At AWS re:Invent in Vegas on December 3rd, Andy Jassy, CEO of AWS announced the general availability of AWS Outposts. Given the foreshadowing of Outposts a year ago, this was expected. By itself, Outposts, similar in some sense to Azure Stack Hub and Azure Arc or Google’s Anthos, promises to alter the way that enterprise customers think about their options for the hybrid cloud by providing an “outpost” of AWS on-premise that provides many of the exact same services as AWS EC2, etc. This type of option simplifies design decisions where latency and concerns on data locality or other governance concerns make it more difficult for an enterprise to consider putting certain workloads in the cloud.
Very importantly, and distinctively, due to the hardware model in which AWS delivers Outposts, in the coming years, we are likely to see the beginning of a potentially very disruptive inflection point in the split between hybrid and private data centers. This inflection point may have major implications in how enterprise customers that work with, or are considering working with, AWS implement their private cloud data centers going forward. Let’s take a look.
Let’s start with the introduction by AWS of Graviton2. AWS introduced a family of three general compute instances, a follow on to last year’s Graviton-1, that is built upon custom ARM silicon designed by AWS’ Annapurna Lab’s team. What is notable about Graviton2 is its pricing – which will be 40% less than the equivalent general purpose EC2 instances that run on the very latest standard Intel X86 cores.
“Virtually all your workloads [will run on Graviton2], and that is a huge game-changer for you all” – Andy Jassy at the 21.30min mark of his keynote.
For users, this option is great news. This represents a dramatic reduction in workload costs. Three new instances, called the M6g, R6g, C6g, support 25Gbps networking, are 64bit and are built on chips using leading-edge 7nm lithography. Moreover, compared to the gen1 Graviton, there are 4X the compute cores with 5X faster memory and 7X the performance – and again, Graviton2 is 40% lower cost than the latest X86 processors.
At first, this may seem like a simple cost and efficiency play by AWS and one that provides competitive differentiation versus GCP or Azure. However, the ramifications to corporate hybrid cloud plans, given the dramatic price discontinuity, may be even larger.
As CIOs and their teams look at their plans for 2020 and beyond, Graviton2 should clearly become part of their thinking for their current or planned AWS workloads. It is likely that for many workloads where internal IT teams used creative math and demonstrated that using on-premise, internally managed to compute and storage for some set of applications was lower cost than the cloud, the Graviton2 announcement has now blown a hole in their story. For many of those companies where the cloud vs. on-prem business case was in the gray zone, moving to AWS with Graviton2 now makes a stronger economic sense. Moreover, it begins to simplify a company’s hybrid cloud strategy.
But there is more.
AWS announced the general availability of Outposts, and the availability in early 2020 of Outposts that runs VMware. Therefore, 2020 should see the beginning of a ramp in the utilization of Outposts in the enterprise, with 2021 likely to be a breakout year. Fast forward a year or so, and you may see AWS Outposts offerings that have been enhanced with the new Graviton2 instances. Should AWS choose to make Graviton2 available in Outposts, this would add even more disruption to the enterprise hybrid data center architecture and cost equation. What would make this particularly disruptive is the potential for an on-premise hybrid cloud that is, in fact, lower cost than any X86 servers the enterprise can otherwise purchase – because Amazon is the only manufacturer of this capability. The internal enterprise CFO-CIO debate about the cloud, hybrid cloud, on-premise vs. cloud changes complexion when the overall cost of the on-premise cloud hardware itself has the potential of being lower cost than standard X86 servers an enterprise would normally use.
AWS also announced Local Zones at re:Invent 2019. Essentially a Local Zone is a large AWS run Outposts to be located in major cities to address the need by some customers and industries for low latency cloud services. LA is the first zone, by invitation only, clearly being established for media companies. This matters because it illustrates that the AWS hardware approach of packaging, essentially a 98% copy of their own cloud resources to build Outposts, enables them to easily leverage their central Cloud hardware investments for this new service. Therefore, much like Local Zones was built from the core technology that powers the AWS cloud with only minimal differences, so to can AWS easily bring this same model of leveraging core technology to Outputs for the Enterprise, but in the future include Graviton2 as an option. I would suspect they’d consider doing this once Outposts has been shipping for a sufficient time and has been gathering momentum in the marketplace.
In simple terms, Outposts, Graviton2, and Local Zones (among others), driven by AWS as the dominant cloud provider, will accelerate the already well-established move from on-premise to cloud and the move from homegrown private cloud, to cloud partner private cloud – be it AWS Outposts or GCP Anthos or Azure Stack Hub. Moreover, given how AWS has set the table here, we can expect that AWS in the short term will influence enterprise customers to reconsider their hybrid cloud plans and the balance of workloads between on-premise and the cloud with the introduction of Graviton2, the general availability of Outposts, as well as new services such as Local Zones.
Longer-term, as Outposts eventually acquires Graviton2 (or 3, etc.) technology, the disruption of the enterprise data center and hybrid data center becomes even greater; potentially hollowing out some of the remaining portions of on-premise solutions. Finally, as this trend begins to take hold, it is reasonable to expect that current suppliers of major components, both hardware and software, to enterprise data centers will be impacted as the cloud descends and swallows an ever-increasing portion of on-premise compute and storage.