Matthew Meckes
Amazon Web Services
Matthew is a Serverless Specialist at AWS and leads the serverless business for AWS in the UK and Ireland. Matthew helps customers architect and execute the most challenging serverless solutions in the region, and works with customers to identify use cases in order to drive the serverless product roadmap.
You want to scale your Java applications to zero, but starting from zero is hard. You have to provision compute, load a JVM, initialise and start your application - often in less than a second.
We will take a deep dive into how AWS uses VM snapshotting at scale. We will discuss why this method is relevant to the JVM and particularly Java applications that use runtime dependency injection such as Spring Boot. We will also discuss some of the solutions we use to deliver a highly performant, production ready, and cloud scale snapshotting architecture, including the invoke routing layer, compute fabric, how we mange snapshot distribution, caching strategies, async chunk loading and chunk load order optimisation.
SnapStart for AWS Lambda solves cold start problems for Java applications. It lets customers run larger, more complex applications on Lambda, and scale up from zero to thousands of execution environments in just a few seconds. We’ll share what we’ve learned from building the feature, and running it in GA for the last 18 months.
More We will take a deep dive into how AWS uses VM snapshotting at scale. We will discuss why this method is relevant to the JVM and particularly Java applications that use runtime dependency injection such as Spring Boot. We will also discuss some of the solutions we use to deliver a highly performant, production ready, and cloud scale snapshotting architecture, including the invoke routing layer, compute fabric, how we mange snapshot distribution, caching strategies, async chunk loading and chunk load order optimisation.
SnapStart for AWS Lambda solves cold start problems for Java applications. It lets customers run larger, more complex applications on Lambda, and scale up from zero to thousands of execution environments in just a few seconds. We’ll share what we’ve learned from building the feature, and running it in GA for the last 18 months.