The past several years have seen fast growth in enterprise hybrid cloud adoption as more and more businesses seek a hypothetical sweet spot between purist extremes of 100% private cloud and 100% public. In the process, according to studies by Intel and others (including Platform9), the industry has refined our understanding of where that sweet spot lies for businesses of different sizes, types, and operational priorities.
We always knew that organizations sought the public cloud — and hybrid strategies — in order to control capex while taking on more-manageable opex, and benefiting from public cloud agility. This vision — and its companion ideas of ‘on-demand’ and ‘pay as you go’ cloud services — has been responsible for much of the explosive growth of AWS and other public cloud providers.
The trouble is that — as appealing as agility and capex to opex migration may appear — private clouds are, by and large, still more cost-efficient for many tasks. As we now know, hybrid cloud strategies achieve low costs by walking a tightrope between the agility of public cloud operations and the control and performance/cost optimization possible on private clouds. Greatest cost savings result when a limited public cloud use is combined with an open-source based private cloud. Public cloud use would arise due to dictates of specific workload requirements e.g. when dealing with burst demand, for achieving resiliency and/or obtaining geographic/data-location benefits for business-critical applications.
No enterprise segment’s needs are addressed perfectly by public or private clouds. Either approach requires trade-offs. Hybrid cloud strategies help organizations manage these trade-offs more efficiently. In general, economics favor heavier use of private cloud in hybrid environments. Cost benefits begin to slip as the percentage of workloads on AWS or other public cloud providers increases — smaller businesses tend to less able than larger ones to exploit public cloud advantages without losing control of costs.
Learn more about key strategic and cost considerations for Enterprises when planning for Hybrid Cloud adoption in our recent webinar – Let’s Talk About Hybrid Clouds
Small enterprises seem to find the greatest benefit with a hybrid cloud approach when workloads on public cloud are > 20% or so of the total (< 80% of workloads remaining on private cloud). Larger enterprises — likely because the proportion of business-critical workloads they need to support is usually higher than for smaller firms — seem to have net positive savings with a hybrid cloud approach with up to 50% of workloads on public cloud.
Given this mathematics, a rational approach to hybrid cloud investment might be a two-pronged strategy, where public cloud is exploited for its agility and convenience — constraining capital expenditure until the needs surrounding workloads are very well understood, after which those workloads are moved to private cloud when possible, keeping costs optimized and public cloud utilization constrained. Workload mobility, in some form, is essential for hybrid cloud strategies to really work. The more seamless mobility can be, the easier it is for every organization to find its own, dynamic hybrid cloud sweet spot between private and public cloud utilization.
VMware on AWS and Hybrid Clouds
This past August at VMworld, VMware announced that it had gone live with a VMware-billed-and-supported hybrid cloud service delivering their flagship Software Defined Data Center (SDDC) on bare metal servers, housed in AWS’ data centers. The long-anticipated offering — in the works for over a year — was met with squeals of glee from tech press and pundits. Wall Street followed suit, rewarding both parties to the announcement — VMware was up 49% in September (and Amazon, 20%), over a year prior.
The AWS announcement appears to have neatly turned around a period of failed hybrid and public-cloud experiments for VMware — experiments ending in the sale of their vCloud Air business to European hosting provider OVH, and ostensible (short-lived) exit from the managed/hosted cloud business. Most analysts greeted news of the exit charitably, agreeing that losing vCloud Air was a good move for VMware, who, despite being the enterprise private cloud leader, was ill-equipped to manage the operational and capital burden of building out and staffing responsive, secure global data centers. But it left VMware customers wondering if the earlier dream of seamless hybrid operations between their private clouds and vCloud Air services run by top-tier global partners had been scuttled.
To many VMware users, therefore, the VMware+AWS announcement came as very good news. Treating VMware as sole provider, they could obtain seamless, AWS-hosted Software Defined Data Center (SDDC) clusters based on vCenter (compute), vSAN (storage), and NSX (network). Live migration (with millisecond downtimes) of workloads between AWS-hosted VMware clusters and on-premise VMware clusters would be supported (or at worst, for users who hadn’t upgraded to NSX and the latest version of vCenter on premise, ‘cold migration’ of stopped workloads from one platform to another). Integration of the hosted VMware SDDC with AWS’ other services (e.g., for storage, serverless compute, etc.) would be provided gradually.
Sounds pretty neat. But closer analysis points to some potential gotchas, as listed below.
VMware on AWS: Hybrid Cloud Challenges
VMware Cloud Foundation/Software Defined Data Center is an all-VMware product: not just single-vendor but extremely opinionated (locked-in). And achieving seamless harmony (including live migration) between SDDC clusters hosted on AWS and on the premise requires customers to buy into SDDC on-premise, rather than staying with recent vCenter, third-party storage, and/or SDN options that may otherwise be perfectly viable.
VMware on AWS represents a way for VMware to ‘white label’ AWS while violating one of the basic assumptions organizations make about how public clouds should be used and hybrid cloud strategies should work: the idea that public clouds let you both pay as you go and stop paying when you don’t need services. VMware’s pricing for its new AWS service revolves around a firm, sunk-cost, one- or three-year commitment to VMware (paid regardless of your actual usage). And this service is billed in terms of clusters whose node-counts vary from 4 to 16 (i.e., not zero to 16), so there’s no way to turn the AWS-based part of your VMware hybrid cloud all the way down.
We think technologies available now, such as OpenStack Omni or containers-as-a-service (CaaS), based on Kubernetes offer a much more cost-efficient, operationally-beneficial, and less complicated route to driving your organization’s hybrid cloud strategy, and achieving truly dramatic savings. If need be, these alternate routes can compliment existing on-premise vSphere/ vCenter deployments exceptionally well. In our next installment, we’ll discuss some of these alternatives.
Learn more about hybrid cloud strategies and cost considerations:
- [Video] KubeVirt – Beyond Containers: Coming full circle back to VMs! - September 12, 2019
- The unforgiving cycle of cloud infrastructure costs (and the CAP theorem that drives them) - April 23, 2019
- Transitioning from managing VMs to orchestrating containers - November 28, 2018