Will Google Kubernetes Engine (GKE) provide an advantage that AWS and Azure can’t match for lift and shift? Both AWS and Azure have gone the route of offering some VMware staging area for legacy applications. AWS has VMC on AWS and Azure has a hosted VMware solution. The idea is to keep operational differences to a minimum between on-premises hosting and cloud. However, the price has been a consistent factor. Customers worry inertia will result in indefinite reliance on VMware vSphere software.
On-Prem vs. Public Cloud – Apples to Oranges
The math behind running traditional workloads in the public cloud is difficult to swallow. On-premises VM customers don’t bear the full cost associated with running a VM. The expense spreads across multiple cost centers. For example, Facilities may pick up the price for both power and cooling while IT infrastructure picks up the costs of servers and OS software while Security picks up their associated cost.
All of these costs are needed to compare the expense of a Googe Compute Engine (GCE) instance to an on-prem VM. From the lens of the cost center paying the GCE bill, a GCE VM costs more than a VMware VM. Regardless of the real cost, the perception is public cloud is more expensive for traditional workloads.
Google Cloud has made a big deal of their Anthos Migrate service. Anthos Migrate gives the ability to move virtual machine hosted applications to containers and therefore Kubernetes. Response to Migrate has been mixed.
The approach isn’t new from a container modernization perspective. Docker announced a similar product in 2017. There doesn’t seem to be wide adoption of the strategy. However, Google may be able to offer something that Docker couldn’t – A Public Cloud infrastructure.
I took note of GKE not because of the advertised ability to “modernize” applications. I took note because of the potential to reduce the cost for lift and shift to the cloud. One of the challenges of the public cloud remains the expense of over-provisioning a VM. In most on-prem solutions, IT departments can be sloppy in provisioning the infrastructure.
If a web server only needs 8GB of RAM, 2 vCPUs and 20GB of HD space but is provisioned with 32GB of RAM, 4 vCPU’s, and 100GB of HD, VMware hides that cost inefficiency. However, in the Public Cloud, the VM owner will pay for that difference in capacity.
Running containers inside of VM’s allow organizations to make more efficient use of cloud-based VMs. Netflix noted this as an unexpected benefit of their Titus container project. The team migrated legacy Java applications into containers. The result – IT could stack web servers on a single Large Cloud VM. Netflix saw a reduction in their Public Cloud hosting costs.
Moving legacy applications to containers isn’t easy. There are multiple reasons why Docker has seen limited success. However, the promise is there as shown by the hyperscaler, Netflix. Netflix is a pure technology company. The company can afford to dedicate an entire engineering team to the effort. Google is looking to productize something that today remains custom.
Google’s conference is called Next for a reason. While the concept makes logical sense, the devil is in the details. There are still plenty of questions to answer around management, visibility, and security. However, the idea isn’t far fetched.
IT infrastructure subject matter expert (Cloud, Virtualization, Network & Storage) praised for transforming IT operations in verticals that include Pharma, Software, Manufacturing, Government and Financial Services. I’ve lead projects that include consolidation of multiple data centers and combining disparate global IT operations. “Three letter” Federal agencies have called upon me to lead the modernization of critical IT communication platforms.