Addressing 'VoIPmageddon': Vetting SIP Trunking Providers
Understanding how your VoIP provider handles calls across its network is as important as knowing how your VoIP implementations run in your own network.
In a recent No Jitter post, "VoIPmageddon: Is Quality Leading to a Telephony Meltdown?," and corresponding session at last week's Enterprise Connect conference in Orlando, Phil Edholm, founder of PKE Consulting, highlighted an important issue in telephony today: the increasing challenge of delivering quality voice as the number of VoIP endpoints grows. The challenge is indeed quite real, but mitigation is possible.
The following best practices can help providers assure delivery of quality VoIP in today's "fractured" world. Enterprise communications decision-makers would be well advised to make sure their providers of choice embrace them.
- Use of SIP trunking: As Phil recommended, VoIP providers need to use SIP trunking as much as possible for delivery of voice calls to the customers they service, and for origination from the carriers with which they peer. In this scenario, the provider would only use TDM in the local access network, and VoIP from a local point of interconnection with the PSTN to the customer. By being as close as possible to the originating carrier, the provider can avoid the multiple TDM/SIP conversions that are so detrimental to quality and latency. Replacement of TDM interconnections with SIP-based interconnections is especially relevant where the core network is already packet-based.
- Work with tier-1 providers: Providers that only work with tier-1 national voice network operators for core services, instead of voice services resellers, further reduce the number of hops. Tier-1 providers operate their own local network infrastructure and have direct interconnections with other tier-1 operators. By doing this, providers can avoid using "grey routes" in the TDM world, where cheap routes introduce delay and quality issues resulting from the usage of tromboning, poor quality interconnects, or even illegal interconnection schemes.
- Active codec management: Where calls are still delivered to a provider as TDM, the choice of VoIP codec should mirror what's in use at the end customer, from the local PSTN-to-VoIP gateway to customer PBX or service platform. This eliminates any transcoding in the delivery. Actively managing codec choices across a global IP-based backhaul network minimizes transcoding steps by obtaining the desired codec as early as possible in the chain of providers.
- Plenty of points of presence: Having a large number of points of presence is a critical quality assurance factor. This is true both for TDM and IP-based interconnections with carriers, and for SIP trunks with customers. By having points of presence in a large number of countries, a provider can assure that traffic comes to it directly from the local carrier, eliminating the multihop SIP issues that Phil discusses in his post.
- Interconnect with the "best" point of presence: By connecting SIP trunks with customers and with local carriers' points of presence with the best IP connections, providers can minimize hops and transcoding for best quality. For best results, the selection of the closest point of presence should be done on a per-DID number basis for each customer. While the connections between points of presences and SIP trunks are static, management should be done in real time through Web-based configuration. As such, providers can easily change the point of interconnection following changes in external network conditions. Using WebRTC, providers can take this a step further. At Voxbone, for example, we have implemented a dynamic "Closest POP Selection" mechanism to reduce the IP network's impact on quality by choosing the point of presence that has the best IP path for quality on a per-call basis. Global providers should have IP points of presence on all continents, connected with global ISPs. This reduces the looping caused by random peering and potential IP hairpins. Providers should allow their most quality-conscious customers to interconnect physically in one of their data centers, but use delivery over the Internet as backup for their private interconnections.
As Phil indicated, voice quality in conferences is more easily perceived because of the statistical probability of having multiple VoIP endpoints in a multiparty conference. So if your company operates a conferencing service, additional requirements apply. For example, your VoIP carrier should provide international and domestic SIP backhaul capabilities to your conference bridge.
To maximize voice quality for conferencing, providers should follow similar steps to those outlined above, aggregating the traffic directly to the conference bridge using SIP, and integrating both the PSTN-originated and WebRTC traffic. This minimizes latency and transcoding. The choice of a dedicated IP-based backhaul network is essential to minimize transmission latency and reduce jitter. As a result of this, conferencing providers can reduce the size of required jitter buffers, reducing the latency for conference calls.
The same goes for contact center environments, where similar approaches as stated above are critical -- a challenge made more difficult given that many remote agents are a far distance from the actual contact center core site. Peer management is particularly important in this market, with the use of local points of presence for both customer and agents advisable. In addition, the provider should optimize the backhaul system for latency and transcoding to ensure delivery of a quality experience.
For customers with multiple locations, having direct SIP trunks at geographically diverse locations versus a single trunk using the enterprise IP infrastructure is also critical for both quality and cost. By tying real-time traffic delivery to the best point of presence for the associated endpoint (the enterprise PBX or UC platform), a provider can actively reduce the loops created in the backhaul system when there's only a single point. For example, if a single SIP trunk point is chosen in California, then an employee in New York talking to a client in Europe would essentially be hairpinned -- adding 70 to 80 milliseconds of delay or more (dependent on jitter buffers) to the conversation. By providing separate SIP trunks in both California and New York, a provider can optimize the quality to the endpoint's location. Easing the mapping of those access points to users, and enabling numbering schemes through per-number call delivery, are very helpful for managing quality in large deployments.
Is My SIP Trunking Provider Following These Best Practices?
All of these best practices are, quite frankly, common sense for a SIP trunking provider looking to deliver a quality voice experience (or in the future, video) in the evolving VoIP world. Organizations buying SIP trunks for their real-time applications should develop a list of these capabilities (for example, location of PBX platforms, IP connectivity at these locations, etc.), and discuss them with their potential providers as part of the acquisition process. In turn, SIP trunking providers should help their customers define elements in their networks that can be optimized to enhance quality.
As Phil pointed out, it is no longer enough to think about how your VoIP implementations run in your own network. In the new world of ever-increasing VoIP endpoint adoption, the choices you make in how you extend VoIP communications outside your network have a much bigger impact on the quality of your service (and your company).
Gaetan Brichet is COO of Voxbone