back to article Really Scary Telecoms Stuff? Nah – telephony's just an app

In 2009, I moved to Jersey to become the network and telecoms manager for a multinational company. It was tremendous fun, as I had a variety of kit to play with. I tended to favour the Mitel 3300 ICP range (still do, actually) that supported about half of our offices, and I did the various engineer courses and exams for the …

  1. Ragarath

    Hosted PBX's are great and I am looking at them the sites I support. These are only small sites and so can easily go talk to one another.

    For big sites what issue would there be if you lost connection?

    1. A Non e-mouse Silver badge


      It depends on the solution ;-)

      For some, you're stuffed.

      For others, you can get local devices which can support a basic level of working if the hosted/cloud data center becomes unavailable.

      Remember, that if you're running business critical services in the cloud, you probably want to have two diverse ISP connections to the Internet.

      1. Anonymous Coward
        Anonymous Coward

        Re: @Ragarth

        Remember, that if you're running business critical services in the cloud, you probably want to have two diverse ISP connections to the Internet.

        That assumes the cloud provider does not go down. If they do you are really stuffed.

        1. A Non e-mouse Silver badge

          Re: @Ragarth

          That assumes the cloud provider does not go down. If they do you are really stuffed.

          Absolutely correct. But the same applies to on-prem kit too: If the PSU in your PABX goes bang, you're stuffed until a replacement/fix arrives.

          1. Anonymous Coward
            Anonymous Coward

            Re: @Ragarth

            Seems an odd way of going about things...for something as critical as a whole-company PABX, shouldn't you have an on-prem set of spares? I would.

          2. Christian Berger

            Re: @Ragarth

            "Absolutely correct. But the same applies to on-prem kit too: If the PSU in your PABX goes bang, you're stuffed until a replacement/fix arrives."

            Actually not: First of all you can have a whole second PBX as a compete spare sitting around... which isn't particularly expensive. Then most PBXes use standard ATX power supplies you can find everywhere. They also use common PC components and since most of the appliances run Linux, you can simply swap the mainboard and it'll just boot.

            Second: Most appliance PBXes are also available as "Software Only" solutions where you get some ISO-image and just boot it up. Since they don't require any special hardware, you can simply install them on any virtual host you want.

            And again, your problem probably won't be that your system will be fully down. Your problem in such situation will be, that occasionally the voice quality will be bad. Or that your uplink will be saturated by internal calls (not every hosted PBX leaves the voice streams locally as in a world of NAT, that's actually a hard problem).

            So essentially it is a stupid idea to use a hosted PBX. There's way more things that can and will fail, and far less things you can fix or even diagnose.

            1. Anonymous Coward
              Anonymous Coward

              Re: @Ragarth

              "If the PSU in your PABX goes bang, you're stuffed until a replacement/fix arrives"

              Since 1996 our PABX has had one replacement PSU. It's got two redundant PSU's though, so the failure was unnoticed until the annual checkup. "they don't build them like they used to" was the comment of the engineer, as he struggled to lift a PSU around the size of a small desktop PC and suitable for providing electricity to a small village.

  2. Christian Berger

    That's pure insanity

    I mean with a hosted solution you will have a lot more parties to shift the blame... and also a _lot_ more things to go wrong. Remember VoIP requires pristine connections with low latency, low yitter and next to no packet loss.

    Typically you are much better off with a small PBX, for example a "Starface" or something else based on Asterisk. There are certain vendors like Mitel or 3CX which tend to offer you solutions that are just plain broken.

    Telco stuff isn't scary if you are prepared to learn a bit and not choose the worst solutions. Unfortunately there are people who refuse to learn things and reliably choose a solution in the lower 90%. Unfortunately many of those believe they know how to install/run some PBX.

    1. Anonymous Coward
      Anonymous Coward

      Re: That's pure insanity

      I mean with a hosted solution you will have a lot more parties to shift the blame... and also a _lot_ more things to go wrong. Remember VoIP requires pristine connections with low latency, low yitter and next to no packet loss.

      There's also the small matter of confidentiality. We require fairly high levels of proof that communication is secure, and that's not possible with a hosted PBX. At most, we host the PBX in a spacial data centre that we have screened, but there's no way we will allow a 3rd party to control the PBX and its VPN circuits..

  3. A Non e-mouse Silver badge


    ..single ISDN links because failures were generally in the PBX, not the line

    Really?!? I've been in the telecoms/IT world for 20+ years, and the BT ISDN lines have always been the most unreliable part of a phone system.

  4. Anonymous Coward
    Anonymous Coward

    If you wanted all the features without the box or the hassle, spread over multiple sites, then why didn't you go withvBT Feature net (5000?). They were doing it from at least 20 years ago.

  5. Slx

    I think people overstate the reliability of the old POTS networks. The exchanges and switches rarely went wrong but the plant isn't all that reliable. If someone digs up a cable somewhere in the local network or cuts an overhead line, your service is gone.

    How often is an enterprise grade IP connection going into fail?

    If you're a serious site, surely you would have multiple redundant links to different networks so that you could at least maintain your core services in case of disaster ?

    1. Christian Berger

      That's actually not the topic

      VoIP can provide reliability advances as you can just switch your IP uplink. What the article is talking about is to use "hosted" PBXes where you have little control over your PBX any more as it's just on some virtual system hosted by some company.

      Such an idea may have made limited sense in the TDM world where PBXes were expensive, but today running your own PBX can be done at next to no cost.

      1. Anonymous Coward
        Anonymous Coward

        Re: That's actually not the topic

        "How often is an enterprise grade IP connection going into fail?"

        If your internet connection is still over copper, then your IP comms are still going over POTS.

        Even if your on FTTP then your still subject to digger related problems, since at least in our area the fiber is in the same duct as the POTS cable. If you have connections via multiple suppliers, then they tend to use the same physical route to get back to the exchange and are still subject to the same digger related issues. So the enterprise connection is likely to fail about as often, if you have madmen with diggers in the area.

        You can eliminate these problems quite quickly if you track the responsible people down... and take a photo of their company van and (probably hired) digger. Passing the photos to BT's responding engineer is like beating them to a pulp, except done financially by BT's billing/legal people when they bill them for the repairs.

        1. Anonymous Coward
          Anonymous Coward

          Re: That's actually not the topic

          Reminds me of a data center that boasted quadruple backbone uplinks. They all entered the premises through the one trench that happened to get dug up by a backhoe.

  6. DonkeyOaty


    My business(s) run a simple asterix install, hand configured. I had no clue when I started, to having full system configured in a couple of weeks.

    If you have < 300 people, and write the macro's carefully, you really can manage this in text-file, for free. I actually host it on a google cloud virtual linux server for good measure.

  7. Christian Berger

    There's actually another very real danger if you host your PBX outside of your network

    If you have a local PBX on your LAN, it doesn't matter how secure your internal credentials are, unless you're very stupid, nobody will be able to pose as an internal phone.

    How if you have a cloud PBX all you need is your credentials to pose as an internal phone... and if you use an app, that means that you store your credentials on a highly insecure device.

    This is a real danger as calling someone involves money. Perhaps not when you call someone locally, but there are providers like which offer your premium rate numbers all over the world. Just get one of those numbers and call them over your snitched PBX login.

  8. The Original Steve

    Even run your own PBX as a hosted solution!

    Just to add something else into the mix...

    I work for a small MSP with a slant towards MS products. Over the last 3 years or so there has been an explosion in requests for Lync / Skype for Business across the industry as a whole.

    Most of the time, we've installed SfB as a single VM (standard edition) using SIP trunks (although sometimes we needed to stay with ISDN so needed a SBC to get that plumbed in), or for larger clients that need HA and even DR for telephony, we've deployed the full enterprise edition of SfB (so 12 VM's for no spof's / HA, and then the same again in another site for the DR element).

    Customers love it, with the biggest challenge being the culture shift. I disagree about the need for DDI's. A lot of modern, SIP focused PBX's these days have phone numbers as a simple abstraction layer to the user's actual SIP address. Nobody in our office has a DDI (other than me, for testing naturally!). Everyone call out, and each department has it's own response group (e.g. hunt group) that people can call in for. We also have a IVR where you can pick your department.

    Users expect:

    - A dial tone

    - A ring-ring noise when a call is holding in a hunt group

    - Phone numbers for each phone

    - That physical handsets are where the number lives

    The above are all perfectly reasonable, as that's how traditional PBX's operate. But in the world of SIP, and in particular SfB (but most other modern PBX's) things work a little different:

    - There is no dial tone

    - When holding in a hunt group, you get music or whatever audio file the PBX admin has specified

    - Phone numbers aren't a real thing. They are simply a way to map a number to the real SIP address that's really used

    - SIP addresses (e.g. phone numbers) are mapped to a user, not to a device

    That last one is a real cultural shift for end users, and one of the hardest parts of a deployment. User centric services is where IT has been heading for years, but telephony has traditionally been somewhat behind the times. Getting users to understand that when dialling a number (SIP address!) the phone system is attempting to reach YOU - the end user - NOT the device. Which is why your handset, mobile app and softphone can all be ringing at the same time. The person calling you is after YOU - the fleshy bit - they caller isn't trying to call a particular endpoint.

    Anyway - back on topic...

    A satellite office for one of our larger clients based in Zurich has about 15 users and needed a new phone system. Building SfB wasn't financially viable, nor was buying a traditional PBX. We therefore put them onto a hosted SfB platform by a company called AlwaysOn. They sent out some awful SNOM handsets and once they had internet access they were working. The reliability and support was truly terrible. After 9 months the client demanded something else as they rely on telephony and this "cloud" solution sucked massively.

    Technically, I couldn't see any reason why it wouldn't work - it's just that the supplier sucked massively, rather than the wrong technology.

    So, I built out a new SfB Enterprise Edition (e.g. no spof HA) platform in our Cambridge datacentres on a shiny new AD domain dedicated to this client. Sent out some decent Polycom phones and a load of Jabra headsets for softphones and 12 months later the client is still loving it. The connection is over the internet from Zurich without QoS or MPLS, with all traffic over the edge servers.

    Essentially it's a dedicated, private hosted SfB.

    The client was so happy with it in their satellite office, the head office in Cambridge wanted to move to it as part of a large office move, giving the platform a total of 280 users.

    As we as the environmental (2 x datacentres, UPS + diesel gen, access control, N+1 AC etc.), physical (Hyper-V cluster, HP SAN's) and application (SfB enterprise edition / clustered) layers all having security, reliability, performance and capacity all baked in as part of the architecture, as the author rightly points out there's no point having all of this but no SIP trunks. Our SIP trunk provider (PureIP if anyone is interested - they're amazing!) has given us two SBC destinations to point our calls to, with each SBC in a different datacentre running as an active/active pair. The same applies for inbound calls - either SBC can send the calls to either of our meditation servers. So should a whole datacentre fail with our telco then there is zero interruption of service.

    With all of this, we've essentially created our own dedicated hosted SfB platform for one of our largest clients.

    I'm aware that SfB can operate as a multi-tenanted configuration, although I haven't tried it. Our client is delighted with the system they have with reliability and also feature set (visual voicemail, IM and presence, conferencing, whiteboards and screen sharing, file transfer, telephony (inc, delegate, IVR's, hunt groups, private line, 2nd line etc.) voice access to their mailbox, enterprise chat rooms etc.), but as it's hosted by us there is no real capital cost involved. Our main cost is licensing, which is 3 x SfB server licences that we brought via SPLA (so about £240).

    I put a similar system in on premise for another large client, but also wanted DR (so the above x 2) which cost around £90k. Compared to the hosted option I have to say it looks like the writing is on the wall for on premise PBX systems.

    With Office365 including SfB Online - and MS now offering Cloud PBX / PSTN Calling (MS provided SIP trunks), combined with the insanely expensive capex needed for on premise, I really can't see many companies opting for on premise in 5 - 10 years time.

    Hence being rather proud of the work we did to get this private hosted system in place. If nothing else it clearly demonstrates that it can be done with good results, and I think the market very much has an appetite for it.

    Sorry of the long and rambling post. After reading this article as well as the ISDN to SIP article I got all passionate again about the SfB projects I've done in the last 2 years and simply wanted to talk codshit about it!


    The market wants hosted. If you're a IT service provider / MSP then it may well be worthwhile looking into YOU creating a multitenant IP PBX system yourself and selling it directly to your clients.

POST COMMENT House rules

Not a member of The Register? Create a new account here.

  • Enter your comment

  • Add an icon

Anonymous cowards cannot choose their icon