Quality of Service, Part 1 of 2: Elements of Enterprise QoS for Voice Over IP
- The Role of QoS in Network Services
- Rationale for VoIP
- Elements of VoIP Delay
- Elements of VoIP QoS
- Application-Specific QoS Requirements
- Migration to MPLS
Voice is just one service of many being rolled out in enterprise networks, branches, and home offices, but voice has its own peculiarities! Specifically, voice-over-IP (VoIP) is a real-time service and is sensitive to delay and packet loss. VoIP deployment may be complicated by the need to maintain legacy equipment such as PBXs, vendor-specific phones, and so on. The need to migrate legacy equipment onto the IP-based cores of the future is increasingly crucial. Happily, the elements of VoIP quality of service (QoS) are essentially the same as the QoS needs of other services: throughput, availability, delay, delay variation, and loss. But other applications have their own QoS needs, and reserving resources for one technology (such as VoIP) may have the disagreeable effect of starving other applications. QoS deployment is a skillful balancing act. Multiprotocol label switching (MPLS) may help the network designer and service provider in rolling out these services.
The Role of QoS in Network Services
The total allowed one-way delay budget for IP-based voice service is generally put in the range of 130150 milliseconds; that is, a maximum of 0.15 seconds. Delays above this level become conversationally noticeable in the form of missing words. Imagine a hypothetical lottery provider phoning Mr. Terry Dactyll to say that he has won $1 million:
Lottery: "Hello, we're calling to inform you of the lottery result." (No delay.)
Terry Dactyll: "Oh, that's unexpected; did I win?" (Followed by a 150 ms delay.)
Lottery: "Are you there?"
Terry Dactyll: "Yes."
Lottery: "Yes." (Answer to delayed question.)
Even though the network loses no data, our unfortunate Terry Dactyll is put through the wringer by the long delays.
Legacy telephone networks have done an admirable job of managing such problems, but service providers increasingly want to migrate to common IP-based cores. This means that time-sensitive applications must be engineered to provide the required end-user experience over an IP-based core. In most cases, this setup looks like Figure 1, in which an enterprise uses IP phones on its various sites and employs one or more service providers to interconnect the sites.
Figure 1 Corporate IP voice service.
Each site in Figure 1 hosts IP phones that plug directly into a LAN interface. In many cases, the headquarters site may host a legacy PBX, perhaps because the enterprise wants to depreciate its expensive legacy telecom equipment and migrate to an IP-based approach. The branch office in Figure 1 uses its WAN link to interconnect its IP phones to the enterprise. In this way, even geographically remote offices can enjoy the productivity-enhancing service benefits traditionally enjoyed only by headquarters.
Home Office IP Phones
Why would home office workers want to use IP phones? In the branch office case, the more applications the home worker can access, the higher the worker's productivity. Another saving comes from the fact that an IP phone user typically doesn't generate additional bills. This means that there's no need for the home user to submit expense reports for use of his or her home phone service. Administrative savings resultno need to process expenses, generate payments, and so on.
Service Provider Links
The links into the service provider core in Figure 1 can use any of a range of technologies, such as Ethernet, asynchronous transfer mode (ATM), frame relay, SONET/SDH, etc. This is called the "handoff" technology. Interestingly, Ethernet looks like it will become the link technology of choicejust as it's the de facto LAN standard. In fact, Ethernet continues to conquer the world and has even appeared in optical transport networksamazing for a flexible 30-year-old technology!
The headquarters site in Figure 1 uses an ATM link, whereas the branch office uses frame relay. The latter may be due to limitations of the local service provider rather than choice. The service provider may not yet have deployed ATM in the branch office location, or the latter may have an unexpired frame relay contract.
The VoIP users in Figure 1 all use their IP phones in a transparent fashion; they don't care that they're not using a PBX-based service. However, the operation of VoIP is very different indeed. Let's take a look at the various delay elements that combine to make up the overall delay budget of 150 ms mentioned earlier. Figure 2 illustrates the end-to-end process.
Figure 2 End-to-end VoIP process.
Depending on the system design, the various blocks in Figure 2 may be executed by the phones, by routers, or by a combination of a PBX and a router. Regardless, the functional blocks are the same, as are the associated delays. As we saw in the earlier hypothetical conversation with Mr. Dactyll, delays have the greatest QoS impact.