Over the past few years Nutanix has had the ability to work readily with customers looking to migrate important, often business critical, legacy applications from a traditional on-premise environment. First, it was migrate to a hyper-converged infrastructure (HCI) still operating on-premise, but now it is increasingly about migrating them to the cloud.
The term ‘legacy applications’ has evolved a bit as well. It used to mean the trusty old war horse applications at the heart of running any and every business. Now, it means any application that is already in place, even if they have not been there that long. Every mainstream business applications is now a target for running on an HCI and for some of them, such as SAP’s HANA, it is arguably one of the best environments available.
It is not surprising, therefore, that companies like JR Simplot have taken up the idea of running HANA on an HCI environment as senior engineers from the company explained at the recent Nutanix.NEXT online conference, breakdown down some of the issues and benefits that came up in getting a new HANA-based application suite up and running.
James (JC) Miller and Troy Ladish are both senior ERP Infrastructure Engineers with JR Simplot, and both deeply involved in a two year-plus project called Bedrock, which they said could be described as the beginning of digital transformation, a foundation for processes and workflow within the company designed to create unity between the varying business groups. From their specific IT perspective, it also meant reducing the amount of managed applications internally, and in particular the fact that they have 30 plus ERP systems.
JR Simplot is a privately-held agribusiness company that spans the entire spectrum of ‘mine to plate’ products and services. This ranges from mining phosphates to make fertiliser, through to advising farmers on their crops and what they need to be successful. It also buys harvested crops for food manufacturing, plus operating a big cattle feed business. It operates across the globe, including the USA, China, Australia and Japan amongst many other countries.
In an organization that big and diverse, operating across multiple countries, the pair faced a number of challenges, not least being the number of unknowns they would have to be working with at the start, as Miller noted. So support became a key element of the mix:
We had to build an infrastructure without even knowing what was going to be running on it with sizes where we had to just kind of make our best guess and start from there. With any big project you're going to run into challenges, no matter who you're working with.
Getting ahead of the future
One of the main challenges of the project was that it was led as an IT initiative, which meant that the first task was to get the business leaders on board. For Miller and Ladish, it was just a matter of building new infrastructure – just another day at the office. But for the business users of the various applications, the project could mean changing the way they'd been doing their job for years. In practice, this gave them the task of getting the entire company on board, and especially the different leaders from each group.
This was not helped by the fact that Nutanix was, when they started the project, still considered a relatively new solution, while running HANA applications on it was still not well-proven. Despite those possible objections, the decision was made to go with Nutanix Prism because an easy, highly scalable solution was needed, although as Miller noted, there were other reasons as well:
It was mainly the desire to get ahead of the future a little bit and have that ease of management and scalability that we didn't get from our old three tier architecture. We wanted something that was a building block for the data center so we could just bring new blocks in when we needed to expand.
In addition, because they didn't have solid requirements to work with, flexibility became a key factor for them, while as noted, support was also a crucial element. Inevitable ‘new project gotchas’ emerged, such as when implementing the RDMA network acceleration technology, contention between drivers for another vendor’s equipment and Nutanix became an issue capable of crashing VMs. Nutanix provided a dedicated support engineer who helped bring all parties involved in the issue together to get it resolved, sticking with the project right through to it going live.
It was the first time the company had worked with a systems integrator of the size of their partner for this project, and there were some challenges faced in that relationship. For example, the fact that sizing requirements were essentially being worked out on the fly, coupled with the fact that the hardware coming from the vendor required multiple changes to be made in order to prep the tin, creating steps that often took more time to complete than expected.
This ran alongside the fact that virtual machines were then being built with little or no requirements information, which led to rebuilding, often multiple times. Miller said:
Add to that, SAP had some very stringent requirements as far as running the HANA databases in a virtual machine that required us to consume at least a full socket of hardware for each of those databases because they did not support crossing Numa boundaries and required a clock timer to come from the CPU directly to the VM.
Fixes for this took away the ability to migrate machine between nodes, which made things tougher for the team. However, the pair reported that positive feedback from SAP in Waldorf is said to be yielding positive results and the problem is expected to be fully resolved very soon.
Despite these problems the final Go Live went very well, according to the pair:
We didn't have a single issue from the IT infrastructure side. I think a lot of this had to do with the fact that Nutanix was so involved throughout the entire process and that just made for a very smooth life. It's really about having the right partners that you can lean on when you run into trouble.
If there was one factor about the project they would change it was, perhaps, the most obvious one – getting some clear requirements on the size and scope of the project upfront. They acknowledged that having such information available would have made the project a good bit less frustrating.
On the other side of that coin however, Miller noted that the situation was eased quite a bit by the nature of the HCI environment and its scalability and being able to just add a building block when they needed to:
So we were never too far behind the curve on that. If we got when some new systems were coming, we could get a good estimate and order a couple of nodes and be ready to build it. And then maybe scaled it a little later if we needed to.