4 Min reading time

Virtual to Bare Metal OpenShift: A1 Journey to Optimized Deployment

Overview

At CROZ, we've led a transformation for a regional A1 operator, successfully transitioning their infrastructure to two robust production-ready Bare Metal OpenShift clusters. This strategic shift marked an important moment for them, since they moved away from Virtualized Environments towards hybrid setups. Our client, has recognized the advantages of Bare Metal OpenShift deployments, particularly when operating at scale.

Bare Metal OpenShift Transition
the_challenge

Challenge

RedHat’s subscription options, tailored to deployment type, have moved our customer in a new era of cost optimization. As opposed to Virtualized Environments, where a single subscription covers up to 4 vCPU cores, in Bare Metal deployments a single subscription covers two CPU sockets or 64 CPU Cores, regardless of which threshold is reached first.

Notably, subscription fees remain a non-issue for Master nodes and infrastructure nodes. In turn, this deliberate decision has ultimately resulted in substantial cost savings and a competitive edge.

For instance, if the client has a request to add an additional 64 vCPU cores to a Virtual deployment; the price difference is 75% in favor of Bare Metal nodes.

When comparing the performance of virtual machine and bare metal workers, we observed several notable differences.

 

Bare Metal OpenShift Clusters
the_challenge

Solution

Firstly, bare metal machines demonstrate remarkable network efficiency, with a network latency that is significantly lower than VMs. This indicates that data takes much longer to travel between different components in a virtualized environment.

Secondly, bare metal machines deliver exceptional CPU performance, surpassing VMs in processing speed by a large margin of around 50%. This suggests that virtualization introduces additional overhead and processing steps, resulting in slower CPU performance.

Furthermore, bare metal machines exhibit far less CPU utilization, almost half of their VM counterparts. This implies that virtualization places extra demands on the CPU, leading to increased utilization.

In terms of RAM latency, both write and read operations were faster in a bare metal configuration. Write latency dropped around 45%, while read latency showed even better improvement of 60%. These findings indicate that accessing and manipulating data in bare metal configurations is more efficient compared to virtualized environments.

Storage performance also exhibited significant differences. The number of storage transactions per second was drastically increased in a bare metal configuration, for an order of magnitude.

Additionally, storage latency was lower in a bare metal configuration by a third.

Taking all of the above into account and pairing it up with major cost-saving along with massively increased performances across the board, choosing a bare-metal cluster was the only natural option for both us and the client.

Bare Metal Nodes

Cost Savings and Performance Gains

In a dynamic A1 landscape, the ability to scale storage is mission-critical. Conventional virtualized environments often rely on attaching virtual disks to VMs, and this way accidentally limiting the ability to scale across multiple nodes. While external storage solutions like NFS or S3 object storage serve as viable workarounds, they come with inherent limitations.

With our approach we were able to achieve the best of both worlds. By harnessing Bare Metal infrastructure for OpenShift clusters and seamlessly integrating OpenShift Data Foundation, we’ve architected an ideal solution.

 

An added aspect of the Infrastructure nodes is the incorporation of additional disk drives, enabling the establishment of distributed storage seamlessly integrated into the cluster through OpenShift Data Foundation. This innovative deployment unlocks a pool of advanced features, including:

  • Built-in S3-compatible object storage
  • Volume snapshots
  • Shared file system (RWX)
  • Shared block storage
  • Horizontal scalability
  • Support for structured data (SQL/NoSQL databases and data warehouses)

The RWX capability, allowing for workload scaling across any number of nodes, unaffected by node-specific storage constraints, stands out as a transformative feature. In contrast to storage deployed via a cloud provider, a common practice in vSphere deployments, OpenShift Data Foundation on Bare Metal nodes provide unparalleled price-to-performance advantages.

“With CROZ’s solution, we have significantly reduced the number of worker nodes, transitioning from many virtual ones to just a few physical bare metal nodes, which consequently optimised our subscription consumption. Additionally, we have freed up a substantial amount of VMware resources and subscriptions previously used for running Openshift nodes.

A key advantage of this transition, beyond significant cost optimisation, has been the increase in available resources, with the new clusters providing 10 times more CPU resources and 3 times more RAM.

Throughout the migration process, CROZ guided us effectively, enabling us to operate and further develop the clusters using in-house resources, which brings long-term value.

We are extremely satisfied with the project’s outcome and would like to thank CROZ for their professionalism, assistance, and expertise. As always, working with the cheerful CROZ team was both enjoyable and rewarding.”

Our client’s strategic decision to maintain Master nodes in a virtualized environment preserves the much-needed stability and high availability. Additionally, the shift to Bare Metal nodes unlocks the full potential of server hardware, effectively optimizing subscription costs and bypassing the typical overhead associated with virtualization environments. All this is achieved while maintaining an exceptionally competitive cost structure.

Get in touch

If you have any questions, we are one click away.

This site is protected by reCAPTCHA and the Google Privacy Policy and Terms of Service apply.

Contact us

Schedule a call with an expert