A forum of IT and telecom professionals were asked what to use, SIP or PRI? Here are their responses:
I need some advice. We are looking for a new phone system. We are leaning toward going with SIP trunks. I get different opinions about SIP and most of the vendors for whatever reason are pushing PRI. Some vendors say PRI is more reliable although my research has shown it is expensive. Other vendors say SIP is not as reliable but it will save us money.
Which way should we go?
I would talk to your carrier and find out what services they offer and compare pricing on the two. You may not be able to obtain SIP in your area or find that you need a gateway on your new PBX to use SIP trunking. If so, your savings would be delayed. Ask for references from customers comparable in size that are using SIP trunking.
Reliability of SIP trunks is a dependent on how you design your environment. The fact is that a great deal of Public Switched Telephone Network (PSTN) traffic is already IP based. I am using SIP trunks internally between PBX and peripheral systems today with great success and hope to have a unified communications service running over SIP in the next few weeks. I’ve also just kicked off a project to convert 90% of our PSTN traffic to SIP via our provider. This will replace over 600 channels with SIP trunks over two diverse paths and Central Offices (CO) with double the bandwidth I have available to me now. If one CO or path fails the traffic will failover to the other automatically. I will spend less per month for nearly double the capacity I have with the PRI’s today. It’s not a huge savings but I’m improving my throughput and business continuity.
More to consider if you are going to install a SIP phone system:
- To bring PRI’s into a SIP phone system will require gateway hardware with PRI modules. This adds additional hardware, software and substantial cost. It also adds complexity, additional points of failure and Firmware compatibility issues during upgrades. You are basically reverse engineering your SIP phone system (PBX) to make it compatible with the PSTN. Bringing trunks in directly via a SIP trunk from your carrier will be simpler to support and troubleshoot.
- Your carrier will likely recommend dedicated MPLS circuits for the SIP interface. You shouldn’t share your circuits with your Internet Service Provider or other services and risk contention issues.
- Carriers will tell you not to route fax traffic over IP trunks but I’m not sure. They wanted me to use G729 codec and I requested G-711 which has no compression and should work. This cuts down on the amount of channels but should allow T.38 faxing.
- A 50mb circuit at G729 can support about 1200 channels. At G-711 you will get about 546 channels.
Your question is a very valid one, but as with many telecom related questions the best answer is that” it depends”. SIP implementations can be very reliable and cost effective in comparison to PRI, but ultimately it depends on your particular needs and infrastructure implementation.
Factors to consider are:
- Is your data network robust enough to handle to additional voice bandwidth needs?
- How critical is your voice traffic to your business? How would you be impacted if the data network goes down?
- Do you need a backup voice scenario?
- What IP PBX hardware are you using or planning on using? (Different SIP services are more compatible with certain hardware solutions.)
- How well does your service provider handle Class of Service provisioning?
- Is your traffic primarily intra or inter company?
- What is your typical voice usage for local, intrastate, interstate and international?
- Are using considering hosted PBX services or a premise based solution?
Are you planning on using unified communications?
There are these and several other questions that must be looked at to answer your question more completely. The potential of SIP is more efficient, flexible, and cost effective traffic handling. That can be realized with proper implementation. PRIs have traditionally been considered more reliable and provide you the ability to separate your voice traffic from your data traffic. Many companies that implement SIP trunking still use PRI and POTS lines as disaster recovery options. SIP can be made very reliable with good network design, but can be a nightmare if implemented poorly. Because there are more variables affecting call quality in a SIP environment, many vendors will recommend PRI because it is a known quantity and easier to implement. However, if you do your homework and what’s necessary to ensure a successful SIP implementation, then the technology can be a very viable alternative.
The other factor to consider is cost. Different providers offer different pricing models, with some based on number of sessions, number of concurrent calls to be supported, or with varying cost per minute charges for different types of voice traffic. In some cases, these per minute rates are more flexible than the rates for traditional voice services. Some SIP service providers may bundle blocks of “free” minutes make their offerings seem more attractive.
There is a quality penalty for going from the uncompressed G.711 to G.729. There is a subjective measure known as Mean Opinion Score that rates quality, 5 being the best. G.711 runs about 4.1 while various offerings of G.729 run from about 3.7 to 3.9. The lower the score the lower the voice quality. Lower voice quality could affect your customers’ opinion of your organization.
Also, don’t be confused if carriers talk about the relative data rates of G.711 and G.729, 64 kbps and 8 kbps, respectively. Once you packetize the data stream with the required overhead, the relative increase in simultaneous conversations supported over a given bandwidth by the lower data rate standard is more like 2 or 3 to 1 instead of 8 to 1.
Separating the voice traffic from the data can be a good strategy, and this can be done logically or physically. Physical separation may offer some added benefits, but may require additional hardware and circuits. Separation can be done at both the LAN and WAN levels. Implementation within an existing voice cable infrastructure is also possible through available hardware products. They can save you money by eliminating the need to run additional cabling to support IP voice traffic. This scenario would also be beneficial in protecting your voice services from a data network outage. The right type of Request for Pricing (RFP) can help you identify the best IP PBX solution at the lowest cost. The same is true when selecting a SIP services provider. Unified communications support and mobility options would help differentiate hardware and service providers.
One thing to inquire about is how E911 is handled by the SIP provider. Proper handling may require the use of a VPC (VoIP Positioning Carrier) which might add to the cost. VPC might be bundled, but ask to have it broken out so you can compare the features with the expense. You may be able to deploy a more general Public Safety Answering Point (PSAP) reporting, and then enhancing internally with a solution that is software based and runs in a Voice Mail server slice.
There are many details to consider and SIP trunking adds some complexity initially, but more flexibility in the long run. It is more in alignment with the NENA i3 NG911 network coming.
As far as I can tell, the cost difference is close to a wash. Faxing is still an issue for many SIP providers, so if that’s important to your business, make certain you discuss this your provider. One provider offered to bring in a test circuit for us to test the faxing capabilities so we can validate the service.
The one thing I haven’t seen mentioned here is security. You’re talking about turning your voice into a network-based product and with that comes a whole new world of security issues. You become vulnerable to hacking on your data network via your voice network. You need to consider the firewall solutions of your provider. What solution do they recommend? Who will provide the firewalls? Where will the firewalls sit – customer edge or provider edge? Who will manage the firewalls? Who will pay for them? How will they handle calls when part of your network fails? You will have a network outage at some point and you will have someone try to attack it. Don’t forget to close and lock the back door.
One thing I’m interested in is how to back up DID numbers. In case of a CO failure, how can calls to Direct Inward Dial (DID) numbers be re-routed? That’s a real tough issue for conventional DID numbers from a conventional central office (CO) but I’m wondering how that might work with SIP trunking? In the case of conventional DID service, I don’t think there is an answer except to wait until the CO is restored.
The solution for SIP is dependent on provider. Some providers will let you have numbers to forward to if the SIP trunk is down. Others, such as AT&T and Broadvox, will let you provision multiple circuits at different datacenters and route the calls Either Round-Robin or Active/Passive.
For DIDs we use AT&T’s CLAR service (which is an MRC per DID) and Cincinnati Bell’s Call-Forward Anywhere Service (which is free). Both of these services need to be activated manually, but you can set up BCP scenarios ahead of time.
If you’re running SIP at G.711 (90-100 Kbps) and not using existing bandwidth infrastructure, you need to factor in the extra cost of new bandwidth (15 simultaneous conversations per T-1 equivalent) vs. PRI (23). It will take 3 T-1’s (4.5 Mbps) of bandwidth with G.711 SIP for almost the same number of conversations as 2 PRI’s.
One thing that I don’t think was mentioned is that SIP offers a great deal more flexibility than PRI. You can order trunks in any number vs. in increments of 23 with PRI. It’s easier to port remote phone numbers, allowing you the ability to consolidate your trunks and experience economies of scale. SIP allows you to avoid some of the taxes and surcharges that come with PRI, as well. Also, additional SIP trunks can be activated in days, not weeks, that PRI requires.
Honestly, DID resiliency is the biggest reason we’re looking at SIP. If you take cost out of the equation, the amount of complexity enterprise-SIP adds is a non-starter UNTIL you talk about DID resiliency. We would bring in two large trunks (in geographically diverse locations) and have our DIDs set to fail over to the alternate location in the event of an outage. We may not be able to get the calls into the site they are supposed to go in a worst case scenario (like the office becomes a smoking hole) but we could route the numbers to a location that would be capable of fielding the calls. This has the advantage of being part of the product offering so there’s not an additional Business Continuity Planning cost. Being able to route calls around a regional outage is a huge benefit.