The Value of Cloud-Neutral Strategy for Containers

Move can be a four-letter word. 

That’s certainly the case when it comes to cloud strategy. 

What may start as spinning up a workload—maybe a small artificial intelligence (AI) or machine learning (ML) service—in a particular cloud often expands in scope. The data created by that single process grows. And grows. Services are added and, as the project becomes successful and mature, that small service needs to be moved to take advantage of other cloud services.

But moving between clouds gets challenging. Moving means that you need to copy and replicate the data in order to be able to grow it out or to allow DevOps team members to be able to access and use their preferred cloud services across all clouds.  

In the contemporary environment, where a cloud-first mindset is rapidly taking hold, many enterprises aim for cloud-neutral container strategies. The goal: the ability to use the data where and when you want. But moving to the cloud (or, more accurately, to clouds) without first evaluating where your data resides can complicate the process. A cloud-neutral, data-first can facilitate a smoother cloud strategy for containers.

Prioritize Where Your Data Lives

Often, in a cloud-first, multi-cloud scenario, businesses turn to more than one cloud for the purposes of gaining advantages unattainable with only a single cloud. In such a case, containers or workloads sit on multiple clouds. This results in paying for them multiple times: securing them multiple times; meeting compliance regulations for multiple clouds; migrating containers between clouds; and paying egress fees to move those containers or workload between clouds. 

As datasets grow, so does data gravity—making all of the above more cumbersome. The pull of that data multiplies the resources required to use it. Moving the data around becomes increasingly expensive and cumbersome, possibly locking you into a single cloud provider. This becomes unmanageable with terabytes, petabytes, or exabytes of data, 

Taking a data-first mindset, instead, shifts the focus on where the data lives and how to access it. Prioritizing how you’re going to map your data to clouds streamlines the strategy. A cloud-neutral approach allows you to store your data once and use it anywhere—providing the ability to present a container or workload to any cloud at any time. 

With your data stored in a central location, it can be accessed as needed by multiple clouds—while you only pay for and manage it once. This cost-effective approach avoids vendor lock-in, while providing access to the hundreds of cloud services available from each of the public cloud providers. You can have a container running a certain service in cloud #1, using a business intelligence (BI) service from cloud #2, and IoT ingestion from a 3rd cloud, for example, with all data stored in a central location.

Advantages of Cloud Neutrality

Streamlined container migration is one of the top advantages of cloud neutrality. Take, for example, a large genomics project, where containers are running in public cloud #1. When working against the data, it may become apparent that the GPUs needed aren’t available in the location where the data is located in that particular cloud, necessitating a move to a lower hardware type for that particular service. With a cloud-neutral strategy, it’s easy to switch containers to cloud #2, with the requisite hardware, within a matter of minutes, then processing data against the desired hardware. With this approach to container migration, it is possible to take a container from cloud #1 and use that same container in cloud #2 (or #3) without having to move the control volumes or the data that the container is working against.

When enough hardware isn’t available, cloud neutrality allows the user to take what was on-prem, push the GPU process to the fullest extent possible, then expand out into the public cloud space and use peak time expansion (cloudbursting). With cloud-neutral containers, it doesn’t matter where they are running; they are able to access the cloud-neutral data from any public cloud. 

As great as they are, cloud services can fail from time to time. A cloud-neutral, multi-cloud strategy allows for service relocation, with automated failover to another cloud. In such a case, failure of one cloud service doesn’t prevent progress. Instead, processing data against the wall-clock times can continue easily.

Finally, among the most significant advantages of a cloud-neutral approach, are the benefits available with parallel computing. With multi-cloud access, you can spin up thousands of instances of a container in a cloud; in our genomic data workload example, this can include processes such as image classification workloads and AI/ML processing models, running thousands of workloads, in all three clouds, simultaneously, all against the ame data set, with results delivered in speeds unachievable through other approaches.

Considerations for a Cloud-Neutral Container Strategy

When evaluating the value of a cloud-neutral container strategy, consider: 

• Connectivity: Determine your current and future needs to connect to different clouds and for parallel computing. What connection, data access, and egress expenses may be associated with your connectivity needs?
• Data gravity: If you’re replicating data to two or more clouds, it takes time to process. No matter what bandwidth you have, you simply can’t beat the speed of light. Cloud neutrality solves for data persistency and data gravity challenges, mitigating the delays and latency between read and write between the different storage capabilities of different clouds. 
• Automation and orchestration: A cloud-neutral strategy is a great way for containers to run on multiple clouds, while all access the same data. It’s also worth prioritizing automation and orchestration for cloud-integrated services across AWS, Azure, and Google (using standalone platforms like Openshift or VMware Tanzu, for example, or Rancher and Mesos). Such tools, layered on top of a cloud-neutral solution, can eliminate concerns about where the container is actually living. (The remaining concern boils down to the cost of that spot instance.)
• Cost arbitrage: When looking at a cloud-neutral workload, consider the various factors that would be in play when presenting data to multiple clouds simultaneously versus maintaining (and moving!) multiple data sets (including security, compliance, validation, and testing) expenses.

How you access and use data—your most valuable asset—should be the central consideration of any cloud strategy. Running data services in containers as part of a cloud-neutral strategy can improve application portability and facilitate the adoption of the multi-cloud data services that help meet your goals for innovation.

Ray Kalmbach, Field CTO for Alliances at Faction, is an evangelist and architect of cloud strategies. Ray has led engineering and architecture teams to push the boundaries of what cloud infrastructures can deliver, providing fast and secure environments for customers, DevOps teams, and vendors. He believes that application environments can be delivered faster and more securely, allowing business needs to be met quicker than ever before.