AWS Case Study: photoWALL
With the popularity of social media, real-time web applications, and mobile devices, photoWALL is a service that aims to tie together all three. photoWALL is a new service developed by AirMe, a Colorado-based company that develops photo sharing applications for the iPhone and other mobile devices. photoWALL allows users to share photos on the photoWALL site as well as on Facebook, Twitter, and Flickr. With one click from a mobile device, the photo will show up on all four sites in seconds, creating a true real-time media stream.
After showing Microsoft an initial prototype of the service, the Microsoft team was enthusiastic and wanted to launch the service as quickly as possible. The AirMe team knew they needed an infrastructure solution that they could deploy quickly and had massive scalability.
AirMe decided to use Amazon Web Services (AWS) and today, the photoWALL web site and server side computing run entirely on AWS. Phil Easter, CTO and Founder of AirMe, recalls, “Amazon Web Services allowed us to create a dynamic, highly scalable ecosystem that would be difficult to duplicate in a traditional server farm, co-location premise. AWS has unique technologies like SQS and EBS that allowed us to architect a new mobile media mash up platform. Add in, auto scaling and load balancing with a CDN…this is not your father’s data center!”
photoWALL uses the entire suite of AWS infrastructure services: Amazon EC2 (with Auto Scaling, Elastic Load Balancing, CloudWatch, and Elastic Block Store), Amazon S3, Amazon CloudFront, Amazon SimpleDB, and Amazon SQS (see Architecture Diagram 9/17/09).
Architecture Diagram 9/17/09
Easter explains their architecture on AWS and how the application works. The application captures and sends a photo, along with 40 different data points of the photo, such as associated media and metadata, directly to their computing resources on AWS. These resources include a large EC2 cluster of 25+ instances hosting their API servers, photo processing servers, and PHP application. The EC2 cluster is fully optimized with Elastic Load Balancing, Auto-scaling, and CloudWatch for seamless scalability. Amazon SQS manages their workflow; once a photo is validated, it is sent for processing, where it undergoes file manipulations (resizing, geo-tagging, indexing, etc.). Amazon SQS is also used for queuing photos and sending to supported social media sites. Metadata information about the photo is stored primarily in MySQL databases (queries enabled with memcached) which are replicated for redundancy. Some metadata is also stored in SimpleDB. Once all the processing is complete, the photo is pushed in Amazon S3 for reliable storage, and then delivered via CloudFront, the global content delivery service, to end users. CloudFront also serves up all CSS and JavaScript.
Shortly after the release of Amazon Relational Database Service (RDS), the AirMe team realized the huge cost and management savings if they converted their MySQL servers. Easter says, “The easy instance and storage upgrades made the decision a no-brainer. The cut-over from one write and two read servers to one RDS instance required four simple configuration changes taking approximately 5 minutes.” After the change, their query times improved by 100%. See Architecture Diagram 10/27/09 for photoWALL’s revised architecture with Amazon RDS.
Architecture Diagram 10/27/09
Easter summarizes, “We chose AWS for the low cost but what we like best about AWS is the ability to instantly turn on new servers with no need to plan. AWS allows us to focus on the business and user experience, not needing to worry about adding new racks, buying more servers, figuring out how to get more bandwidth, etc. Without AWS we would not exist – not due to lack of knowhow, but due to time to market!”
To see photoWALL in action, visit http://www.photowall.com 
or learn more about AirMe at http://www.airme.com .
©2013, Amazon Web Services, Inc. or its affiliates. All rights reserved.
