TMCnet - World's Largest Communications and Technology Community




Delivering Reliable Quality of Service

By: Richard "Zippy" Grigonis

Quality of Service (QoS), which refers to simple network quality parameters such as packet loss, delay and jitter, long ago gave way to Quality of Experience (QoE) which takes into account echo, comfort noise, volume, and Mean Opinion Scores (MOS) in a total end-to-end scenario. Although things have improved over the years with regards to VoIP and video, triple and quad-play service bundles, multivendor environments and hybrid networks still present challenges to service providers and their customers.

Cisco (News - Alert) was one of the first major vendors to tackle QoS / QoE. Cisco’s IOS Multiprotocol Label Switching (MPLS) and its underlying concept of assigning high delivery “priorities” to real-time voice and video packet streams serves as the basis for much of QoS and QoE in networks ranging from IP, Frame Relay, ATM, to Ethernet. Subscribers with differing access links can easily be aggregated on an MPLS network edge since MPLS is independent of access technologies.

Today Cisco offers AutoQoS, which incorporates intelligence involving queuing, dropping, traffic conditioning, queue-depth, drop thresholds, burst parameters, etc., into Cisco IOS Software and Cisco Catalyst Operating Service Software to provision and manage large-scale QoS deployments. It automates consistent deployment of QoS features across Cisco routers and switches for Cisco QoS features and variables with the correct parameters. Users can subsequently tune parameters generated by Cisco AutoQoS to suit their particular application needs, as desired.

With Cisco AutoQoS, service providers can use its ‘template driven’ approach to reduce the operational expense of provisioning managed services and QoS for voice traffic to large numbers of customer premise devices. Enterprises can also benefit from deployment costs and time up to three times cheaper and faster than a manual approach.

Some forms of quality control are byproducts of new technologies appearing in certain layers in the OSI stack and specific areas in the network. Matisse Networks (News - Alert), for example, pioneered the field of Optical Burst Switching (OBS) with its EtherBurst packet optical transport system that delivers both the efficiencies of Ethernet packet switching and the huge bandwidth of Dense Wave Division Multiplexed DWDM optical technology. Optical burst switching uses optical burst transponders that can communicate directly with all destinations across a metro network. This means no optical circuits need to be pre-provisioned, and expensive circuit transponders don’t need to be dedicated for every single communication path. This frees up expensive capital, simplifies network design, and enables the creation of pure packet metro networks where bandwidth shifts in real-time to where it is needed in the network. Relative utilization is further increased by finer-grain QoS and more efficient handling of multicast traffic.

As for companies that specifically target QoS/QoE capabilities, they obviously try to encompass as many networks and services as possible. Brix Networks (News - Alert), for example, offers a wide range of converged service assurance solutions that address the specific VoIP, IPTV, and VPN management needs of carriers, cable MSOs, managed service providers, large enterprises, and mobile operators. Their “Brix System” includes BrixWorx (service assurance correlation and analysis software engine), various monitoring sources (Brix Verifiers, Brix software agents, and other performance verification options), BrixCall (voice quality and performance management software), BrixVision (comprehensive IP video quality and performance management software), BrixNGN (scalable software that verifies the quality of next-generation networks), and BrixView (advanced analytics and business intelligence software).

In general the Brix System features what’s called Brix Tri-Q Analysis, that provides complete VoIP and IP video-based service testing across the three vital areas that impact the success of a real-time service: signaling quality (signaling path performance), delivery quality (media transport performance), and content quality (overall quality experience).

Charles Baker, Director of Product Management at Brix, says, “We definitely see many operators offering bundled services. T-Mobile (, for example, offers their WiFi (News - Alert)-based HotSpot @ Home service, which enables T-Mobile cell phone subscribers to seamlessly transfer calls seamlessly between the cellular network and a WiFi hot spot in the home. It’s better than a femtocell, because you don’t need a specific device out there. You can just use a Cisco Linksys (News - Alert) router. They’ve even added an ATA [Analog Telephone Adapter] to that endpoint, so you can use their Talk Forever home phone service by plugging any standard home phone into the HotSpot @Home wireless router, which connects over your broadband Internet connection. So you can replace your home phone service too. Internet does the backhaul.”

“This trend is absolutely happening,” says Baker. “The endpoints are becoming more intelligent; service providers are extending their reach with these intelligent endpoints. They’re first making sure that their core network is performing well from a QoE perspective. Wireless carriers, wireline carriers and everybody is measuring quality across their MPLS cores. They’re looking at their service bundle specifically as to how their data, VoIP, IPTV (News - Alert) and wireless traffic interacts with each other in the MPLS core. That’s really the key to their services. They they’re extending outward as far to the network edge as possible. In the case of T-Mobile (News - Alert), Linksys ATAs today provide functionality called SIP media loop back — you can actively test from the center of the network out to an endpoint, and proactively determine that connection’s quality.”

“So, basically, operators are taking different approaches” says Baker. “They’re actively testing the network, they’re running transactions across it all the time. They’re baselining the network and new customer services, and they’re also passively monitoring the network to see what’s going on. They’re calling up statistics from endpoints, they’re monitoring at different points in the network, they’re doing classic signaling monitoring, and they’re putting all of this information together in different views so they can say, ‘My network is performing at this level, my VoIP service at this level, my data service at this level,’ and so forth. They’re using key performance indicators to do this, such as latency, loss, jitter, MOS scores for VoIP, video quality measurements and so forth. They’re looking at what’s happening with their service, and then they evaluate what’s going on in the network that’s affecting that service, and how the services interact with each other.”

“Resident generating units are the real key these days,” says Baker. “In the case of T-Mobile, they took what was a single revenue-generating unit and they turned it into two by modifying an off-the-shelf device.”

A Chip Off the Old QoS

Centillium (News - Alert) Communications is a major vendor of high performance, cost-effective semiconductor solutions for broadband access. Centillium’s complete, end-to-end System-on-Chip (SoC) solutions accelerate development time-to-market for “last mile” products with Fiber-To-The-Premises (FTTP) and VoIP technologies. Centillium products include digital and mixed-signal integrated circuits and related software for FTTP central office and customer premises equipment and VoIP solutions for carrier- and enterprise-class gateways and consumer telephony.

Centillium’s Dr. Majid Foodeei, VoIP Director of Technical and Strategic Marketing, says, “If you’re referring to VoIP service, and you look at IP data networks, you can see that providing QoS has a long history there and tons of work has been done concerning that — DiffServ [Differentiated Services] and MPLS is the way the world at the protocol level brought QoS in and ‘filled the gap’ that had been occupied by ATM [Asynchronous Tranfer Mode]. ATM and TDM in general were originally a lot more reliable by nature, but the IP world has long since closed the QoS gap. However, providing quality real-time services over IP and mobile networks has proven to be far more challenging in reality than in theory. Certainly I’m not happy with many mobile phone calls and these days I’m not happy because I hear things that I recognize as artifacts that have been attributed to VoIP: delay, poor quality comfort noise, echo and that sort of thing.”

“Centillium has a VoIP play that runs from the CPE side to the network core, from access to wireless to fixed-line,” says Foodeei. “We could potentially be an important piece of the puzzle. I emphasize the ‘puzzle’ aspect of all this because there are many pieces to today’s telecom world, and that’s also part of the challenge. It’s hard to find problems in the network when end-to-end service quality is quite a challenge because there are many multi-vendor components in the network and you could blame everybody and yet nobody wants to admit it’s their fault.”

“What we’ve done is a lot of traffic management work on the CPE side of the equation,” says Foodeei. “which is where Centillium’s Entropia™ devices are used as the core technology for VoIP gateways. At the core we’ve enabled QoS elements such as type-of-service, DiffServ and MPLS. Then we focused on making many measurements of quality, such as latency and the VoIP management protocol RTCP-XR [Real-time Transport Protocol Control Protocol Extended Reports] and RTCP-HR [Real-time Transport Control Protocol High-Resolution Report], but customers are not yet widely using everything that we can make available to them. That’s a problem. On the carrier side, we’ve done such things as reduce packet delay in the AT&T Cingular (News - Alert) network by half.”

Centillium’s Bill MacDonald, Director of Product Marketing, says “At the edge of the network, you must address QoS at all of the different layers when it comes to not just VoIP but all of the different types of services supported within a residential customer’s home, SOHO or even an SMB. You have to start at the physical layer. There are a number of things that the industry is doing to improve QoS at the physical layer. Much of it has to do with the transport technology used. If it’s optical, you typically have a low bit error rate and low delay, and so QoS is not as much an issue at the transport or physical layer. But with technologies such as DSL, which are widely deployed worldwide, you’re dealing with a ‘pipe’ that supports limited traffic and so you have to be careful to prioritize traffic correctly, putting in place mechanisms that give priority to VoIP. Even when transmitting a large data package, you might want to break that packet in the middle of its transmission and transmit a VoIP packet so you don’t incur too much delay or jitter [delay variation].”

“For DSL you can implement a ‘dual latency’ type of architecture whereby you basically enable virtual channels within the DSL pipe,” says MacDonald. “For example, services such as VoIP which aren’t tolerant of delay you would place in a channel having a low latency, which we call the ‘fast path’. You always give those packets high priority. In fact, with some implementations, if you’re transmitting a long data packet, you can break that transmission and insert small voice packets, then continue the data packet transmission.”

“You might have another ‘channel’ we call the ‘interleave path’ where you put services which aren’t delay tolerant at all,” says MacDonald, “such as video services, and which are very sensitive to packet loss. Impulse noise and crosstalk are issues with DSL. For video and some data services, your interleave path needs to have a high level of impulse noise protection. You can achieve that via such things as interleaving to achieve a data path providing an almost error-free transmission path. However, in providing that path, packet delay is incurred, to which video services tend not to be sensitive.”

“Another mechanism now being standardized by the ITU is called SOS (News - Alert),” says MacDonald. “That tries to eliminate line drops for video cell services by detecting increases in crosstalk and rapidly adjusting the data rate to compensate, so the line isn’t dropped if somebody else turns on their VDSL modem, and it cross-talks into your line.”

“Other mechanisms have been proposed, such as a retransmission mechanism at the physical layer,” says MacDonald. “That attempts to reduce the data loss associated with impulse noise on DSL lines.”

“So, lots of things can be done at the physical layer to improve QoS for broadband access,” says MacDonald.

MacDonald elaborates: “As you work your way up the protocol stack, there are various QoS mechanisms you can put in place at the different layers. For example, in our devices we have hardware-based mechanisms in our gateway / router devices which can classify services at layer 2, 3 or 4 using different criteria. The services are classified into different hardware-based queues, and then we provide a scheduling mechanism to ensure that the high-priority services get the highest priority and the low priority services don’t get completely starved. So, there’s a lot that can be done in gateway-chip type of devices that can be implemented in hardware so as to guarantee QoS for the different service types.”

“Then there are ways that you can architect your processor in the gateway,” says MacDonald. “With our System-on-a-Chip [SoC] architecture, we have one or two RISC processors which handle routing, management-type functions and customer applications, but we offload the voice processing – the codecs, the echo cancellation – to a separate DSP [Digital Signal Processor (News - Alert)]. That’s important because you if have a high call volume at a particular time, your DSP can handle those voice calls without impacting the routing being done on the RISC.”


Bridging the Service Quality Gap

By: Yaron Eisenstein

Guaranteeing reliable QoS seems to be an elusive goal for many telecom service providers, particularly given that defining “reliable QoS” in meaningful, concrete terms can be a challenge. We have seen from enough cases that even when a service provider is meeting its own internal KPI benchmarks, customers may still complain of poor service quality. Reliable QoS is QoS which meets subscriber expectations. Therefore, the issue is much larger than simply a healthy network; the service delivery process must be managed in relation to subscriber expectations, their perception of the end-user experience, and how well the gap between these points is minimized.

New service management technology has enabled an improved set of best practices that enable service providers to define and deliver reliable QoS in a way that is more meaningful in terms of customer retention and other business-level objectives.

In order to reliable QoS from the subscribers’ point of view, service providers first need to gain a clear understanding of their priorities and expectations. To create a solid knowledge base for policymaking, the necessary data can be gathered to a large extent from within the network, such as current performance and quality levels, usage trends, and any other information which may be leveraged provide insight on subscriber needs and demands. External sources must also be included, such as feedback from customer care centers or other business intelligence. For example, if subscriber expectations change to emphasize disconnection rates, responsiveness to customer feedback, or any other aspect of the user experience, then the operator must update their definition of “reliable QoS” accordingly and adjust their quality monitoring standards.

Having outlined their business-level objectives, service providers can then move on to define meaningful metrics from the customer point of view. In order to take concrete steps towards meeting customers’ expectations, operators can take advantage of new OSS tools to create a wide variety of KPIs and KQIs to measure service performance and quality. Since customer expectations — and therefore the Service Level Objectives (SLOs) — can rapidly shift with a new service offering or other market development, a service provider needs tools which can display these custom indicators according to each key customer or customer group, rather than by just by service or network domain. For example, if a provider’s subscriber base includes VNOs, enterprises, and private users, each will have different expectations, and so the operator must monitor its performance for each. Flexibility is also key, as each new service launched may require new custom indicators.

With the tight competition in today’s marketplace, it is not simply enough for an operator to meet its SLOs, but it is necessary to do so as efficiently as possible. First of all, operators need end-to-end visibility of traffic and services, with the ability to present service performance in a customer-centric display. A unified mediation layer is an important basis of any OSS platform, consolidating vast amount of data feeds and enabling smooth integration with new data resources as they become available. With a certain level of integration between Fault and Performance Management systems, this information can then be leveraged by more advanced algorithms, such as service-impact analyses, displaying network and service events in terms of their impact on each user’s experience. Robust root-cause and drill-down OSS functionalities can analyze these data sources to monitor individual customers or subscriber groups as part of manual or automated processes in the fault resolution cycle, helping to align all aspects of network operation with the organization’s SLOs as well as the larger business-level objectives.

Reliable QoS doesn’t end with SLO assurance. Service providers must continually manage perceptions, working to improve customer awareness of their capabilities and commitments. With key business customers, this can be done through tailoring SLAs to realistically define what their expectations should be. For the general consumer base, a provider can define reasonable SLOs for both technical indicators such as call success rate and service availability, as well as broader aspects of service delivery such as customer support responsiveness.

Service providers must address the issue from both angles: it is equally as important to meet customer expectations as it is to let customers know what to expect. It can be easy to place more of an emphasis on the delivery aspect, and rightly so, as this is the area which makes the most concrete impact on the end user experience. Nevertheless, in designing their SQM solutions, operators must keep all aspects of operations aligned with their larger business-level goals in order to achieve meaningful success for the organization as a whole. IT

Yaron Eisenstein is the Director of Marketing of TTI Telecom (News - Alert). For more information, visit the company online at


Quality Ethernet

Extreme Networks (News - Alert) designs, builds, and installs Ethernet infrastructure solutions for enterprises and service providers demanding high performance, converged networks that support voice, video and data, over a wired and wireless infrastructure. Extreme Networks provides automation functionality with its Universal Port software, integrated in their ExtremeXOS operating system.

Peter Lunk, Director of Service Provider Marketing, says, “If you look at both telephony and triple play services, and if you think about residential networks, we fit in there at an aggregation layer, where we might be bringing in a number of DSLAMs or similar equipment and we aggregate the traffic. We don’t typically make the ‘first mile’ gear, that’s usually done through partners. We find that the place that we must perform the QoS function is at the first aggregation point. ATM-based DSLAMs have more or less gone away, and there’s been a big move to Ethernet, so there’s been an increased need for QoS methodologies on Ethernet aggregation gear.”

“With triple play services, in the past there was a model where you would assign a VLAN [Virtual LAN] for each service type running back to some sort of service router in the POP [Point of Presence] or the cable head-end,” says Lunk. “But now, we see service providers adopting a different model where they actually offer a VLAN for each subscriber, and they would like to provide distinct QoS for each subscriber application, whether that be voice, video or just data connectivity. It becomes a tiered QoS where, on a port on an Extreme Networks aggregation switch, you want to be able to limit how much bandwidth is coming in on that port. We might have 300 to 500 subscribers coming through if there’s a DSLAM on that port, and we might want to limit each one of those subscribers to some fixed bandwidth. And, with each one of those subscribers we also might want to classify their voice traffic as having the highest priority, even above video, then make sure video is prioritized above the data bandwidth, but then allowing the data to peak if the subscriber isn’t watching video. You also want the flexibility to be able to offer different packages.”

“If you have 300 to 500 subscribers on each Ethernet port, and then you have a switch that has 20 ports, you’re talking about tens of thousands of subscribers on a single module in the worst case,” says Lunk. “So, one trend we see is a move toward silicon-based QoS processing, specifically around the residential model where you have a tiered, hierarchical QoS. For the past ten years we’ve had Ethernet switches equipped with a hardware-based QoS capabilities with, say, eight queues going onto a backbone, but now we’re dealing with something different – controlling things down to the application level and going out to the subscriber-facing ports. We were one of the first to do that in silicon.”


Multilayered QoS for Reliable VoIP Service

By: Oded Tubias

The advancement of broadband technologies has opened the frontier for wide-scale service provider deployment of VoIP services. Service providers that take advantage of this opportunity cannot compete on price alone. For a VoIP service to be a success, it must rival PSTN reliability and voice quality. This can be achieved by deploying a VoIP system with Quality of Service (QoS) mechanisms that ensure the delivery of the toll quality voice that customers expect from traditional fixed line telephony.

Despite the standardization of IP telephony protocols, not all VoIP systems are created equal. A system with a multilayered QoS architecture best provides the safeguards to neutralize the factors that can degrade voice quality in transporting IP phone calls. The failure of a VoIP service to deliver quality communications can quickly eat into profits and run it out of business. Effective QoS thereby enables VoIP service providers to keep Opex down while maintaining customer satisfaction.

The Resource Layer

While the beauty of VoIP is cost-savings from the convergence of voice and data networks, the flip side is the co-sharing of the data communications bandwidth resources at the customer’s disposal. To overcome the resource competition between voice and data that can potentially harm voice quality, a VoIP service provider must enable customers to control resource distribution, giving processing preference to real time voice packet traffic. The base of a multilayered QoS design for ensuring voice prioritization is therefore the resource layer.

To ensure that the quality of phone conversations is never compromised, the CPE should carry out traffic shaping, giving priority to the real time traffic. In this method, bandwidth use is controlled dynamically, reducing the amount of bandwidth serving data applications and making more available for voice. This QoS provision controls bandwidth according to demand, with voice always getting preferential use of customer resources. The passage of data should be halted completely by the CPE if necessary to ensure maximum bandwidth for voice traffic.

The best solution to the QoS resource challenge is to offer customers a “private virtual connection” (PVC). In this architecture, separate channels are created for voice and data transmission from the CPE to ensure there is no competition for resources, with each application having its own reserved path. In this way QoS is guaranteed, since there are dedicated channels for separate voice and data traffic, removing the risk of congestion created by the flow of two traffic types.

The Priority Layer

In the best-effort world of IP, standards for conduct are being widely adapted throughout networks, resulting in the rapid transformation of VoIP into a service with end-to-end QoS. This is achieved by using tagging, the priority layer of multilayered QoS architecture. The labeling of the header of a packet according to Type of Service (ToS) gives instructions to the IP network on how to handle the packets and how to prioritize the traffic according to the type of packet’s content. According to this regime, real time voice packets should be given the right of way by all networks.

VoIP QoS can also be guaranteed at the priority layer by offering customers a system with VLAN, which is another type of tagging. If the provider supports VLAN along the network, it can give priority to a specific VLAN tag (News - Alert) according to a pre-defined prioritization mechanism, handling VLAN-tagged traffic with its highest QoS capabilities. The priority layer is, therefore, the heart of VoIP QoS, transporting instructions to the network to ensure the uninterrupted flow of voice traffic in the network.

The Management Layer

The top layer for delivering QoS to customers is a centralized VoIP network management system. A central integrated management and provisioning system supplies the voice quality monitoring that is crucial for providing a quality service. The management layer needs analysis tools for end-to-end, real-time service monitoring. The provisioning system should receive call quality data on the levels of delay, jitter and packet loss, delivering alerts when problems arise.

VoIP QoS does not only depend on the service provider’s equipment and technology. Many service disruptions arise from problems with Last Mile (News - Alert) access technology, usually DSL lines provided by a third party. A service provider has to be able to monitor the quality of the access line and send the information to the central provisioning system for analysis. This capability has the added benefit of enabling service providers to set up QoS before the VoIP telephony service is operated, thereby ensuring customer Quality of Experience right off the bat.

Quality of Service brings Quality of Experience

A VoIP system with the right composition of QoS tools can ensures service provider delivery of a Quality of Experience that rivals the toll quality of PSTN service. VoIP service providers that provide customers with a smooth transition to VoIP with consistent voice quality can succeed in this competitive market. IT

Oded Tubias is VoIP Product Line Manager at RAD Data Communications (News - Alert), where he is responsible for product marketing for the RAD VoIP System. He has more than 15 years of experience in product definition, design and management. Reach him at [email protected].


QoE Drills Down

Psytechnics (News - Alert) is a leading provider of software solutions for assessing and managing QoE for real-time voice and video. Service providers and enterprises worldwide use their products to monitor, manage and improve IP telephony, video and unified communications solutions in both fixed and mobile environments. Psytechnics’ Experience Manager product augments traditional data network management applications by delivering real-time alarms, drill-down analysis and advanced diagnostics involving voice and video. Experience Manager can identify call-ruining issues that many other solutions are unable to detect, such as echo and noise. Besides analytics and diagnostics, Experience Manager aggregates Quality metrics for SLA reporting and trending.

Benjamin Ellis, Psytechnics’ Vice President of Marketing, says “We see that VoIP is really bringing home the practical issues and applications of QoS. People have been talking about QoS for a long time but now it’s becoming more of a visible service issue since people are making heavier and heavier user of VoIP technologies. A few years ago, people took a network view of QoS, measuring packet loss, delay and jitter at one point in the network. But as soon as you introduce VoIP or videoconferencing into the network, existing definitions become problematic for a number of reasons. The first is that you’re looking at links, and when delivering voice or video, people care less about the link and more about what happens with the actual user. When you look at QoS for non-real-time applications such as web access, you can get away with a fair amount of variation. But as soon as you start dealing with real-time voice and video, quality becomes really visible to the end user in terms of what the actual QoS is that’s delivered. Now you’re looking at individual user streams or user experiences, as opposed to what’s happening with a whole ‘pipe’ of packets.”

“If I have 100 voice calls traversing a circuit,” says Ellis, “I need to understand what’s happening with every single one of them, as opposed to what’s happening with the whole gigabit Ethernet link.”

“The term QoS itself is evolving,” says Ellis. “Many people use the term QoE for Quality of Experience. We’ve used it for a long time, just to differentiate what we call a network view, which is operationally useful. But when you talk about real-time voice and video, it’s more about QoE, which includes what’s happening up at the application layer. QoE is more than just packet loss, delay and jitter, it’s about echo, whether the call was loud enough, whether it was distorted in some way, and so forth. These things are completely invisible to the IP world, and many of these other problems come from voice gateways. The challenge is that they don’t show up as packet loss and jitter. You have to look at the streams, and you have to be aware of what’s happening in the network too.”

Making packetized voice and video an enjoyable user experience is a major challenge these days, but it’s also a major opportunity for those innovative companies with the technologies that successfully help service providers as well as their corporate and residential customers. IT

Richard Grigonis (News - Alert) is Executive Editor of TMC’s IP Communications Group.


The following companies were mentioned in this article:

Brix Networks (

Centillium Communications (

Cisco Systems (

Extreme Networks (

Matisse Networks (

Psytechnics (


» Internet Telephony Magazine Table of Contents

Today @ TMC
Upcoming Events
ITEXPO West 2012
October 2- 5, 2012
The Austin Convention Center
Austin, Texas
The World's Premier Managed Services and Cloud Computing Event
Click for Dates and Locations
Mobility Tech Conference & Expo
October 3- 5, 2012
The Austin Convention Center
Austin, Texas
Cloud Communications Summit
October 3- 5, 2012
The Austin Convention Center
Austin, Texas