I’m at Sys-con’s Cloud Computing Expo in NY today and tomorrow. Wifi willing, I’ll be live blogging several sessions.
First up, is a keynote by Dr. Werner Vogels, CTO of Amazon on The Power of Infrastructure as a Service. Here’s the abstract:
“Building the right infrastructure that can scale up or down at a moment's notice can be a complicated and expensive task, but it's essential in today's business landscape. This applies to an enterprise trying to cut-costs, a young business unexpectedly saturated with customer demand, or a start-up looking to launch.
There are many challenges when building a reliable, flexible architecture that can manage unpredictable behaviors of today's internet business. This presentation will review some of the lessons learned from building one of the world's largest distributed systems - Amazon.com. The focus will be on state management which is one of the dominating factors in the scalability, reliability, performance and cost-effectiveness of the overall system.”
Goals of presentation:
1. How infrastructure as a service is a reality through Amazon Web Services
2. Explain how Amazon needed these services themselves
Werner shows video made using Animoto, Animoto owns no infrastructure, all on Amazon’s services: S3, EC2, SQS.
Key to Animoto is maintaining SLA with users, the service depends on instant gratification. Recently launched Facebook application, which significantly spiked service within hours. Using EC2, could accommodate growth.
Another example, Nasdaq Replay. Didn’t know if there was market for this application, so leveraged S3 for data storage, minimized risk.
Shift of infrastructure from being capital expense to variable cost.
Another example, SmugMug. Started with own infrastructure, ended up focusing too much time on storage management rather than on furthering application and business. Moved to Amazon S3.
Forces driving new resource models:
- uncertainty: acquire & release resources on demand; pay as you go
- more missed them
Another example, Eli Lilly, acquire and fire-up single service in-house, 3-4 months, on EC2, 5 minutes. Werner stresses more important, is when you don’t need that computing power anymore, you can get rid of it. Not that easy in the enterprise in-house model.
More examples, NY Times Time Machine and Washington Post.com. Washington Post example is Hillary Clinton’s 8 year schedule as first lady. Received as PDF, made searchable and available on web within 6 hours.
Bild.de citizen journalism site built, running and reliable in 4 weeks.
Shows chart of how Amazon’s web services are using more bandwidth than Amazon’s retail sites. Highlighting the rapid growth of Amazon Web Services.
Now, talking about Amazon’s internal history. First, app servers and databases that required continued work on scale, for Amazon, and hosted store sites (Marks & Spencer for example). Next, moved to SOA, no direct database access allowed anymore, databases don’t scale, only access data through API.
Service-orientation not just architectural principal, but also organizational principle. Team that built it, maintains it, keeps it running.
Problem with this became that each team spent 70% of time working on infrastructure issues – network, highly available storage services etc. So, moved to shared services model, virtualized compute (EC2), storage (S3, Simple DB, EBS) and messaging (SQS). Freed up engineers to spend time on delivering business capability.
Noticed that engineers would access and hoard compute resources. Built EC2 with easy access, release and reacquisition to change engineer behavior. Would release if confident could get the resources back.
Speaking of how different storage offerings evolved from application data storage & access patterns (use cases).
Never expect infrastructure work to go to zero (from 70%), it is about 20-25% now, and that’s acceptable.
Amazon Infrastructure Services design points: Scalable, Cost Effective, Reliable & Secure
A point Werner is emphasizing is that Amazon’s services can be used independently.
Now, he’s talking about Stax, Java applications in the cloud, Wolfram’s Mathematica and a few others. Lots of examples.
Enterprise partners, less cool, but certainly important: IBM, Oracle, Microsoft, Red Hat, Sun, Salesforce & Capgemini
Overtime cloud providers will be compared on these 5 points: Security, Scalability, Availability, Performance, Cost
Werner says that today is still day 1 in cloud computing.
“Cloud computing is a reality today, only because of Amazon”. All you need is AWS Url & credit card.
Comments