CMTG boosts private cloud capacity and prepares for AI with Lenovo
New servers deliver faster workloads and greater capacity.
Cloud provider CMTG has partnered with Lenovo to upgrade its private cloud infrastructure with AMD EPYC-powered servers, reporting a 70 percent reduction in CPU utilisation following the deployment.
The upgrade has also improved workload performance, increased computing capacity and laid the groundwork for CMTG to expand into private AI services. The workloads weren't simply consuming less CPU, they were completing faster.
“Customers running traditional database reporting workloads saw some reports complete approximately three times faster than they had on the previous infrastructure. We also saw significant improvements across everyday compute-intensive activities such as SQL reporting and Exchange filtering,” said Carl Filpo, MD, CMTG.
The performance improvement was largely attributed to the new AMD EPYC 9475F processors, which offer higher per-core performance and more cache, alongside a move to faster DDR5 memory. Each new server has 96 physical processor cores, giving CMTG significantly greater computing capacity.
Filpo explained that for CMTG, the biggest economic benefit is the ability to support substantially more users and workloads from the same infrastructure before needing to make the next capital investment.
CMTG purchased ten new servers and built additional capacity into the platform, which it believes will allow it to double its current workload without further hardware investment.
“The ability to defer further hardware investment is significant when an equivalently configured host in today's market can represent approximately $300,000 to $400,000 in capital expenditure,” Filpo told CRN Australia.
Building a platform for private AI
With the upgrade, CMTG sees an opportunity in private AI inference and specialised workloads that can securely access customers' business data.
The company is developing an AI worker platform for its own managed services operations, aiming to reduce administrative workloads by analysing service tickets, engineers' notes and customer interactions to identify recurring issues and flag matters requiring attention.
"The objective isn't to replace the engineer or the customer relationship. It's to use AI to perform analysis at a scale that would be difficult to achieve manually, surface the items that require attention, and give our people more time to act on those insights," Filpo said.
CMTG also plans to offer customers isolated AI environments that can securely access information held in on-premises and cloud systems, with data sovereignty and security central to the approach.
CMTG's platform has been built with the Australian Government Information Security Manual (ISM) Protected guidelines in mind and is currently going through an IRAP assessment.
“We’re also designing the AI platform around the same principles of separation between customer environments and CMTG's own systems,” he said.
The other significant investment is in people and CMTG is developing its AI capability and skills to support customers with AI adoption.
“Ultimately, we're not trying to recreate a hyperscale AI training environment,” he said.
“Our opportunity is to provide Australian organisations with private, locally operated AI compute that can securely interact with their own business data, remain isolated according to their requirements and scale as their adoption of AI grows,” he added.