From the beginning we've worked hard to ensure that building and maintaining applications on Google App Engine is as easy as possible and that those applications can be made to scale very well. Scalability can mean different things to different people though. For us, platform scalability (today) means efficiently serving 250 million daily page views for 250,000 developers. For you, application scalability means being able to serve a large number of requests to your web site when you need it.

From the beginning we've worked hard to ensure that building and maintaining applications on Google App Engine is as easy as possible and that those applications can be made to scale very well. Scalability can mean different things to different people though. For us, platform scalability (today) means efficiently serving 250 million daily page views for 250,000 developers. For you, application scalability means being able to serve a large number of requests to your web site when you need it.






But serving all of that traffic is just half the challenge. Even for applications with very predictable traffic patterns it's often impossible to know how much traffic your site will see at any given time in the future. You need flexibility and a platform which will automatically adjust to your needs, minute by minute. This is especially true for sites that are used in conjunction with large media events. We've been working with one such company and we'd like to share their scalability story with you today.






Gigya’s technology enables their customers to connect seamlessly to popular social network platforms for user authentication, sharing and engagement. One product feature is a social live chat widget used by their customers to accompany live-streamed events. When I spoke with Gigya's Raviv Pavel, VP Research and Development, he told me, "Although we typically host all our services in-house, on our own infrastructure, we felt that GAE would be a better fit for the live chat feature because of its unique traffic pattern, which is characterized by very low traffic most of the time with very high bursts during high profile events."






Below is a graph from the January 22nd Hope for Haiti Now: A Global Benefit for Earthquake Relief during which Gigya had deployed their social gadget to multiple web properties. The vertical scale shows number of requests or queries per second, which peaked at around 1600 QPS. Each of their events has its own unique characteristics. So far we've seen Bactrian camels, Dromedary camels, parts of the Rocky Mountains and this next graph which vaguely resembles a submarine poking through the polar icecap.













Gigya wins because Google App Engine maintains the infrastructure needed to support those traffic bursts for the live chat widget. And because of App Engine's granular billing model, they only pay for the resources they use. When Gigya's live chat sees reduced traffic, we make sure our machines are busy serving other requests.






If you have an interesting App Engine story you'd like to share, let us know. We'd like to hear about it.






Posted by Fred Sauer, App Engine team

We are excited to announce the release of version 1.3.1 of the App Engine SDK for both Python and Java. While this release contains plenty of new features and fixes, we've concentrated on using our very first SDK release of 2010 to improve the heart of many App Engine applications: the Datastore.
We are excited to announce the release of version 1.3.1 of the App Engine SDK for both Python and Java. While this release contains plenty of new features and fixes, we've concentrated on using our very first SDK release of 2010 to improve the heart of many App Engine applications: the Datastore.



Here are the three major improvements that 1.3.1 has in store for datastore users:
  • Datastore Query Cursors - Cursors allow applications to save and 'bookmark' their progress through a query, so that it can be resumed later. This works great in combination with paging URLs, as well as processing in the Task Queue API, but there are many other uses. Watch for an upcoming blog post that explores Cursors in the near future. They're also really handy in the context of the next change...
  • No more 1000 result limit - That's right: with addition of Cursors and the culmination of many smaller Datastore stability and performance improvements over the last few months, we're now confident enough to remove the maximum result limit altogether. Whether you're doing a fetch, iterating, or using a Cursor, there's no limits on the number of results.
  • Reduced error rate with Automatic Datastore Retries - We've heard a lot of feedback that you don't want to deal with the Datastore's sporadic errors. In response, App Engine now automatically retries all datastore calls (with the exception of transaction commits) when your applications encounters a datastore error caused by being unable to reach Bigtable. Datastore retries automatically builds in what many of you have been doing in your code already, and our tests have shown it drastically reduces the number of errors your application experiences (by up to 3-4x error reduction for puts, 10-30x for gets).
But even with our focus on the Datastore, we were able to sneak in a quite a number of other new goodies into 1.3.1 as well:



For Python, we've included the AppStats RPC instrumentation library. AppStats lets users easily profile the performance of calls from their app to the App Engine backend services to identify and isolate issues such as ineffective caching, bottlenecks, and redundant RPC calls in their app. (A Java version is in beta testing now.)



And for Java, we've included a comprehensive new unit-testing framework for your App Engine apps. The unit-testing framework enables you to test your application code in a natural, fully supported manner, and also allows you to integrate your App Engine apps into other existing testing and automation frameworks.



The list of changes in 1.3.1 goes on and on (Custom Admin Console pages! Support for wildcard domain mappings! Java precompilation on by default for all applications!), so make sure to download the new version and read our release notes for the complete list (Python, Java).



Posted by the App Engine Team