Some interesting developments in grid computing.... ----- Forwarded by Ranga Nathan/AMERICA/BAX on 09/22/2006 04:29 PM ----- *Tom Metro <[EMAIL PROTECTED]>*
Sent by: [EMAIL PROTECTED] 09/22/2006 03:45 PM To L-boston-pm <boston-pm@mail.pm.org> cc Subject [Boston.pm] Jeff Barr's talk: Amazon's web services I wrote up my notes on Jeff Barr's talk to Boston.pm on 9/13/2006 for some colleagues, and figured I might as well post a copy here for anyone who wasn't able to attend. Here are some of the highlights: The Elastic Compute Cloud http://www.amazon.com/b/ref=sc_fe_c_1_3435361_1/104-6093751-9591919?ie=UTF8&node=201590011&no=3435361&me=A36L942TSJ2AJA A general purpose "grid" or cluster of machines. It uses Xen virtualized machines and they'll rent you the equivalent of a 1.7 GHz Xenon CPU w/1.7 GB of RAM, 160 GB HD, and 256 Mbps network (probably 1/4 of a blade) for 10 cents/hr. You provide your application as a Xen image, and load it up onto as many machines as you need. This can be used for a variety of tasks - from compute intensive operations, to bulk data processing, to web hosting. One example the speaker mentioned was a telephone company using it to process end-of-the-month billing, where the compute capacity is only needed for a few days out of the month. The Elastic Compute Cloud is only 4 weeks old (and still in limited beta), but it potentially will have a big impact on Internet startups. Now it is no longer necessary to pay for substantial infrastructure up front. You just expand to more machines almost transparently as your business grows. (Though you could always rent machines, the ability to expand and contract capacity wasn't as flexible.) The speaker told a story of a friend who declined a VC investment because this service allowed him to avoid the up-front expenses. Alexa web services http://www.amazon.com/b/ref=sc_fe_l_2/104-6093751-9591919?ie=UTF8&node=239513011&no=3435361&me=A36L942TSJ2AJA It's basically the database and processing infrastructure of a search engine that you rent, which you can then layer your specific search engine design on top of. Their database includes a bunch of meta data on the spidered sites, including a "web map," which is the linking relationships between sites. They also have a thumbnail service that returns an image for a specific site. And they have a custom processing cluster that can execute custom data processing code (at a rate of $1/CPU/hour; the speaker speculated the price may drop now that the ECC service is priced so much lower). Amazon's Simple Storage Service (S3) http://www.amazon.com/S3-AWS-home-page-Money/b/ref=sc_fe_l_2/104-6093751-9591919?ie=UTF8&node=16427261&no=3435361&me=A36L942TSJ2AJA Amazon's answer to online storage. This has been available for 6 months or so, and already several startups (such as photo sharing services and PC backup services) are using it as their back-end storage infrastructure. It's been written about a lot, so I'll just mention that the rates are 15 cents per GB per month, and 20 cents per GB for transfers. It was also mentioned that S3 is on the same internal network as The Elastic Compute Cloud, and that there are no transfer costs between the two. Some related links: JungleDisk (open standard/closed source/multi-platform, runs as a WebDAV server, $0.15/GB) http://jungledisk.com/ filicio.us (web-based file storage and sharing front-end for S3) http://www.filicio.us/ Amazon Web Services Developer Connection: s3sync -- a simple rsync look-alike http://developer.amazonwebservices.com/connect/thread.jspa?messageID=37272&tstart=0#37272 S3Fox Organizer (Firefox extension) http://www.rjonna.com/ext/s3fox.php s3/fuse (Linux file system for S3) http://dev.extensibleforge.net/wiki/s3/fuse Windows-only commercial backup services using S3: http://elephantdrive.com/ http://altexa.com/ Amazon Mechanical Turk http://www.amazon.com/Mechanical-Turk-AWS-home-page/b/ref=sc_fe_l_2/104-6093751-9591919?ie=UTF8&node=15879911&no=3435361&me=A36L942TSJ2AJA http://www.mturk.com/ Described as "Artificial Artificial Intelligence" or "people as a service." This service provides the infrastructure for automating the process of assigning small tasks to people - tasks that are too difficult or impossible for computers to do. An example of a task is image processing: does this image contain a picture of a human? This task can be given to thousands of people, with multiple people processing the same images to improve accuracy. Another example is transcribing podcasts. It can also be used for marketing surveys. (Who would have thought Amazon would end up dominating the "get paid to take surveys" market.) One entertaining example use was for an art project, where people were paid to draw sheep (http://www.thesheepmarket.com/). HIT-Builder for Amazon's Mechanical Turk (a tool for creating tasks) http://www.hit-builder.com/ CastingWords (podcast transcription services that uses AMT) http://castingwords.com/ Amazon's E-Commerce Service http://www.amazon.com/E-Commerce-Service-AWS-home-page/b/ref=sc_fe_l_2/104-6093751-9591919?ie=UTF8&node=12738641&no=3435361&me=A36L942TSJ2AJA This was their first web service offering. It basically lets you create your own Amazon clone, if that's what you want to do, or build any kind of a supplementary service that depends on Amazon's product catalog. The speaker demoed a couple of sites built on top of ECS that included some interesting visualization technology. One, hivegroup.com, lets you select products from the Amazon catalog based on parameters, while in real time updating a graphical grid (Java applet) representing the matching products. Another, liveplasma.com, was a music and movie "discovery engine." You supply the name of a music group, and it produces a slick connected graph representation of related groups, with each node linking to a product page on Amazon. Also briefly mentioned were: Amazon Historical Pricing http://www.amazon.com/Historical-Pricing-AWS-home-page/b/ref=sc_fe_l_2/104-6093751-9591919?ie=UTF8&node=15811391&no=3435361&me=A36L942TSJ2AJA Amazon Simple Queue Service http://www.amazon.com/Simple-Queue-Service-home-page/b/ref=sc_fe_l_2/104-6093751-9591919?ie=UTF8&node=13584001&no=3435361&me=A36L942TSJ2AJA And general links: http://aws.amazon.com/ The main site for Amazon's Web Services. http://www.awszone.com/ A web front-end for interacting with Amazon's services. http://aws.typepad.com/ Jeff Barr's blog jbarr (at) amazon.com Jeff Barr's email -Tom -- Tom Metro Venture Logic, Newton, MA, USA "Enterprise solutions through open source." Professional Profile: http://tmetro.venturelogic.com/ _______________________________________________ Boston-pm mailing list Boston-pm@mail.pm.org http://mail.pm.org/mailman/listinfo/boston-pm -- __________________ Ranga Nathan Work: 714-442-7591