Not All Workloads Work Well in a Hyperscale Environment

December 10, 2020

How do you determine what the right infrastructure is for the right workload? Typically, when you think cloud, you automatically think of cloud computing platforms on a metered pay-as-you-go basis. However, a hyperscale environment may result in much higher costs than what the simple price calculator suggests, like if the workload is designed on a more traditional multi-tier architecture, with isolated security subnets for the relational database and the web and app tier.

Some workloads may be better redeployed into a managed colocation environment where there are lower costs of hosting, cooling and power, and organizations can be freed up from maintaining legacy infrastructure from a patching and security hardening perspective. This is the managed services component that comes into play when choosing a cloud managed services provider. 

Many companies are realizing their push into the hyperscale was ill-advised for a particular workload and are pulling it back into a hosted private cloud or colocation environment. There’s even a name for this in the industry: The Boomerang Effect.

It’s not that hyperscalers are bad, they are great for the right workload and the right client, but not all workloads work well on hyperscalers. Here is a use case for an instance where a workload is not a good fit for a hyperscale environment:

A bank has two large virtual machines (VMs) with a minimal amount of storage for ~$500 per month. Within the virtual private clouds (VPCs) is a networking construct. It’s very common to separate out your web and application servers from your database servers for security reasons. And they are even in different security groups, in this case, the EC2 security group and MySQL database security group.

The problem comes in with VPC peering which has to be set up between these two VPCs—meaning the virtual machine in subnet A has to talk to the VM in subnet B because the database and web and app tier are all intended to work together. With this example, which is very typical for a banking OLTP application, data is stored in a relational database but the web and app tier are interacting with it.

All the traffic that goes between the application server and the database server is going to get tracked and charged as an egress fee between VPC A and B, even though they are in the same region and availability zone. If you add up that traffic for what is just a 1 GB connection (and most are 10 GB) and plug it into the simple monthly calculator, it comes out to $3,700 a month, a 750% increase over the flat fee for just the two large VMs.

Not All Workloads Work Well in a Hyperscale Environment

Cloud adoption continues to grow at an unprecedented pace:

  • 90% of companies use some type of cloud service
  • 85% of organizations have deployed a hybrid cloud environment
  • 77% of enterprises have at least one application or portion of it on the cloud
  • 1,427 average number of cloud services consumed by enterprises
  • 60% of organizations use cloud technology to store confidential data

Picking the right cloud is critical to the success of your digital transformation and hybrid IT strategy. You need a partner with a large cloud portfolio service provider which also provides a fully recoverable production-ready cloud services design, who will work with you to get to that right answer, using your workloads to guide recommendations.