More details have come out on AWS Outposts, Amazon’s on-premises cloud solution coming out fall 2019. Pricing is still not available, but I see two reasons AWS Outposts can dominate the hybrid cloud market as AWS has in public cloud.
First, on its technology approach. AWS Outposts runs Amazon’s existing cloud technology, top to bottom. According to AWS, they had to add minimal additional technology to extend management to the customer site.
Amazon’s approach is much simpler than Microsoft’s competitive AzureStack hybrid cloud solution. AzureStack runs on top of third-party hyperconverged solutions, which then require an integration layer to work with the Microsoft Azure cloud. Microsoft adds to its integration pain by supporting multiple hyperconverged partners. Each partner will require ongoing compatibility testing, which further constrains Azure development velocity.
AWS chose to use the same hardware used in the AWS data centers. Amazon has already developed and proven their own Nitro version of hyperconverged at much greater scale than anyone else. For Outposts, they just added a chip to each device for zero-touch provisioning on-premises. Simple. It also creates a big moat since the hardware includes technology proprietary to Amazon.
AWS will offer a flavor of Outposts running VMware Cloud. This may seem similar to the Microsoft AzureStack partnerships with DELL/EMC, Cisco, HP, and others, but it is distinctly different and a much cleaner approach. This is the key difference: Amazon partners build on top of the AWS foundation while Microsoft relies on partners for its AzureStack foundation.
In Amazon’s case, they put the burden on VMware to develop its VMware Cloud overlay on top of an existing AWS service (the AWS EC2 Bare Metal Instances). The success or failure of VMware Cloud is up to VMware. Outposts can thrive with or without VMware. This is just another route to market for Amazon. Meanwhile, Microsoft is entirely dependent on its infrastructure partners for the success of AzureStack, and has to share the customer wallet with these partners.
AzureStack is not the only other hybrid cloud option, but other packaged or one-off stacks suffer similar challenges – multiple solutions are painful and brittle to integrate, limit capabilities, and reduce agility. The integration challenge is ongoing as each product is independently developed. I am a big fan of Cloud Management Platforms (CMPs) to execute a multi-cloud strategy for this exact reason.
Second point, Amazon’s costs should be much lower than other managed public cloud providers. Outposts runs on commodity components purchased at mega-cloud scale. But they aren’t really commodity components. They include Amazon’s proprietary chips and connectors for snap-in, zero-configuration provisioning. Amazon eliminated expensive, high-failure power supplies from each component with their custom cabinets with DC power PDUs. Yet the Outposts hardware can be low cost because Amazon has vertically integrated, displacing the traditional hardware manufacturer and reseller/integrator, with unmatched economies of scale. Nobody has a more labor-efficient data center solution. Now it will be available at any location.
The support and maintenance of Outposts will be much more efficient than other hyperconverged solutions because of Amazon’s unique hardware and services approach. It is promised to be a true plug-and-play architecture with efficiencies that only AWS could develop. Amazon handles tech support and repairs in their facilities. Its a single vendor managing the entire stack – pretty compelling. Customers just swap out bad components for the new ones shipped from Amazon. The components simply snap in place, leaving little room for error and little need for outside services to swap hardware. The labor savings should be significant for larger enterprises.
Outposts’ proprietary intelligent cabinet has to be more costly than a standard rack because it incorporates a PDU (a power strip at the back of the cabinet). But at the same time, it eliminates costly power supplies in each component. It also makes installation a matter of rolling the cabinet into place and plugging it in.
So, why not jump on AWS Outposts for all your on-premises workloads? The AWS architecture is now very mature, but your applications need to fit. You are locked into the supported AWS services within AWS Outposts just as in any public clouds. AWS security is also strict. For example, networking is locked down and does not support multicast or broadcast required by some applications.
Cost is also a factor. Your hardware load balancer may need to be converted to a virtual edition on AWS. You can’t build and operate all same capabilities at a lower cost than AWS, but that does not mean its the cheapest on-prem or colocated solution.
It remains to be seen how aggressive AWS will be with Outposts pricing. But Amazon has demonstrated its willingness to sacrifice profit for growth. AWS has the opportunity to kill the managed hybrid cloud competition on price and capabilities while keeping healthy profits.
As I see it, Microsoft is betting customers will continue to trust their existing hardware partners and resellers for hybrid cloud. Amazon has big technology and cost advantages to compel customers away from their traditional partners. That proved true in public cloud and I see every reason AWS can replicate that success on-premises.