sobolev Verzonden: woensdag 13 augustus 2008 20:43 Aan: RTPproxy Users Onderwerp: Re: [RTPproxy Users] About Jitter Ronald Voermans wrote: > Maksym, > > That seemed to do the trick. I increased the POLL_LIMIT to 500, and all the jitter I experienced is over. Looks like it still was a problem of our ITSP (who uses a Cirpack platform), because their jitterbuffer was set at 120ms, which should be more than enough for the jitter we were experiencing. But changing the POLL_LIMIT made the problems go away!!! > > What could be the maximum setting for the POLL_LIMIT? Glad to hear that. The reasonable values are within range 100-1000. Above 1000 it should make no measurable difference on additional jitter introduced by the proxy. I would consider increasing default value 2-3 times in the next release, thank you for your input! Regards, -- Maksym Sobolyev Sippy Software, Inc. Internet Telephony (VoIP) Experts T/F: +1-646-651-1110 Web: http://www.sippysoft.com _______________________________________________ Users mailing list Users at rtpproxy.org http://lists.rtpproxy.org/mailman/listinfo/users From sobomax at sippysoft.com Wed Aug 13 15:41:47 2008 From: sobomax at sippysoft.com (Maxim Sobolev) Date: Wed, 13 Aug 2008 15:41:47 -0700 Subject: [RTPproxy Users] About Jitter In-Reply-To: <1EBF014F0B26AF42B43FBC1BB5974F76AE9E6E1C@gi-glogil-ex1.gilze.global-e.nl> References: <1EBF014F0B26AF42B43FBC1BB5974F76AE9E6DEA@gi-glogil-ex1.gilze.global-e.nl> <489E2A59.7040308@sippysoft.com> <1EBF014F0B26AF42B43FBC1BB5974F76AE9E6DF7@gi-glogil-ex1.gilze.global-e.nl> <48A1EAF9.1090203@sippysoft.com> <1EBF014F0B26AF42B43FBC1BB5974F76AE9E6E1A@gi-glogil-ex1.gilze.global-e.nl> <48A32B41.1050505@sippysoft.com> <1EBF014F0B26AF42B43FBC1BB5974F76AE9E6E1C@gi-glogil-ex1.gilze.global-e.nl> Message-ID: <48A3632B.1070406@sippysoft.com> Ronald Voermans wrote: > OK. Glad to help! > > When we're discussing the variables in the rtpp_defines.h file, what does > > #define LBR_THRS > > mean? > > The other variables I understand... Ah, it's very old hack that allows rtpproxy to send each packet twice. Could be helpful in the case of very lossy links. Basically if this feature enable the proxy will duplicate every packet with the size less than LBR_THRS. Regards, -- Maksym Sobolyev Sippy Software, Inc. Internet Telephony (VoIP) Experts T/F: +1-646-651-1110 Web: http://www.sippysoft.com