Farewell EC2-Traditional, it’s been swell

Retiring providers isn’t one thing we do at AWS. It’s fairly uncommon. Corporations depend on our choices – their companies actually stay on these providers – and it’s one thing that we take critically. For instance SimpleDB remains to be round, though DynamoDB is the “NoSQL” DB of alternative for our clients.

So, two years in the past, when Jeff Barr announced that we’d be shutting down EC2-Classic, I’m positive that there have been a minimum of a few of you that didn’t consider we’d really flip the change — that we’d let it run endlessly. Nicely, that day has come. On August 15, 2023, we shut down the final occasion of Traditional. And with the entire historical past right here, I believe it’s price celebrating the unique model of one of many providers that began what we now know as cloud computing.

EC2 has been round for fairly some time, virtually 17 years. Solely SQS and S3 are older. So, I wouldn’t blame you if you happen to had been questioning what makes an EC2 occasion “Traditional”. Put merely, it’s the community structure. After we launched EC2 in 2006, it was one large community of 10.0.0.0/8. All cases ran on a single, flat community shared with different clients. It uncovered a handful of options, like safety teams and Public IP addresses that had been assigned when an occasion was spun up. Traditional made the method of buying compute lifeless easy, though the stack working behind the scenes was extremely advanced. “Invent and Simplify” is likely one of the Amazon Leadership Principles in spite of everything…

If you happen to had launched an occasion in 2006, an m1.small, you’d have gotten a digital CPU the equal of a 1.7 GHz Xeon processor with 1.75 GB of RAM, 160 GB of native disk, and 250 Mb/second of community bandwidth. And it could have value simply $0.10 per clocked hour. It’s fairly unimaginable the place cloud computing has gone since then, with a P3dn.24xlarge offering 100 Gbps of community throughput, 96 vCPUs, 8 NVIDIA v100 Tensor Core GPUs with 32 GiB of reminiscence every, 768 GiB of complete methods reminiscence, and 1.8 TB of native SSD storage, to not point out an EFA to speed up ML workloads.

However 2006 was a special time, and that flat community and small assortment of cases, just like the m1.small, was “Traditional”. And on the time it was really revolutionary. {Hardware} had grow to be a programmable useful resource that you could possibly scale up or down at a second’s discover. Each developer, entrepreneur, startup and enterprise, now had entry to as a lot compute as they needed, every time they needed it. The complexities of managing infrastructure, shopping for new {hardware}, upgrading software program, changing failed disks — had been abstracted away. And it modified the best way all of us designed and constructed functions.

After all the very first thing I did when EC2 was launched was to maneuver this weblog to an m1.small. It was working Moveable Type and the this occasion was adequate to run the server and the native (no RDS but) database. Ultimately I turned it right into a highly-available service with RDS failover, and so on., and it ran there for five+ years till the Amazon S3 Website feature was launched in 2011. The weblog has now been “serverless” for the previous 12 years.

Like we do with all of our providers, we listened to what our clients wanted subsequent. This led us to including options like Elastic IP addresses, Auto Scaling, Load Balancing, CloudWatch, and numerous new occasion sorts that might higher swimsuit totally different workloads. By 2013 we had enabled VPC, which allowed every AWS buyer to handle their very own slice of the cloud, safe, remoted, and outlined for his or her enterprise. And it grew to become the brand new commonplace. It merely gave clients a brand new degree of management that enabled them to construct much more complete methods within the cloud.

We continued to assist Traditional for the following decade, at the same time as EC2 developed and we applied a wholly new virtualization platform, Nitro — as a result of our clients had been utilizing it.

Ten years in the past, throughout my 2013 keynote at re:Invent, I instructed you that we needed to “assist right now’s workloads in addition to tomorrow’s,” and our dedication to Traditional is the most effective proof of that. It’s not misplaced on me, the quantity of labor that goes into an effort like this — however it’s precisely the sort of work that builds belief, and I’m happy with the best way it has been dealt with. To me, this embodies what it means to be buyer obsessed. The EC2 crew stored Traditional working (and working effectively) till each occasion was shut down or migrated. Offering documentation, instruments, and assist from engineering and account administration groups all through the method.

It’s bittersweet to say goodbye to certainly one of our unique choices. However we’ve come a great distance since 2006 and we’re not executed innovating for our clients. It’s a reminder that constructing evolvable methods is a technique, and revisiting your architectures with an open thoughts is a should. So, farewell Traditional, it’s been swell. Lengthy stay EC2.

Now, go construct!

Beneficial posts