Replacing PRI and ISDN with SIP trunking
ISDN switch-off dates are now fixed in most major markets. Moving from PRI to SIP is not just a like-for-like swap - it changes how you buy capacity, how you fail over, and what a phone number is attached to.
The switch-off is no longer hypothetical - BT has committed to withdrawing ISDN in the UK, Deutsche Telekom has already completed its all-IP migration in Germany, and incumbents across Europe and Asia Pacific have published equivalent timelines. If your voice still arrives on a PRI, you are running on infrastructure with a published end date and a shrinking pool of engineers who know how to fix it. That is the practical case for SIP trunking. The commercial and technical cases are stronger still, but only if you plan the migration rather than treat it as a cut-over.
Capacity stops being a physical constraint - A PRI delivers a fixed number of channels: 30 on an E1, 23 on a T1. Adding capacity means ordering another circuit, waiting weeks, and paying for the channels whether they are used or not. A SIP trunk delivers concurrent call sessions over an IP connection, and the number of sessions is a commercial setting rather than a physical one. Most Dollu customers size their trunk for the busiest hour of a normal week and add burst capacity for seasonal peaks, so a retailer running 60 concurrent sessions in March can be at 200 in the week before a public holiday without a change order. The bandwidth arithmetic is straightforward: G.711 needs roughly 87 kbps per call including overhead, G.729 about 31 kbps, and Opus sits in between depending on configuration. A hundred concurrent G.711 calls fit comfortably in 10 Mbps.
Resilience becomes a design decision - A single PRI is a single point of failure, and most enterprises knew it and lived with it. SIP lets you design around that. Register the same trunk from two locations, or configure the carrier to fail over to a second IP or a second site when the primary stops responding to OPTIONS pings. Point the trunk at a data-centre SBC and route calls to whichever office is up. Have inbound numbers overflow to mobiles or to a cloud contact centre if the SBC itself is unreachable. None of this was practical on TDM without duplicating circuits at every site.
Numbers detach from the circuit - On ISDN, a phone number is tied to the exchange that serves the PRI. On SIP, the number lives on the carrier's platform and can be pointed at any endpoint. That has three consequences worth thinking through. First, geographic numbers in most countries can be ported to a SIP provider and served from anywhere, subject to local rules on address verification. Second, you can consolidate numbering from many offices onto one trunk and one SBC, which simplifies dial plans and call recording. Third, emergency calling now needs an explicit location record per number or per registered endpoint, because the network can no longer infer where the caller is from the circuit.
Cost is usually, but not always, lower - The obvious saving is line rental: SIP session pricing is typically a fraction of PRI channel rental, and international and inter-office calls route over IP at wholesale rates. The less obvious costs are the ones that catch people out. You will need an SBC or a firewall with proper SIP awareness, a QoS-capable connection or a dedicated voice VLAN, and possibly new handsets or gateway cards for legacy PBXs. Budget for those, and budget for a period of running PRI and SIP in parallel while you migrate numbers.
Migration steps that avoid the common failures - Start with an audit: every number, every DDI range, every fax line, alarm panel, lift phone and payment terminal that touches the PRI. Analogue devices are the migration's hidden risk; some will work behind an ATA and some will not. Second, order the SIP trunk and test it with new numbers before porting anything, so you validate codecs, DTMF (RFC 2833 versus in-band), fax handling (T.38 or G.711 pass-through) and call recording without customer-facing exposure. Third, port numbers in batches, starting with the least critical range, and schedule each port for a quiet period with the PRI still live as a fallback. Fourth, only after all ranges are ported and a full billing cycle has passed cleanly, cancel the PRI.
Security is different, not worse - A PRI was hard to attack because it was physically isolated. A SIP trunk is reachable over IP, which is why the SBC matters. Use TLS for signalling and SRTP for media where the carrier supports it, restrict the trunk to the carrier's signalling IPs, use strong registration credentials or IP authentication rather than shared passwords, and set per-destination call limits so a compromised PBX cannot generate a night of premium-rate traffic. A carrier with real-time fraud detection will catch anomalies you miss, but the first line of defence is on your side of the trunk.
What good looks like six months later - The customers who are happiest post-migration are the ones who treated it as a chance to redesign, not just replace. They consolidated three regional PBXs onto one platform, added a cloud contact centre for overflow, put a second SIP registration in a different data centre, and now scale sessions with a portal change. The ones who struggled tried to make SIP behave exactly like the PRI it replaced. It can, but that leaves most of the benefit on the table.
