This is an example of how the tools that are part of modern cloud computing, including automation tools, elasticity and dynamic pricing can come together to help address major scientific workloads.
A Cloud Guru is now offering a free one-hour course explaining the basics of what cloud computing is (and isn’t) and why you keep hearing about it. I recommend it to anyone who’s not sure where to start on this whole “cloud” thing.
Amazon has begun rolling out a new web console. We’re expecting existing accounts to be updated by the end of the year, but there’s no telling exactly when a given account will see the update.
A Cloud Guru has put together a video walking through the new console and highlighting some of the changes. Have a look, and hopefully it will be less of a surprise when the update hits your account.
We’ll be diving into Amazon Lambda with next week’s training, so it seems like a perfect time to highlight how event-driven function calls can allow extremely cost-effective web applications. One developer delivered 30,000 page views for less than a quarter and documented his experience.
I’d like to share an Amazon blog entry detailing Georgia Tech’s use of AWS to guarantee emergency communications.
In the event of an emergency, Georgia Tech’s communications team will direct clients to their emergency website, a quick-loading static site hosted on Amazon S3. By using Amazon’s services, Georgia Tech can be confident that they have the capacity to communicate to any size audience at any time.
S3’s standard tier stores a minimum of three copies of data across multiple physical locations within a region.
S3 is built to be scalable, spreading load across a huge number of systems. Adding thousands of additional requests per minute is inconsequential.
Since Amazon bills for actual usage, the main steady state cost is for the web content. The site the blog describes probably totals a few megabytes of storage. At $0.03 per gigabyte per month, that’s going to round up to a penny.
Even during a crisis, with a large number of clients repeatedly reloading the page, Georgia Tech estimates their costs will max out at $20-some dollars.
Instant updates via a simple API
Content is managed through a simple web application running on EC2 instances. When new information is available, the communications team can log into that application, update the content, and publish a new page to S3. The updated content is immediately available to clients.
I’d initially thought that CloudFront might make a good addition to this design, since it speeds up content delivery and lowers costs even further. But CloudFront requires 10-15 minutes to process a cache invalidation request, which could delay critical communication in an emergency situation.
We ran two AWS 101 labs today, with 19 total participants. The lab guided users through launching two web servers with SSD storage and using an Elastic Load Balancer to direct HTTP traffic to the highly-available cluster.
I checked our bill after the labs concluded:
$4.01 comes out to just over twenty cents per participant. I feel like that’s not too bad.
Amazon Web Services is now available at Illinois for research, instruction, and administrative use. You can request access or log into an account at the University of Illinois AWS service gateway: http://aws.illinois.edu/
Amazon will be on campus regularly to provide guided labs and consultation. More information is available in our Spring 2017 Lab Schedule.
You can contact the Illinois AWS team at firstname.lastname@example.org. We’re always available to sit down and discuss whether AWS is a good fit for your unit’s needs, design solutions, or take feedback on how we can improve the service.
Amazon Web Services turns ten years old today. Werner Vogels, Amazon’s CTO, has shared a retrospective of the lessons Amazon has learned in that time:
These ten lessons learned are surprisingly relevant today as we develop plans to make AWS available for campus use and consider the opportunities we have to change our practices and make excellent use of new tools.
We considered setting this site up in AWS. It would have been an interesting exercise that would have produced a useful resource, and give us some practical experience doing something in AWS. But as we thought about it, we realized that publish.illinois.edu (PIE) is already an established and solid service, we’d get the job done faster with PIE, and we wouldn’t have to build, run, update, manage and pay for the EC2 instances. AWS is a great tool, but it’s not necessarily the best solution for everything. We needed a web site, not IaaS, so we chose PIE.
This is an interesting article about building an application without just spinning up EC2 instances. Developing applications this way will likely be more reliable and scale better, as well as cost a lot less than running an EC2 instance 24×7.