Media Gateway Benoit Bégué 2006 Study for EE department. EE526 with Professor Dan Keun Sung
Table of Content • Introduction: what is a Media Gateway ? • Operation principles • Related standard • Performance issues • Problems in performance measurements • Experiment environment • Tests and results • Others: quick market analysis • Conclusion
What is a media Gateway ? • Media Gateway :translation unit between disparate telecommunications networks such as • PSTN • Next Generation Networks • 2G, 2.5G and 3G radio access networks • PBX • Media Gateways enable multimedia communications across Next Generation Networks over multiple transport protocols such as ATM and IP
Operation principles • A couple translating and linking different protocols and technologies: • Gatekeeper or Media Gateway Controller (MGC) • deals with the signaling • Media Gateway (MG) • deals with the data/voice
Related standards : signaling protocols • Signaling protocol between an MGC and gateways : • ITU-T Rec. H.323. This is employed where all network elements (NEs) have software intelligence. • SIP (Session Initiation Protocol, Ref. 2) is used when the end devices have software intelligence and the network itself is without such intelligence. • MGCP (media gateway control protocol) is another gateway control protocol. • MEGACO (ITU-T Rec. H.248, Ref. 13) is a gateway control protocol applicable when end devices are without software intelligence and the network has software intelligence.
Related standards : voice codecs • The terminals may have different capabilities. That’s why we need a negotiation about terminal compatibility. • PCM, ADPCM, LPC, AMR, A-CECP,G.7xx • H.245 is used to negotiate service capabilities between terminals and gateway controllers
Related standards : terminations on a GW • Terminations are logical entities on a GW: • Analog telephone line • Digital telephone (64 Kbps) • T1 trunk line (1.544 Mbps) • E1 (2.048 Mbps) • RTP (VoIP) Example of Cisco T1 card interface (RJ48 port)
Performance analysis • Reference Paper • A Study of IP Network Impairments Impact on Media Gateway Performance • Sherry Wang, Luis Nieto, Edward Zielinski • Next Generation Networks, Bell-Labs, Lucent Technologies • Ask one question : • Can voice quality be guaranteed in a connectionless and best effort data network ?
Performance issues • A GW involves many complex processes: • packetization / depacketization of voice frames • jitter smoothing • delay • error treatment (lost packets) • If Quality of Service (QoS) engineered IP network is available, the GW is a bottleneck
Problems in performance measurements • Voice quality is defined as the qualitative and quantitative measurement of the sound and conversation quality. • Subjective ! • Mean Opinion Score (MOS) • Gives a subjective opinion score • expensive in terms of time and human efforts • Perceptual • Speech Quality Measurement (PSQM) • defined by ITU-T Recommendation P.861 • Repeatable • Objective • Reasonably inexpensive
More about PSQM • PSQM compares an original voice signal to a received copy of this signal • PSQM measures the dissimilarity of the received signal to the original signal from human auditory perceptions. • A lower PSQM score implies higher similarity between two signals (good !). The score of zero corresponds to equality of two signals to the human ear (excellent !). • Big PSQM score -> bad !
Experiment environment Performance measurement tools (acts as 2 terminals) IP Network (can simulate Impairments)
Experiment environment • G.711 vocoder was used through out this study • The choice of the number of voice frames per IP packet is a trade off between a fast delivery and network bandwidth efficiency. In this study, 10ms voice frame per RTP packet was used as recommended by the manufacture. • An adaptive jitter buffer was used to accommodate jitter smoothing and reduce delay at the same time.
Tests • 1st: Analog communication: only switch • 2nd: Analog and IP communication (using switches and gateways) in ideal IP network condition • 3rd: 2 with impaired IP network • Delay • Jitter • Packet Loss
Theory for IP Network Delay Impact (on Voice Quality) • The end-to-end delay or the speech latency can be summarized as a function of • gateway performance g • IP network hop behavior h • Dpkt is packet-processing delay • Th is transmission delay • Qh is queuing delay • Ph is propagation delay
Theory 2 • The hop behavior is replaced by a fixed delay D(i) generated by the IP network emulator. • D(i) is a constant in ith test, i = 1, 2... • N is the number of frames per packet –constant in this study- • f is the frame size C was measured at approximately 26.5ms that is about 6ms in switch and about 10ms in each gateway
Results for IP Network Delay Impact (on Voice Quality) D(i) was selected from 10ms to 70ms with 10ms increments.
Theory for IP Network Delay Variation Impact (on Voice Quality) • Quality is affected by not only the end-to-end delay, but also the delay variation = jitter • gateway jitter: J(di, ji) • where di is a probabilitydistribution • ji is an average jitter value in ms for the ith test. • A uniform distribution and an Internet distribution were used.
Theory 2 • An adaptive jitter buffer is used in a gateway to effectively smooth out jitter without introducing too much delay • Speech latency is affected if • Jitter exceeds buffer capacity • Jitter triggers buffer adaptations frequently
Theory of IP Network Packet Loss Impact • Voice quality, in terms of intelligibility, is directly affected by packet loss. Two types of packet loss can occur: • Single-packet loss: occasional bit error on the transmission facility • Consecutive packet losses: due to traffic congestion • We will see which one is worst
Others: Market analysis • Cisco Systems is the leader for general VoIP network equipment • Huawei is the current market leader for media gateway port shipments for 2005 • With a projected increase of over 300% in the next 5 years and a total world market of over $8.7 Billion, demand for Media Gateways signal the accelerating transition from PSTN to VoIP
Conclusion • The network delay affects only the conversation quality, not the voice quality • This is achieved through an optimal selection of a jitter buffer size and a packet drop rate in a gateway • The burst packet loss is more detrimental to voice quality than the single-packet loss • What is missing in this presentation: • Codec comparison • and their echo cancellation capability • From a network-service provider’s perspective, simple system configurations and consistent parameter settings are desirable features of a gateway and of other network components for a speedy VoIP network deployment.