Breaking Through the Infrastructure Software RoadBlock
CenterGrid’s journey to VergeOS
Summary
CenterGrid’s legacy infrastructure software restricted its hardware choices, consumed up to 20% of CPU before the first customer VM ran, and charged per CPU and per core. After testing almost every commercial hypervisor and several open-source options, CenterGrid chose VergeOS. It outperformed every alternative, proved the easiest to run, and went from pilot plan to production in under a week.
The roadblock became a runway for dense, GPU-driven growth.
About CenterGrid
CenterGrid started operations in 2009 as a small business unit inside Ascendum Solutions, LLC. As the unit grew, it spun out of Ascendum. Under its new name, CenterGrid has achieved significant revenue growth and established itself as a leading managed service provider (MSP) and cloud service provider (CSP).
A Turnkey Service Model
CenterGrid goes beyond providing infrastructure. It delivers a complete turnkey managed IT service. The company has built its name on responding quickly to unexpected customer requests, such as a new deployment or the expansion of an existing one, and on strict adherence to project timelines. Rapid response at that pace requires a flexible and robust infrastructure.
CenterGrid serves a wide variety of traditional enterprise businesses and has developed specific expertise in the media and entertainment (M&E) industry. M&E workloads need rapid stand-up and push the limits of the virtualized environments typical in CSP data centers. High-powered GPUs drive CenterGrid’s M&E services for visual effects (VFX), computer-generated imagery (CGI) and a growing demand for artificial intelligence (AI).
Technical and Business Limitations
According to Chief Operating Officer Chris Beard, CenterGrid’s legacy infrastructure software had a restrictive list of supported hardware and was slow to add new hardware. It also exacted a significant performance penalty, often called the virtualization tax. That tax can consume as much as 20% of CPU resources before the MSP creates a single customer-facing virtual machine. Development of the legacy solution had also not kept up with the GPUs that M&E and many other industries count on.
The financial picture added pressure. CenterGrid used the vendor’s complete stack: virtualized storage, hypervisors, a management console and software-defined networking. Per-CPU and per-core licensing made that software a significant part of the IT budget and made it harder to invest in the ultra-dense physical servers that save data center floor space and reduce power and cooling costs.
The Selection Process
The CenterGrid team describes itself as a bunch of hardware nerds and welcomed the challenge of finding new infrastructure software. The team built an extensive lab and a detailed test suite that mimicked customer workloads, then ran it against the existing environment to set a baseline. The goal was to reduce costs and simplify operations without compromising the customer experience.
CenterGrid then tested almost every commercially available hypervisor and several open-source solutions. Each had advantages, and each had downsides, including forced hardware purchases, added complexity, no entry-level interfaces, poor cohesion among modules and only minor cost reductions.
Discovering VergeIO
After early disappointment with other alternatives, CenterGrid discovered VergeIO. VergeIO is a pure software company. It does not sell hardware and does not even keep a hardware compatibility list, only a modest set of minimum requirements. Its UltraConverged Infrastructure (UCI) offering integrates networking, virtualization and storage into one cohesive operating platform, VergeOS.
The licensing model caught CenterGrid’s attention first. VergeOS is licensed by the physical server, not by the number of CPUs or cores or by RAM or storage capacity. CenterGrid can build highly dense servers without raising software licensing costs, and the single license includes the networking, virtualization and storage that other platforms sell in pieces. Demonstrations of VergeIO’s migration capabilities also promised a rapid exit from the legacy platform if VergeOS passed testing.
Putting VergeOS Through the Test Suite
CenterGrid ran VergeOS through its full test suite. Its features were at least on par with the existing solution, and many exceeded it. VergeOS also offered better hardware support, especially for the GPUs CenterGrid’s M&E clients need. In benchmarking, VergeOS outperformed every other solution and was the easiest to set up and operate.
“One of the things that shocked us was how responsive the interface was even in a large environment with dozens of virtual machines. Screen refreshes that can take minutes in our legacy hypervisor happen instantly in VergeOS.”
Chris Beard, Chief Operating Officer, CenterGrid
Beard points out that it sounds like a small thing, but waiting on screen refreshes can consume hours per day for an operations team.
The tests also covered failure conditions such as drive loss and server downtime. In each case the workload kept running and moved to surviving nodes. Adding new nodes proved the most straightforward process of any product CenterGrid tested, including so-called turnkey solutions. Pricing and ease of migration both exceeded expectations, and CenterGrid projects licensing savings measured in six figures over the next few years. Those savings compound as each server delivers more performance and CenterGrid buys denser, space-saving servers.
VergeIO stayed responsive throughout the test period, especially on support and knowledge transfer. By the end, the entire CenterGrid IT team was on board. No one felt forced into the decision, and the team was excited to begin.
From Pilot to Production in a Week
With the proof of concept complete, the team drew up a plan to migrate infrastructure to VergeOS gradually. Then CenterGrid signed a new customer who needed an environment stood up quickly, and the organized pilot became a full production rollout. Within a week, CenterGrid had its first production VergeOS instance running, providing high-performance NVMe storage and GPUs and hosting a public real-time rendering website across dozens of nodes.
Operational Excellence
CenterGrid keeps adding to its VergeOS instance, taking advantage of easy node addition and the efficiency of each node. The platform is rock-solid and reliable, and the support quality continues to impress the team.
“The VergeIO Customer Success team has been amazing; they won’t let go of a problem until it’s solved, even if it is not necessarily a VergeOS issue.”
Chris Beard, Chief Operating Officer, CenterGrid
The team uses the VergeOS API with Terraform and Ansible and automates more tasks than ever. The responsive interface saves hours per day, and VergeOS Virtual Data Center technology lets CenterGrid securely deploy new customers in minutes.
| Area | Legacy stack | VergeOS |
|---|---|---|
| Hardware support | Restrictive list, slow to add GPUs | Modest minimum requirements, better GPU support |
| CPU overhead | Up to 20% before the first VM | Outperformed every solution tested |
| Licensing | Per CPU, per core, sold in pieces | Per physical server, full stack included |
| Interface | Refreshes take minutes | Refreshes are instant |
| Scaling | Dense servers raise license costs | Straightforward node addition at no added license cost per core |
| New customers | Slower stand-up | Virtual Data Centers in minutes |
The Bottom Line
With VergeOS, CenterGrid’s infrastructure is where it needs to be. The team can focus on serving its customers and winning new ones, not on whether the infrastructure will hold up to growth.
See What VergeOS Does for Your Environment
CenterGrid replaced its legacy stack with VergeOS and moved from pilot to production in under a week. Talk with a VergeIO engineer about what that looks like for your workloads.