Bits & Bytes #1- A deeper look into the technology that makes Aptela work

Posted by: Mahesh Paolini-Subramanya

This is the first post in our ongoing technology "Bits & Bytes" series by Aptela CTO Mahesh Paolini Subramanya. He'll provide a deeper look into VoIP technology and what makes it all possible.

The business of doing telephony hasn't really changed in a long time.  Oh, there has been plenty of evolution.  But revolution? Not really.  Or at least, not until we came along.  Pull up your chair, get closer to the screen, and read along…
   
Telephony systems have always been complicated things, regardless of whether they are the big ones in the Central Offices of your local quasi-monopoly that they call Switches and the little ones in your office that everyone calls a PBX, even though it’s really just a PC running some fancy software.  The big ones are, by their very nature, amazingly good at Just Working.  The problem - of course - is that you can't really afford one, unless you have pots-o-cash, or are a Large Company like GE or IBM.  As a result, you end up with a PBX, which remember, is still just a PC.  You want reliability?  Get a second PC to hang around in case the first breaks or just get a really expensive PC with dual power supplies, ECC RAM, and tail-fins.  Which works really well till you decide that you want Conferencing or want to add that 51st user or want your voicemail to show up in your email just like on your iPhone…when you are informed that you should have gotten a completely different PC with the special add-in Conferencing module, and different  tail-fins, and Just Sign Here...

Cue Aptela. When we got into the biz ten years ago, we decided that just doing things the old way - but better, faster, cheaper - just didn't cut it.  We saw the net change the world around us and felt that there had to be a better way to do telephony.  There were all these great technologies just being abirthed, such as SaaS, Cloud Computing, virtual services, and so on.  And they were just begging to be used.  Unfortunately, Cloud Computing is difficult and getting to leverage its capabilities is, well, ludicrously difficult.  You can get a neat Software as a Service environment by sticking the greatest API with the coolest GUI on the front of a ginormous phone switch and at the end of the day, you have a ginormous phone switch that is rapidly growing obsolete, costs a bundle to maintain and is only scalable through the addition of wheelbarrow loads of cash.  However, take the same great API and cool GUI, slap it on a truly distributed and virtual environment, which is horizontally scalable through the simple racking of additional inexpensive servers, ah now we're talking.

Of course, that is exactly what we did at Aptela, i.e., re-engineer the very basic meaning of what it means to be a Telephony System.  In Aptela calls are handled by our Cloud, i.e., any available server in our data-centers.  These self-same servers move calls around as necessary to Make Things Happen.
For Example, say you call in to one of our hosted clients, and are listening to the company IVR ("Thanx for calling WidgetCo, Your source for Widgets....").  Your call is being handled by a random server in the cloud - call it server Alice.  You then decide to "Press 6 to join the Exciting WidgetCo 10-Q Conference call!" which happens to be running on a completely different server - call it server Bob.  Alice moves your call seamlessly (and in the background) to Bob so that you can participate in the (Exciting) conference call.  Trust us, there is a lot of science and some small amount of magic in that happening seamlessly. And that is just the smallest, tiniest subset of the remarkably cool stuff that happens in our cloud.
   
All this, mind you, happens through the magic of Erlang - a tremendously cool and geeky environment/language (http://trapexit.org/) of which we will have a lot more to say in subsequent editions.  Suffice to say that thanks to Erlang and our implementation of it, we have a telephony system that is:

• Able to scale extremely inexpensively via racking inexpensive white boxes.
• Fast to develop against.  A typical feature takes less than a week to develop, test, and deploy.  A week!
• Trivial to develop against.  Remember the bit about Cloud-Computing being difficult?  Well, we lied.  It isn't for us.  But trust us, for most of the folks out there, it really is.
• Fault-tolerant and reliable.  Ericsson's Erlang based switches have annual downtimes measured in milliseconds.
• Clearly amazingly buzz-word compliant.  (Some people really dig this feature.)

Time is running short on this edition of "As Aptela Turns," but stay tuned for the next chapter, when we bring you more exciting news from the Wonderful World of Telecom Engineering!