AWS re:Invent 2018 took Las Vegas by storm last week, and the Trek10 team was out in force, keeping up with all the new announcements from AWS and our other partners. (We even got a shoutout in Werner Vogels’ keynote!)
You’ll find lots of posts this week summing up big themes and high-level takeaways from re:Invent. We’ve certainly got our share of opinions, but at Trek10 we care primarily about what these announcements allow us to build.
While many re:Invent announcements were featured in keynotes and got lots of media attention, some of the most game-changing feature upgrades slipped out quietly in the days leading up to the big event. But never fear: Trek10 keeps a close eye on every AWS announcement (we even have a Slack bot that routinely diffs the publicly-available CloudFormation spec — we’ve caught some features that way before they’ve even been officially released!)
I polled some of Trek10’s engineers about the re:Invent announcements that excite them most, and asked them to provide real-world use cases showing these new features in action. So without further ado, here are the new AWS capabilities and services that Trek10’s serverless experts are building on:
- Amazon Textract (currently in limited preview)
We have a few Clients running fragile, bespoke OCR processes on EC2 instances or in database stored procedures. This code can run to thousands of lines and is difficult to migrate or maintain. Textract, a service that can automatically pull text from scanned documents such as medical forms, should help these customers remove technical debt while empowering the team to reduce dependence on legacy architectures.
For some of our large serverless architectures, the Amazon API Gateway costs are even more expensive than the underlying Lambda functions that power the API. Placing Lambda functions behind an ALB, rather than API Gateway, is a whole new way of building a serverless architecture that should be dramatically more cost-effective.
One of our Clients has experienced DynamoDB read/write spikes that were too fast for auto scaling to handle, but the costs to provision for the maximum size are too costly. Now that DynamoDB is a true serverless datastore, the Client will incur a higher base cost, but a far lower overall expense than by explicitly provisioning capacity. And, no need to worry about handling spikes again!
— Joel Haubold
I’m working with a client that has had limited ability to migrate to a serverless architecture: the codebase has a heavy Ruby focus, the existing app is behind an ALB, and they use a relational database on RDS. Now I have a very elegant set of features to start slowly migrating a Ruby on Rails application to serverless bit by bit.
— Sean Summers, Sr. Cloud Engineer
- Amazon Timestream (currently in limited preview)
One of our large IoT clients needs to coalesce multiple read buckets over time from IoT devices, joining the most recent known values of different ages. Instead of updating wide keys in DynamoDB or reading, updating, and rewriting s3 objects, we can now use Amazon’s managed time series database to stay on top of this fast-accumulating data.
— Sean Summers
Today, one of our Clients with bursty workloads uses a complex system with multiple autoscaling groups to direct traffic to on-demand or spot EC2 instances for the most cost-effective performance. By combining spot and on-demand instances in the same ASG, that whole system can go away. Talk about an undifferentiated heavy load off our minds!
— Joel Haubold
While we love serverless at Trek10, we still spend our share of time with those pesky things called containers. Blue/green deployment abilities for AWS’s native container services give us a whole new set of tools for rolling out changes to Fargate-based APIs.
- AWS Security Hub (currently in preview)
As part of our Landing Zone deployment, we typically implement a “Security Services” AWS account. Deploying the Security Hub in this account will allow us to aggregate incidents and data across numerous AWS services. Additionally, we are heavy users of Alert Logic and are excited to dive into the Alert Logic / Security Hub integrations.
Nothing sexy to say about this one! We have Clients who run SFTP servers on EC2. We would much rather decommission those servers and remove the operational burden of managing those VMs (not to mention the out-of-the-box HA and addition of all the features of S3 like lifecycle policies, storage classes, Athena, etc.).
— Josh von Schaumburg
As for me, I summed up my reaction to some of the new serverless features in this interview with AWS’s Randall Hunt.
If you’re overwhelmed by all the great content put out during this year’s re:Invent, our own Ryan Scott Brown has made a handy searchable index of session videos. If you’d like to chat more about how AWS re:Invent 2018 affects your cloud strategy, feel free to reach out @Trek10Inc, and happy building!