In a major departure from its usual cloud-only stance, Amazon Web Services has announced it’s now possible to run EC2 instances with on-premises hardware – but only its own Snowball Edge devices.
The Snowball Edge is sold as a data transfer appliance that improves on the base Snowball by adding the ability to use it as local storage in hard-to-connect locations and to run Lambda and Greengrass functions on the edge to do some basic data processing.
Now AWS has made it possible to run full EC2 instances on the devices.
Amazon’s announcement explains that the Snowball Edge contains a Xeon D CPU running at 1.8 GHz, and can run instances that consume a total of 24 vCPUs and 32 GiB of RAM.
A new “sbe” instance type runs on the devices and comes in six sizes, from the “small” size at one vCPU and a single gig of memory up to the “sbe1.4xlarge” that offers 16 vCPUs and consumes all the RAM inside a Snowball Edge.
Amazon reckons you’ll create images in its cloud, give them a thrashing to make sure they work and then drop them onto a Snowball Edge.
AWS seems to be targeting two markets here.
One is the classic play of doing some data processing data on the edge rather than sending everything to a cloud or other data centre. That scenario is agreed as a sensible use for edge devices as it reduces the use of expensive WANs and avoids latency hassles for time-sensitive jobs.
The second is remotely-managed PCs and servers. AWS’ Jeff Barr wrote that the instances will meet the needs of an IT manager he recently met who “told me that he wanted to be able to set up and centrally manage the global collection of on-premises industrialized PCs that monitor their machinery as easily and as efficiently as he does their EC2 instances and other cloud resources.”
By making instances on a Snowball Edge just another EC2 instance to manage, AWS pretty much delivers on that scenario.
It also delivers a kind of inside-out hybrid cloud. Today most discussions of hybrid clouds expect that on-prem workloads will stretch into the cloud. AWS’ plan here is for workloads cooked in the cloud to be pushed out onto edge devices.
Amazon’s hybrid and edge products have never been strong and it’s never shown much interesting in making them better. Now it’s showing us that it has a different take on hybrid cloud.
It won’t have this field to itself, because all the big server-makers have conventional boxes ready to splash down at the edge and the likes of VMware, OpenStack and Windows Server aren’t shabby at shunting workloads around to edge servers.
But none also operate the world’s premier cloud and offer the kind of heterogeneous experience AWS is aiming for.
The big question now is whether this is all AWS has got in the hybrid cloud or if there’s more to come. We’ve already had a hint of AWS on-prem switches. So perhaps the cloudy colossus has decided it’s time to come for more frequent visits.
In other AWS news, it today previewed three new instance types that will run on its Nitro architecture and accompanying hypervisor. The three instance types are:
- Z1d – for compute-intensive workloads, They’ll run at up to 4.0 GHz, “powered by sustained all-core Turbo Boost”.
- R5 memory-optimized instances running at up to 3.1 GHz, also in all-Turbo mode and “with up to 50% more vCPUs and 60% more memory than R4 instances.”
- R5d instances just like the R5 but with local NVMe storage up to 3.6 TB
The preview is notable in and of itself, as AWS seldom teases new stuff. It instead just announces it exists. The cloud caper is changing, fast. ®