MessageSling and Amazon Web Services

Slideshow Transcript

1. Slide 1: Ryan Angilly Founder / VP Engineering 22 September 2008 rangilly@messagesling.com --
2. Slide 2: Voicemail is old  Stuck on the phone  Deleted after two weeks  Tough to share  Tough to forward  Don't have your phone? No voicemails.  Don't have service? No voicemails.  Battery dead? Can't get your contacts either. 22 September 2008 2008 AWS Start-up Tour Boston
3. Slide 3: What is MessageSling? (for those reading these slides on the web)  MessageSling makes it easier to integrate voice messaging into your daily workflow, which over the past few years has come to be mainly email and instant messaging for a large number of people. Voicemails and contacts, for most people, are stuck on the phone. Voicemail suffers the additional problem of not having a lifetime of much more than 2 weeks before auto-deletion by the carrier. MessageSling takes your voicemails and puts them into your daily workflow by making them available at any computer, and in the process allowing you to store them indefinitely.  On the flip side, there are many times where contact information is stuck on a computer -- an email, a phone number -- they never all get onto your phone, and sometimes your phone isn't available. With the integrated, voice-searchable contact list, your MessageSling \"voicemail box\" becomes much more, allowing you to send audio messages to people's email from your phone. 22 September 2008 2008 AWS Start-up Tour Boston
4. Slide 4: Shameless plug  Started in January  Bootstrapped for three months  Raised some convertible debt  Launched at DEMOfall08 in September  Response was awesome  Lifehacker, CIO, VentureBeat  Thousands of users  Looking for $$ 22 September 2008 2008 AWS Start-up Tour Boston
5. Slide 5: How we use AWS 22 September 2008 2008 AWS Start-up Tour Boston
6. Slide 6: Benefits  Cost  $500/month will support 4 developers, QA and ~50,000 users  Performance  Latency between DB, Web, and Tel is awesome  Scalability  S3 and SQS response times are the same for 10 users or 1000 users 22 September 2008 2008 AWS Start-up Tour Boston
7. Slide 7: Gotchas  When things go down, they are sometimes out of your control  Outages  Lightening strike killed EC2 for 5 hours in June  AWS Backend node failures killed S3 for a couple hours  Upgrade increased error rates on SQS for a few days just last week  Eventual Consistency  S3 inconsistency windows  Save/get times  SQS inconsistency windows  Testing issues 22 September 2008 2008 AWS Start-up Tour Boston
8. Slide 8: Then again... Then again.... Things are rarely completely under your control.... 22 September 2008 2008 AWS Start-up Tour Boston Source: http://kecute.files.wordpress.com/2008/02/lipstick.jpg
9. Slide 9: AWS: Not just for Production  Dev machines brought online in seconds  News ways to solve problems  Problem  Poor audio quality on small instance  Possible Solutions  Recompile asterisk with a different set of drivers  Change the kernel on the EC2 instance to increase the clock rate  Throw more horse power at it and see what happens  A tough problem was solved in about 15 minutes because of High CPU instances 22 September 2008 2008 AWS Start-up Tour Boston
10. Slide 10: The Future  SQS used to queue failed jobs  Autoscale RoR Workers and Speech Rec instances based on  Requests  Amazon SES: Static Email Service  Despite Elastic IPs, mail from EC2 instances gets hosed by some spam filters  Amazon STCWAE: Something That Competes with App Engine  Dont make me deal with Apache and IPs and scaling instances. Just make it work :-) 22 September 2008 2008 AWS Start-up Tour Boston
11. Slide 11: That's all folks Questions? http://messagesling.com Scot Junkin sjunkin@messagesling.com Ryan Angilly rangilly@messagesling.com 22 September 2008 2008 AWS Start-up Tour Boston