The best evangelist in the biz, Jeff Barr, recently announced the latest EC2 enhancement – Auto scaling groups with multiple instance types. “How do you guys feel about that announcement?”, a customer asked me over email. The answer is – we’re excited. The latest move from AWS is a huge validation for the things we’ve built over the last few years here at Spotinst.
It is great to work side by side with AWS, a cloud provider that truly wants to help customers optimize their costs and leverage its idle compute capacity — EC2 Spot Instances. With the recent announcement, AWS is making Spot more accessible and versatile than ever.
It’s becoming clear that AWS is embracing Spot Instances and pushing them hard. Now, Spot Instances are extremely easy to get started with. Without any bidding system or pricing models to learn, the barrier to running workloads on Spot is lower than ever.
So What’s Next? We’ve got your back.
In order to fully leverage EC2 Spot in your production environments, we’ve continued to focus on our customers and make sure they’ll be covered from all angles, including —
Prediction & Fallback to On Demand
Maintaining mission-critical application’s SLA with fallback to On-Demand Instances when Spot Capacity isn’t available
Elastigroup guarantees application uptime by shifting Instances from Spot to On Demand (or RIs when appropriate) ahead of time, to be precise, up to 15 minutes in advance of interruption(!). This allows you to run reliably your production services on Spot Instances. Whether it’s several dozens, or several hundreds of servers running fully on Spot Instances, Elastigroup makes sure that you have a solid plan for SLA when Spot capacity isn’t available.
Maneuver between On-Demand, Spot & RI to self-optimize workloads.
One of the things we found to be pretty complex when we built Elastigroup, is why, when and exactly how to switch instances back from On-Demand to Spot Instances or RIs that were in use but no longer available. This may cause downtimes if not handled properly and with great care.
Preferred Instance Types
Defining preferred Instance Types is key when working with heterogeneous groups. Often, a specific Instance Type is a great fit for a workload and should be favored when possible. Elastigroup lets you pick a preferred instance type and a list of ‘secondary’ instance types. Elastigroup will compose your cluster from preferred instances only, however, when your preferred one isn’t available as EC2 Spot, Elastigroup will launch Spot Instances from the ‘secondary’ list. These instances will, of course, auto-replace themselves once the preferred instance is available again.
Application & Containers Awareness
One of the major points of differentiation Elastigroup has over any other scaling tool is its application awareness. Elastigroup knows what type of application runs on top of the cluster, from basic Web Services behind Load Balancer to complex heterogenous Kubernetes Environments with thousands of Containers and Deployments. In the event of an instance interruption, Elastigroup not only scales up ahead of time, it also takes the necessary actions to drain Instances properly from the cluster (close HTTP Connections / Safely evict Containers).
Yes, There are still applications that need their OS/Logs/Application data. There are still applications that can only live on a single VM and not as part of a Horizontal scaling cluster. Elastigroup has a unique capability called “Stateful Spot Instance” which lets application maintain their state across interruptions, including the instance’s Private & Public IPs, Root Volumes, Data Volumes and even its Memory*.
Elastigroup supports your workloads where they are. Spotinst offers the same auto-scaling, cost-reducing, instance-type-mixing goodness on Google Cloud Platform and Microsoft Azure. Elastigroup supports Azure Scale Sets with Low-Priority VMs, and GCP Instance Groups with Preemptible VMs to ensure maximum availability at minimum cost across your multi-cloud infrastructure.
Elastigroup is the #1 cluster service on earth.
Simplify Deployments, Reduce Costs and Gain Deep Visibility
We’ll continue to focus on our customers and deliver the best of breed cluster service for every workload and across all cloud providers.