13 Comments

Summary:

It’s true that Twilio just named successful Silicon Valley prognosticators Founders Fund and Mitchell Kapor as initial investors, but the real Twilio story lies in its infrastructure. Or rather, its lack of infrastructure. The San Francisco-based startup hosts all the computing functions of its web developer-friendly […]

twilio_logoIt’s true that Twilio just named successful Silicon Valley prognosticators Founders Fund and Mitchell Kapor as initial investors, but the real Twilio story lies in its infrastructure. Or rather, its lack of infrastructure.

The San Francisco-based startup hosts all the computing functions of its web developer-friendly telephony service on Amazon Web Services, a business model that enables Twilio to differentiate itself even further from its voice API cohorts. Upfront CAPEX savings by Twilio trickle down to users in the form of an inexpensive per-call pricing model. By leveraging Amazon’s cloud platform, Twilio’s infrastructure grows (and shrinks) right along with customer demand.

As with all things invoking “cloud computing,” however, there’s a catch: Twilio customers’ applications are at the mercy of Amazon. If EC2 goes down, phone calls don’t go through. CEO Jeff Lawson’s faith in Amazon might provide some comfort (“If I say, ‘I’m a startup and I run my own data center,’ you should run like hell,” he joked to me), but the prospect of uncontrollable downtime might give users pause if their telephony apps are tied to measurable revenue. Twilio doesn’t offer an SLA, either.

That said, Lawson says Twilio’s platform has been up and running at 100 percent since the company launched in November, and users can check current availability at status.twilio.com.

Another noteworthy aspect of Twilio is that it might not be alone for long on the development front. Voxeo will announce today that it’s launching a platform to let developers write telephony apps in standard web languages.

  1. Sounds like a non-problem. If Twilio has run at 100% and outsourced any potential problems to a reliable web behemoth, why expect it to go down? Moreover, EC2 is “uncontrollable” only insofar as Amazon is unresponsive to its web-services customers. Unless you have proof of that, your argument appears to be a non-starter.

    Share
  2. Great piece Derrick.

    Not only are we providers of cloud services, we’re also pushing the economics of telecom by consuming cloud services as well. This is a very insightful look how our infrastructure enables our customers’ applications scale up and down, with no pre-defined “per-channel” caps as most telecom providers impose. And because we haven’t had to invest millions in enterprise hardware, we can pass the savings on to our customers.

    I also wanted to note that we maintain infrastructure not only in Amazon’s 3 geographically distributed data centers, but also at other cloud hosting providers, including as Rackspace. Between Amazon’s reliability (which we’re confident already surpasses in-house hosting or colocation) and the ability to move capacity to another cloud on demand, we’re confident that our architecture will meet the needs of our customers.

    Thanks again for the insightful post,

    -Jeff

    Jeff Lawson
    Co-Founder & CEO, Twilio

    Share
  3. I would have thought that the Amazon services are not nearly deterministic enough to ensure QOS for VOIP processing and call setup / routing. If my assumptions are indeed wrong, as they are offering these services, then this is a big boost for AWS.

    My question: IS the Twilio call routing infrastructure supplemented in any way by dedicated VOIP processing hardware?

    Share
  4. Additionally, there has to be an interface to the PSTN.

    Share
  5. Derek, your status link actually goes to Twilio’s pricing page. Looks like an edit artifact. Good write up otherwise.

    Share
  6. [...] for sales and marketing and infrastructure services. In addition to TechCrunch, posts appeared on GigaOm and TelephonyOnline and included details about Twilio’s new customers including Cheetos, [...]

    Share
  7. Alan, I cannot answer for Twilio, but I would guess that dedicated VoIP hardware is not used.

    I am saying this based on our architecture at Ringful.com, which has been around a little longer and plays in a similar space. We offer a RESTful Web Services API for telephony mashups to subscribers worldwide.

    Ringful.com has been running on a combination of Google App Engine and Amazon EC2 since day one. We’ve had two occasions for the past year when one of the EC2 nodes became unresponsive to the point that we couldn’t ssh and had to restart the image. Other than that, the cloud services have been great for customers calling all over the world. Even quality of conferencing, which requires real time processing of multiple voice streams has been great.

    Share
  8. I think you inadvertently stumbled on the correct headline. AMZN will make or break Twil not on reliability — AMZN needs that for far greater reasons, but it could easily break Twil by expanding its notion of cloud services to exactly what Twil is offering. With all due respect to the good work Twil has done, such a move would be a fairly incremental conceptually and make a lot of sense in AMZN’s cloud offerings.

    Or they could make Twil by buying them, which is more likely.

    This is the company eBay should’ve bought instead of skype.

    Share
  9. [...] As an interesting side note, Twilio is built for and runs on Amazon’s cloud infrastructure (link). [...]

    Share
  10. Twilio can always build their stack from ground up. Others are doing it: http://tinyurl.com/dgva2d

    Share

Comments have been disabled for this post