Sure, AWS is now embracing hybrid computing, but with a nod to the future, not the past. Here's how they're using it as an on-ramp for public cloud.
Some people are starting to suggest that Amazon Web Services (AWS) is finally having to acknowledge the reality of hybrid computing. Those people are wrong.
At AWS re:Invent, it's true that AWS announced a slew of products or programs (like Greengrass) to enable local-plus-cloud processing of data, but anyone reading this as anything other than an on-ramp to a public cloud future simply aren't paying attention.
The hybrid on-ramp
Ovum analyst Tony Baer asked, "The question for Amazon is how far it will be willing to grow its on premises footprint to accommodate this [hybrid] reality," but I think he's got the emphasis wrong. Despite Oracle and its ilk still owning the majority of enterprise workloads, the industry direction is clear, as Baer earlier noted: "[B]oth [Oracle and AWS] predict that the bulk of on-premise servers will disappear by the middle of the next decade."
Though Baer is correct to surmise that legacy servers will keep breathing in data centers, I don't believe AWS will do much to fight for yesterday's war. Already, Gartner analyst Thomas Bittman has uncovered a massive shift in where workloads are moving: According to Bittman, public cloud VMs have grown 20X over the years while private cloud growth putters along at 3X growth.
Why? Because "New stuff tends to go to the public cloud, while doing old stuff in new ways tends to go to private clouds. And new stuff is simply growing faster."
In other words, I see the AWS nod to hybrid workloads as more a matter of collecting modern workloads rather than trying to rescue legacy workloads. It's also a way, as AWS product strategy chief Matt Wood told me in an interview during AWS re:Invent, to "skate to where the puck is heading."
Hence, AWS Greengrass is a way to ensure that IoT workloads can thrive in the AWS cloud, despite latency and bandwidth constraints. In discussing this with Wood, he pointed out IoT is the obvious next level of infrastructure that needs to run in the cloud. With minimal compute power in sensors and other IoT devices, the cloud offers a great way to bring to bear massive compute power without having to run it on limited devices.
However, some applications can't successfully manage round-tripping data to the cloud, due to latency and other issues, so AWS is introducing a hybrid way of managing it: Run necessary processes locally on the device. Is it hybrid? Sure. But it's hybrid with a purpose: Move as much data as possible into the public cloud where, as Wood assured me, "In the fulness of time we continue to believe that most companies will not run their own data centers. The customers at scale are trying to migrate as many workloads as they can onto AWS with the stated aim of getting out of the data center business."
Greengrass is just a path to getting into the cloud, not to necessarily accommodating risk-averse CIOs who want to park data in their data centers indefinitely.
Where the puck is heading
Coming back to Baer's question, the momentum is not for private data centers, whether we dress them up as "private clouds" or simply call them what they are: Poorly secured silos. Even companies that want to believe in a private cloud future, as UBS analyst Steven Milunovich concluded, are coming to a "general consensus...that private cloud implementations generally are not working," further adding that "many companies that begin on a private cloud path end up going down a public cloud path."
SEE: AWS to Oracle: Now it's our turn and we got next (ZDNet)
This is why AWS cloud chief Andy Jassy was so comfortable slamming Oracle in his keynote, introducing a PostgreSQL-backed Aurora cloud database service to encourage even more companies to migrate. Based on Indeed.com jobs data, the market is listening:
So, does AWS care about hybrid computing? Sure, if it helps to move more workloads into the public cloud. Just don't for a minute think that AWS has changed from its original belief that a "model that requires companies to invest heavily in virtualization and maintaining their own data centers is not a 'real cloud' model," as Jassy declared years ago.
- Hybrid cloud isn't cloudy enough, suggests new Forrester data (TechRepublic)
- AWS to Oracle: Now it's our turn and we got next (ZDNet)
- Public cloud crushing private cloud in growth and revenue (TechRepublic)
- Are Amazon and Google the hybrid cloud advocates they claim to be? (ZDNet)
- Report: Big banks to move 30% of workloads to public cloud within three years (TechRepublic)