«End-to-End QoS Provision over Heterogeneous IP and non IP Broadband Wired and Wireless Network Environments A dissertation submitted in satisfaction ...»
1. Static guaranteed: This policy guarantees a static bandwidth to each virtual channel. A guaranteed bit rate value has to be speciﬁed so that the actual bit rate is guaranteed up to this boundary value. The unused bandwidth (guaranteed bit rate -instant bit rate) is reserved and cannot be allocated to other virtual channels. q
2. Dynamic guaranteed: This policy guarantees a dynamic bandwidth to each virtual channel. A guaranteed bit rate value has to be speciﬁed so that the actual bit rate is guaranteed up to this boundary value. On the contrary to the static guaranteed policy, the unused bandwidth (guaranteed bit rate
-instant bit rate) is not lost, but can be allocated to other virtual channels.
3. Best eﬀort: This conventional policy allocates bit rate to various virtual channels based on the available bandwidth.
The DVB domain employs the implemented priority policies in order to preserve traﬃc classes deﬁned in the IP domains. The binding among Bandwidth Aggregates (BA) and the corresponding priority policies is given in Table 5.2.
AF1x BAs and BE BA can borrow bandwidth beyond the guaranteed. Whilst the EF BA is statically allocated a maximum value, hence cannot borrow unused bandwidth.
5.2.4 DiﬀServ/UMTS Class Coupling
In order to integrate the UMTS network domain with the others networks domains, and to achieve QoS consistency across DiﬀServ IP network and UMTS network, by mapping UMTS classes to predeﬁned DiﬀServ classes. Here two approaches can be envisaged: (1) One-to-one mapping which maps each UMTS class to a corresponding DiﬀServ QoS class. However, one-to-one mapping might not always be possible since networks may support diﬀerent sets of QoS classes.
(2) Many-to-one mapping this approach can map a number of DiﬀServ QoS traﬃc classes into a single UMTS QoS traﬃc class. A DiﬀServ core can deﬁne many QoS traﬃc classes, when compared to only limited QoS classes supported by UMTS; then a close set of DiﬀServ QoS traﬃc classes having almost similar
QoS requirements can be merged into a single UMTS QoS class.
The Table 5.3 shows the mapping based on - of the predeﬁned DiﬀServ classes according to DiﬀServ speciﬁcation, where the ﬁrst digit of the AF class indicates forwarding priority and the second indicates the packet drop precedence, and the UMTS QoS classes for both mapping approaches.
Policing of traﬃc levels for each UMTS class might mean that within an agreed bucket level the DiﬀServ class might change from AF1x to AF2x for a partially ﬁlled bucket and to AF3x for anything over the limit. These suggested values treat the dropping of streaming layers as less critical than those for background traﬃc.
The actual QoS that can be obtained depends on detailed traﬃc engineering for both radio and DiﬀServ networks.
The packets, with assigned priority, are sent to the DiﬀServ network to receive diﬀerent forwarding treatments. Mapping these prioritized packets to diﬀerent QoS DS levels causes them to experience diﬀerent packet loss rates with this diﬀerential forwarding mechanism. In addition, to the prioritized dropping performed by DiﬀServ routers, traﬃc policing can be carried out at intermediate video gateways (between diﬀerent network domains), using packet ﬁltering.
5.3 Testbed Conﬁguration This section evaluates the performance of the proposed architectural framework through a set of experimental cases. I study the performance of our framework by enabling or disabling scalable video coding and/or by enabling or disabling prioritized transmission. The quality gains of scalable video coding in comparison with non-ﬁne grain scalable video coding and the quality gains of prioritized transmission in comparison with non-prioritized transmission applying two different DiﬀServ/UMTS traﬃc classes mapping approaches are discussed in detail.
The conﬁguration setup is depicted in Figure 5.3.
Eight YUV Quarter Common Intermediate Format (QCIF) 4:2:0 color video sequences consisting of 300 to 2000 frames and coded at 25 frames per second are used as video sources. Each group of pictures (GOP) is structured as IBBPBBPBB. and contains 25 frames, and the maximum UDP packet size is at 1024 bytes (payload only). The Microsoft MPEG-4 FGS encoder/decoder  is used for encoding YUV sequences. A number of background ﬂows is transmitted in the network, in order to lead the system in congestion.
A unique sequence number, the departure and arrival timestamps, and the type of payload that contains, are obtained identify each packet. When a packet does not reach the destination, it is counted as a lost packet. Furthermore, not only the actual loss is important for the perceived video quality, but also the delay of packets/frames and the variation of the delay, usually referred to as packet/frame jitter. The formal deﬁnition of jitter, which is used in this analysis, is given by the equation 5.3 and 5.4. It is the variance of the inter-packet or inter-frame time. The frame type is determined by the time at which the last
segment of a segmented frame is received. The packet jitter is deﬁned by:
The packet/frame jitter can be addressed by so called play-out buﬀers. These buﬀers have the purpose of absorbing the jitter introduced by the network delivery delays. It is obvious that a big enough play-out buﬀer can compensate any amount of jitter. There are many proposed techniques in order to develop eﬃcient and optimized play-out buﬀer, dealing with this particular trade-oﬀ. These techniques are not within the scope of the described testbed. For our experiments
the play-out buﬀer is set to 1000 msecs.
In order to measure the improvements in video quality by employing H.264/MPEGAVC, I use the Peak Signal to Noise Ratio (PSNR) and the Structural Similarity (SSIM)  metrics. P SN R is one of the most widespread objective metric for quality assessment and is derived from the Mean Square Error (MSE) metric, which is one of the most commonly used objective metrics to assess the application level QoS of video transmissions .
Let’s consider that the video sequence is represented by V (n, x, y) and Vor (n, x, y), where n is the frame index and x and y are the statial coordinates. The average P SN R of the decoded video sequence among frames at indices between n1 and
n2 is given by the following equation:
Note that, the P SN R and M SE are well-deﬁned only for luminance values.
As it mentioned in , the Human Visual System (HVS) is much more sensitive to the sharpness of the luminance component than that of the chrominance component, therefore, it is considered only the luminance P SN R.
SSIM is a Full Reference Objective Metric  for measuring the structural similarity between two image sequences exploiting the general principle that the main function of the human visual system is the extraction of structural information from the viewing ﬁeld. If v1 and v2 are two video signals, then the SSIM
is deﬁned as:
where L is the dynamic range of pixel values and K1 = 0.01 and K2 = 0.03, respectively.  deﬁnes the values of K1 and K2.
5.4 Results This section evaluates the performance of the proposed framework conﬁguration through a set of experimental cases. I study the performance of our framework by enabling or disabling scalable video coding, or by enabling or disabling prioritized transmission. The quality gains of scalable video coding in comparison with non-scalable video coding and the quality gains of prioritized transmission in comparison with non-prioritized transmission are compared in detail.
The implemented DiﬀServ mechanism is incorporated into the two IP autonomous systems of the heterogeneous IP/DVB/UMTS testbed. Each autonomous
system consists of three PCs (at least PIII CPU with 512MBytes of RAM) running Linux Operating System (kernel version 2.6.11)  with iproute2 package and tc utility support. Each IP domain includes two edge routers and one core router. The supported BAs are EF, AF1x and BE. The Hierarchical Token Bucket (HTB) packet scheduler with three leaf classes is used for the realization of the supported BAs. Specically, a pFIFO queuing discipline is adopted for the EF BA. Three Generalized Random Early Detection (GRED) virtual queues with diﬀerent drop precedences are implemented for the AF1x BA. The BE BA is served through a RED queuing discipline. The maximum bandwidth allocated at the parent HTB class is 13Mbps shared among the BAs. Each leaf class can borrow excess bandwidth from another leaf class.
The DVB domain includes two full uplink/downlink conﬁgurations. The uplink involves an encapsulator, a multiplexer and a DVB-S modulator. The downlink is realized through a DVB/IP gateway, which is a standard PC running Linux operating system (OS) equipped with a standard Ethernet controller and a DVBS PCI card capable of demodulating the DVB-S signal and de-encapsulating the IP packets. Note that in order to deal with the IP to MPEG-2 encapsulation overheads, the total link bandwidth is 14 Mbps, which is 1 Mbps bigger than the IP domains one.
For UMTS, NS-2 based simulation environment with the appropriate Enhanced UMTS Radio Access Network Extensions for ns-2 (EURANE) package extensions for simulating a UMTS network is adopted. A single UMTS cell
of 1Mbps with the following rate allocation for the supported traﬃc classes:
200Kbps for the Conversional class, 300Kbps for the Streaming class, 200kbps for the Interactive 1 class, 100kbps for both Interactive 2 and 3 classes, and 200Kbps for the Background class, is simulated. In order to ﬁll in the UMTS class capacity, a number of background ﬂows are transmitted in the network. The background traﬃc is increased from 210Kbps to 540Kbps leading the UMTS network in congestion. Two mapping approaches, presented in Table 5.3 are employed.
The ﬁrst experimental case refers to a single layer MPEG-4 stream transmission and is encoded at 384kbps. For this scenario, I use EF for transmitting I-frames and AF12 and AF13 for transmitting P- and B- frames respectively. The mapping of DiﬀServ classes to the UMTS ones is performed through Table 5.3.
The second experimental case concerns a scalable MPEG-4 stream transmission consisting in two layers. The BL packets are encoded using the MPEG4-FGS codec with MPEG-2 TM5 rate control at 128kbps and the EL ones are encoded at 256kbps. For this case, I have direct application of Tables 5.2 and 5.4.
The third experimental case concerns a scalable MPEG-4 stream transmission consisting in one BL and two ELs, i.e., EL1 and EL2. The encoding of BL packets remains at 128 kbps as in the second case, while the encoding of packets of both ELs is at 128kbps. For this scenario, I use EF for transmitting BL, AF11 for transmitting EL1, and Best Eﬀort (BE) for transmitting EL2. The mapping of DiﬀServ classes to the UMTS ones is performed through Table 5.3.
The fourth experimental case adopts the setup of the third case, while it applies the prioritized packetization scheme of the second case to the packets of
the ﬁrst EL, i.e., for this scenario, I use EF for transmitting BL, Table 5.2 for transmitting EL1, and Best Eﬀort (BE) for transmitting EL2.
Tables 5.5 and 5.
6 depict the results from experiments in terms of PSNR and SSIM video quality metrics for eight diﬀerent YUV video sequences for all cases (1 to 4) for the two mapping settings concerning DiﬀServ/UMTS classes coupling (one and many-to-one mapping). For one-to-one mapping, each conﬁguration case increases the video quality and the gain increment that oﬀers each case is around 2dB in terms of PSNR. For many-to-one approach, the Cases 3 and 4 produce the same results.
For the Highway video sequence, I measure the packet/frame loss for I-, Pand B- frames for the four scenarios fro the two mapping approaches. Fore Case 3 and 4 the depicted measurements concern EL1. The results presented Tables
5.7 and 5.8 are in accordance with the ones depicted in Tables 5.5 and 5.6. For one-to-one mapping, each case improves the previous one and Case 4 oﬀers the
best video quality gain as it experiences the lower packet frame losses. For manyto-one mapping, Case 2 oﬀers the best video quality.
By isolating the losses and the delays to P- and B- frames it is achieved signiﬁcant gains to video quality. Packet losses, which P-frame content, can aﬀect not only the decoding process of P-frames but also the B-frames. This lead to higher percentages of B-frame losses but it is a signiﬁcant aﬀect to the overall video quality. In the fourth scenario, the user can achieve the same video quality, compared to third scenario, without using only the AF11 traﬃc class of the DiﬀServ. Distributing the traﬃc to all traﬃc classes, achieving the same video quality, in the lowest price, by sending lowest traﬃc to the cost eﬀective AF11 traﬃc. From the network provider perspective, the providers network can use more eﬃcient its bandwidth, by serving more users, at the level of quality they pay.
As an overall remark of the above results, I could note that Case 4 for oneto-one mapping could oﬀer almost the same video quality as Case 2 of many-toone mapping approach, without employing the cost eﬀective conversational class.
However, one-to-one mapping might not always be possible since networks may support diﬀerent sets of QoS classes.