BarcampBrighton and Cloud Computing

Jeff Barr, from Amazon, was at BarcampBrighton as part of a long European tour, talking up cloud computing.

• apologised for going to be a little more commercial than others….but taken out the prices so it’s not a sales pitch!
• been at Amazon for 6 years.   Saw real potential, the first catalog service.  they started sending me out to conferences, I was good at it so they made me the evangelist and that’s what we do, me and 3 others. we have a wiki that people can request us.
• FOr the last 5 years we have been opening up the infrastructure – the product catalog, alexa and now infrastructure services. putting APIs out there, charging models around them. we try and make it really simple to get started. Sign up and away you go
• we offer access in a number of ways. a lot of what we have had to do over the years has to do with scale, have the infrastructures.  we offer that out to developers.   The developers can focus on the innovative and creativity part..we do the hardware.
• we rely a lot of input and feedback from developers. let me know what you like and don’t like. we write a trip report everyday to get the feedback to the company.
• Cloud computing- emerging trend. you look to the cloud to do stuff for you. You can treat it as infinite capacity, scale on demand.  With cloud computing, what were fixed costs turn into variable costs, where you pay for what you need.  You get possibly better staff to keep it up, get economies of scale.
• It can reduce your time to get things up and running,
• We have 6 different services – Queue, storage, elastic compute cloud, flexible payments,
• S3: object based storage…1b-5GB. private or public, redundant and dispersed. storage. US and Eu locations. (Ireland). EU added due to latency and data legal issues. We dynamically manage the copies of the data.   We are redundant enough. 99.99% availability goal. Can organise in buckets/ each bucket is a flat storage model. Can use it as a bit torrent seed. Complete API around it.  We have libraries for a number of different languages, that we have built or other developers.
• Elastic Compute Cloud: (not the ZX81): use Xen, take lots of machines, slice into small components, using a indi OS. when you have an instance it is yours while you pay. you get root level, elastic capacity,. There’s a lot of apps built on top as well for you to use. you get to scale in minutes.  Up and down control. We have 3 different instance types – small, Large and XL.  You have an Amazon Machine Image – which is your’s.  You put this in S3, then roll it out to the servers you need as you scale.  Full API into the cloud, you can start machines with one call.  Used in lots of scientific research. MapReduce and Hadoop, for engineering and science calculations.   For Fortune 500 companies, often for high impact, short term projects, as a dev host.  One example is the NYT archive, When it was a closed service, they re-rendered PDF from TIFF every time. They decided to use EC2 with Hadoop to pre-render everything.  They tested it..then ran it over 24 hours over 100 instances. Far better than having to do internally.  Some one has built EC2 Firefox UI – a browser addon that allows you to control the instances

Tags: jeff barr, amazon, s3, ec2