“Existing conditions” is a term for conditions that don’t last very long. Amazon changed the face of retail. Google democratized access to information. Facebook and pretty much every social media site has changed the way we interact with each other.
Change came to Carrs Tool Steels, too. The U.K. supplier of tool steel and high-strength aluminum alloy products to manufacturers has seen its customers’ purchasing habits change from bulk-buying of large pieces of material to precision buying of blocked-up tool steel for just-in-time production.
It’s done a lot of work on its ERP system and its business model over the last decade so that it can more efficiently respond to clients whose own models lead them to place smaller — but more — orders. The upgraded Priority ERP software the company uses enables it to record, trace, and search the exact dimensions of every single piece of stock in the warehouse at any given time, ensuring within seconds that customers receive the closest match to their requirements.
But the hardware supporting it hadn’t changed. “The hardware has to be up to the job,” said Mark Mullaly, Carrs Tool Steels Systems Manager.
The infrastructure was comprised of a series of stand-alone, non-virtualized servers that were overloaded with distinct ERP applications that were all interconnected with each other. “We had multiple workloads on each server and should one workload cause problems, it would have a domino effect on other workloads,” Mullaly said. A problem with one piece of software could bring the whole operation down.
And indeed, Carrs did have a situation where its ERP system was offline for more than a day. Back when larger but fewer orders were being placed, he said, the company could work around infrastructure hiccups. But tight timelines in lean manufacturing means customers expect fast and reliable services — no excuses. For Carrs, a full set of backup and recovery capabilities to ensure always-on data protection was a priority.
Data Protection with Teeth
When first considering hardware options, Mullaly thought that it would make sense to just replace systems with other brand-name servers and storage or go with a build-it-yourself virtualization solution. Buying separate servers to run VMware with a single storage unit would be a single point-of-failure problem, though.
Costs and training were a concern. The company couldn’t afford to opt for a solution which, with licensing fees and support costs, would cost somewhere near £100,000 over five years. Plus, Mullaly said, he didn’t have the time to spend on high-end training for managing systems.
“We have a small in-house IT team, and that is me,” he said. “There is a limited amount of time that I can spend away from business, and I need a solution that doesn’t require learning thousands of different skill sets just to manage hardware.”
Hyperconvergence systems were the next option to consider as a scalable approach for pooling data center functions onto hardware and sharing them between virtual machines running on any host in the cluster. Equally important was that the HCI system have real teeth when it came to all-in-one data protection.
Carrs’ choice depended on the system being suitable for a small business and offering the same reliability that systems in much larger organizations usually have. Mullaly decided to go with Scale Computing’s HC3 solution, which integrates storage, compute, and virtualization layers of infrastructure into a single solution architecture for automating the management of its IT environment.
Mullaly said he appreciated the fact that data protection was at the core of the system, not just added as an after-thought. The integrated snapshot data replication for improved business continuity on the Scale Computing system is important to the company. Every business needs to be able to restore its data off-site or restore it from an earlier point in time to help it recover from an unplanned event.
“It is just so simple to set up schedules that can produce hourly and/or daily snapshots and then choose which schedule to apply to which VM,” he said. “The more impressive aspect, though, is that the production of these snapshots has negligible effect on system performance.”
Carrs wouldn’t have to think about licensing issues, or anything that adds complications by buying an HCI system. “We are not going to push the limits of any system we have. It just has to do the job of running our software reliably and consistently,” he said. Another plus of choosing a fully integrated solution is that data loss is minimized. There is more data loss — 2.36 terabytes on average — when more vendors are involved in data protection.
Carrs actually chose a slightly larger scale computing system than it had to, but Mullaly thought it was smarter to have its data center prepared for the next three years rather than having to increase the size of the system along the way. “We ended up going with a system that was about twice as big as our biggest estimate, to allow for data and processing growth.”
In the longer term, he could expand the HC3 solution by just bolting on another node. Carrs currently has one three-node cluster on the HC3 Platform running six VMs, and in the future it is looking to run up to 8-10 VMs. The Priority ERP software, file and print services, remote Desktop Services, anti-virus and Active Directory run on the platform.
Keeping an Eye on Costs
It’s common for organizations, particularly smaller ones with few IT resources to draw upon, to move their infrastructure to the cloud. But Mullaly was concerned that costs would unexpectedly ramp up.
“We wanted something that was ours and that we knew the costs of,” he said. “We still have work to do to make systems more reliable and use Scale Computing to the best of its capabilities.”
Scale Computing helps keep total cost of ownership down for customers by using the open source KVM hypervisor that offers similar functionality to VMware.
Carrs attributes, at least in part, its move to a new infrastructure to run its key ERP solution to its on-time delivery performance dramatically increasing over the last six months. There hasn’t been any unplanned downtime in the past seven months. “We are able to satisfy customers better now than last year.”
The popularity of hyperconverged solutions is reflected in the fact that the global HCI market size is expected to grow from USD 4.1 billion in 2018 to USD 17.1 billion by 2023, at a CCAGR of 32.9 percent.
Image used under license from Shutterstock.com