March 2008 | Volume 11/ Number 3
QoS in Enterprise Networks
By Richard "Zippy" Grigonis
Back in the good old days before VoIP, an IT guy like Yours Truly could rely on service level agreements (SLAs), even if they were totally wrong, since data is more forgiving of network hiccups than real-time services such as voice and video. In those days, a simple handheld reflectometer was all you needed to test your network’s Quality or Service (QoS) – which was whether or not a cable was broken.
Today, we have far more sophisticated tools that must work in contemporary IP and Fixed-Mobile Convergence (FMC) environments. For example, AirMagnet (http://www.airmagnet.com) makes the ingenious AirMagnet VoFi Analyzer that is claimed to be the first of its kind – a professional monitoring and troubleshooting tool for Voice-over-WiFi. With VoFi Analyzer a technician can directly monitor the quality of every call whizzing over an organization’s wireless LAN and automatically detect the sources of voice problems. The device can actively troubleshoot problems from the phone to the wired call manager. VoFi Analyzer has a built-in expert system that considers all the things that can go wrong with wireless voice, enabling it to quickly diagnose device-related problems, QoS problems, RF problems, WLAN problems and even wired network problems associated with Voice-over-WiFi.
As it happens, QoS is ultimately much bigger than packet loss, delay and jitter. It’s about end-to-end quality of experience (QoE) for the user, which usually involves a set of hops through a hybrid network as complicated and bizarre as a picaresque novel. Companies both large and small have risen up to accept the challenge and help enterprises achieve toll-quality voice and video communications over IP.
At Computer Associates, now more commonly known as CA (http://www.ca.com), CA eHealth for Voice combines data collection across the voice network with performance monitoring, alerting, reporting and policy management for Avaya, Nortel and Cisco PBX, IP Telephony, voice messaging or unified messaging platforms. It helps you to pro-actively manage legacy, hybrid or IP telephony systems from one console. With CA eHealth for Voice you can identify degradation of service issues such as maintenance warnings, trunk or port capacity thresholds, and other service-affecting issues on the PBX, IP PBX, voice or unified messaging platforms. Its capacity analyzer tools can determine the grade of service and assure sufficient capacity to handle traffic loads.
CA’s Steve Guthrie, Director of Product Marketing, says, “When CA acquired Concord Communications in June 2005 it actually ended up with several best-of-breed solutions which Concord had acquired relating to voice. These have been integrated into our CA product line.”
“When you say ‘Quality of Service’ people immediately think of packet delay, jitter and loss,” says Guthrie. “They want a calculated value which, more often than not, is some sort of MOS [Mean Opinion Scale] rating. I’m here to articulate today that, QoS is more than MOS. However, I want to acknowledge that we have at least four ways to compute MOS within our product set for managing voice and surely MOS is important. Indeed, we enable it in the Cisco and Nortel worlds, for example, and our eHealth for Voice solution can pulls stats from a Cisco Call Manager or from a Nortel CS1000, and we can compute MOS just like everybody else. We also have a voice quality monitor component for eHealth for Voice that does simulated tests. So we have both proactive as well as static ways of computing MOS. We can also use the IP-SLA capability within Cisco routers to set up tests and we can do that with both our eHealth and Spectrum products.”
“Thirdly, we have a growing partnership with Psytechnics [http://www.psytechnics.com],” says Guthrie. “There are some common Psytechnics and eHealth customers that have done their own integration so they can use the Psytechnics tools. We’re making our partnership more formal, so that customers can have a fourth choice, if your will, for computing MOS.”
“MOS is important, and people like it, and we enable its measure,” says Guthrie. “But we go further. It’s more than QoS. When I think of quality of service, I think about the voice service as a holistic approach. It’s not just MOS, it’s how my voice service is performing from an infrastructure level as well. I’ve got to think about infrastructure management and how important that is.”
“We hear people say they need to do four things regarding quality of service: First, ‘How can I manage my service level and how can I communicate a service level performance against, say, a service level agreement?’ That’s classic SLA management,” says Guthrie.
“Second, they want to resolve problems faster,” says Guthrie. “These problems, while they may permeate as quality of service, or as bad phone calls, it’s more than that, and the customer wants to look at what the problem is and how he can resolve it quickly. They also want to be able to use historical trending and analysis to determine when the problem actually began. What was the seed that germinated that caused the problem? Customers also want to minimize the duration of downtime. One can use root cause analysis, for example, to discover that just because somebody picked up a phone and couldn’t make a phone call doesn’t mean that the phone is broken. It could be any number of other issues, such as a DHCP server that’s gone off line, thus the phone no longer has an IP address and it can’t register to a call manager.”
“The third issue customers talk about is how they can minimize the cost of downtime,” says Guthrie. “In this perspective they’re thinking, ‘How can you give me some advance warning that the service is starting to degrade and that if it’s left untouched, it will lead to a total failure or a disruption?’ They want to know how to minimize the number of outages, how they can minimize the number of outages and how they can have some early warning system that says, ‘You’ve got a Call Manager that’s running out of memory’ or ‘You’ve got a router that supports the voice service that’s running out of memory’ or ‘You’ve got a DHCP server that’s malfunctioning and is disrupting the service’.”
“Fourth and finally customers want to know about capacity planning,” says Guthrie. “They want to be able to use historical analysis and trends to say, ‘If my network usage continues at the following pace, or if my voice service continues at the following pace, I’m going to require additional circuits.’ Often times when I talk about capacity planning, I think about growing capacity but I also in this regard want to talk about de-commissioning capacity. So, for example, in the voice business today, it’s important that as people make the transition from, say, a TDM-based voicemail system to an IP-based voicemail system, they can go and scrub the old data and look at it themselves, but what they really want to do is to have the ability to automatically scrub the data and analyze it and have something help them understand if their voicemail usage is growing the way it once did or if it is in fact declining. As they go to transition from TDM to IP, they want to have the ability to pause and say, ‘Wait a minute. Do I need the same capacity that I once needed or have trends and behaviors changed to the point where I don’t need as many voicemail ports as I once did?’ Something similar can happen with the total number of calls. The number of calls is important because that’s how we engineer our network, but if people are not making as many phone calls as they once did because of email, IM and so forth, customers ask ‘Do I need to plan for a different, perhaps lower, capacity?’.”
“I mentioned that de-commissioning underutilized assets is important here,” says Guthrie, “and what we’ve found early in the IP telephony era was that people over-commissioned or bought more bandwidth than they needed because they thought that over-provisioning was the cure for quality of service. That’s not only not a cure-all, but an expensive remedy. Now we’re seeing people reverse course and use capacity planning as a way to ask themselves, ‘Hey do I have more circuits than I really need?’.”
Taking Things Seriously
Ceterus Networks’ (http://www.ceterus.com) UTS and UTX platforms are transport solutions that enable Ethernet and legacy circuit-switched service delivery over copper or fiber to any location. They can deal with anything from carrier-class, fully redundant, multi-service applications, to small business integrated voice and Ethernet access services.
Mark McDonald, Vice President of Product Management at Ceterus, says, “I see QoS becoming more important every day. We do enterprise Ethernet deployments. Our particular technology carries Ethernet over DS1s, DS3s, OC3s and a SONET infrastructure. Up until probably the summer of 2007, our customers used fault monitoring capabilities in the DS1, DS3 or SONET gear to ensure that circuits were up and running and stable. Now things are more sophisticated. QoS also includes the ability to classify the traffic and prioritize it and schedule it. We have that capability. We have a number of stabilizing queues in our product and so if it comes in by port or by VLAN – primarily Layer 2 – they can ‘mark’ or ‘color’ the packets and if there’s a congestion or failure in the network, we can adjust the bandwidth so that the high-priority traffic gets through. That’s pretty standard stuff.”
“But now I see that people are thinking more and more about how they would do all this on a native Ethernet domain,” says McDonald, “where they don’t have SONET or some other backhaul infrastructure to determine where the faults are or if the circuit is even operational. There are two standards that have been kicked around. One standard that’s been approved since last summer is the 802.1ag continuity fault management, which defined the parameters and techniques you should use to determine the bit error rate on a native Ethernet link, so you can figure out the delay and delay variation, continuity checks, and things like that.”
“People are now asking about this capability,” says McDonald, “and they’re putting dates on when they want it in their network and interoperating with their own equipment, dates which are generally in early 2008. By the second half of 2008 they’d like some standard inter-vendor interoperability.”
McDonald continues, “The second standard in this group is the ITU standard Y.1731 or ‘OAM functions and Mechanisms for Ethernet-based Networks’ to support point-to-point connections and multipoint connectivity in the ETH layer along with capabilities to operate and maintain the network and service aspects of ETH layer. You hear both 802.1ag and Y.1731 mentioned together. A third item that’s been around a bit longer and you hear more often is ‘Ethernet in the First Mile’ [EFM] which doesn’t improve or replace the existing Ethernet. It’s a set of additional specifications enabling users to run Ethernet over previously unsupported media, such as single pairs of telephone wiring and single strands of single-mode fiber [SMF], thus making EFM port types suited for use in subscriber access networks; i.e., networks connecting subscribers to their service provider. It also happens to involve some performance monitoring and fault monitoring capability for a single link, which is 802.3ah.”
“Product selection in the near future will be influence by compliance with these new standards,” says McDonald.
U4EA Technologies (http://www.u4eatech.com) is a leading provider of multi-service business gateways that enable service providers to deploy IP communications solutions to SMB and enterprise customers. U4EA’s all-in-one customer premises devices are supported by the company’s QoS (GoS) mechanisms to protect delivery of converged VoIP, data and video services. The U4EA Fusion Series includes next-gen network signaling gateways used by carriers to interconnect legacy networks and equipment with next-gen networks.
U4EA’s Peter Thompson, Chief Scientist, says, “One of the things that we see is a much broader-based consensus that you really need QoS. That’s a change from a year or two ago, when many people thought that all you needed was enough bandwidth. You don’t hear that so much anymore. That’s because more people are now actually trying to use VoIP and other kinds of multimedia applications on converged IP networks, and they realize that these don’t work all the time unless you actually do something about things such as QoS. So there’s been more of a progression and consciousness that this is important, and that’s particularly the case in the service provider space, where providers know they must satisfy certain quality-of-experience expectations on the part of their customers.”
“In the enterprise space there’s a feeling that people are starting to get their QoS issues under control,” says Thompson, “which tend to involve efficiency and cost. The price of having reliable VoIP may be high since some people still maintain a completely separate network connection, just for the VoIP. There’s a sort of eternal vigilance concerning the utilization level of the WAN link and companies are pretty quick to upgrade bandwidth if it looks like the utilization is getting too high. Interestingly, Cisco still recommends that you must never use more than 30 percent of your link capacity for things such as VoIP. Of course, we’re pushing the fact that if you use superior QoS technology, then you can get much more out of your link.”
QoS is a Many-Faceted Thing
Virtela (http://www.virtela.com) is global network integrator enabling business solutions for multinational organizations. They enhance business processes by creating customized business solutions and managed network services that integrate into your existing network architecture. They can do everyting from Mobile Workforce Enablement to HD Videoconferencing.
Rob Pfrogner, Security Services Product Manager at Virtela, says, “Our play is as a global network provider, so we have global reach through our partners. We’re able to reach pretty much anybody anywhere, because we have relationships with the local access providers worldwide. We’ve also strategically located our Policy Centers – which are basically data centers where we place our own equipment in order to install our own ‘intelligence’ into the network – and that makes up our backbone that stretches from Policy Center to Policy Center. One of our challenges is providing a QoS when we happen to be stitching together different types of customer network segments. We have to make sure that everyone ‘plays correctly’ across our backbone, even though they may be using different technologies for QoS on their own network. A challenging aspect of QoS is making sure that whatever we start with is recognized and carried through or translated when we have to traverse, say, AT&T’s network which reaches customer premises #1 and then Level 3’s network which extends to customer premises #2.”
“Ultimately, QoS is a complicated and many-faceted technology whose challenges are exacerbated when connecting multiple disparate networks,” says Pfrogner. “Reserving bandwidth may be required for company A to assure their mission-critical application functions acceptably, while priority queuing is the preferred method for Company B. Virtela matches the needs of the application to the technology that fits best. If a deployed solution is underperforming, Virtela can change the QoS method to achieve the desired result.”
In the face of a growing, hybrid network maintaining end-to-end voice and video quality is an ever-increasing challenge. Fortunately, sophisticated tools and policies for the enterprise – along with a bit of help by technology running on the service provider side – should keep multimedia communications crystal clear for some time to come. IT
Richard Grigonis is Executive Editor of TMC’s IP Communications Group.
Today @ TMC
ITEXPO West 2012
October 2- 5, 2012
The Austin Convention Center
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
Cloud Communications Summit
October 3- 5, 2012
The Austin Convention Center