ScaleMatrix Sets New Benchmarks for Performance and Efficiency

I’ve been obsessed with computers since I was four years old. I like to joke I was on amber-and-black monitors before there were black-and-white screens. And it’s not just computers—as a gamer and photographer, my world revolves around technology. When I clock out at the end of the day, I return to my home lab and continue tooling around on personal IT projects well into the evening.

I’m a technologist through and through, driven by all things technical.

Offering Customers Answers That Make Sense

I have worked alongside similarly dedicated people at ScaleMatrix for nearly seven years. We’re a hybrid service provider delivering cloud, colocation, managed services, data protection, and connectivity options under one roof. I came here after three years at a cloud service provider, where I believed I had learned everything I could—especially how to work with clients.

I enjoyed working with my boss there, who had left the company for ScaleMatrix. So when I started looking for a new opportunity, I reached out to him, and he told me about some things going on at ScaleMatrix. They offered so many services, which meant there were opportunities to grow and have fun learning all kinds of technologies. 

ScaleMatrix works with a wide variety of customers. We offer every level of engagement, whether clients want us to fully manage their solution, hand them the keys and send them a bill, or anything in between. We handle managed and hybrid services for SMEs and provide colocation services for large enterprise customers. We offer a public cloud, white-label cabinets at our data center, and other services of varying complexity.

As our lead engineer for customer-facing solutions, I give our clients answers that make sense for their circumstances. Most technology gets easier to use every year. When tech components fit together like building blocks, the more pressing challenge is to listen to our customers, understand their goals, figure out how to get there, and ensure we’re all on the same page. I’m a problem solver at heart, and I love finding the perfect solution for our customers’ needs. 

The Ease of Working with a Single Vendor 

One of the many incredible aspects of my job is working with HPE technology. ScaleMatrix started using HPE Nimble Storage solutions around 2018. That initial purchase included HPE Synergy servers for our public cloud and customer deployments in our data center. Our edge systems also run on HPE ProLiant servers, and we deploy fully-managed HPE ProLiant rack servers to customer sites for those who want on-prem solutions. 

We began working with HPE because the company offers a wide umbrella of solutions that equal or exceed the competition. Choosing several HPE solutions allows us to operate our data centers and cloud solutions from a centralized location. We don’t need separate systems to monitor, configure, and repair our storage and compute; we access everything through an HPE dashboard. And when HPE acquired our disaster recovery vendor, Zerto, we gained the ability to manage our backups under the same umbrella as our composable infrastructure, which is incredibly attractive and allows us to engineer better products for our customers.

It’s easier to deal with a single vendor, especially an organization with the manufacturing, research, and support capacity of HPE. Working with a single vendor also helps when dealing with an issue affecting multiple systems. For example, if we’re facing simultaneous storage and compute issues or a compute problem turns out to be a storage problem, all tickets are with HPE. We may have to speak to different support teams, but we don’t have to start from scratch with two separate companies.

We have 1:1 interactions with HPE’s teams and have even been in the room with senior executives and high-level technologists to discuss our concerns. In return, we provide actionable data and insights about HPE products that can help them refine future releases. Although HPE is a multinational corporation and we are a small colocation and cloud provider, they treat us as equals. 

Experimenting with ARM-Based Servers

In August 2021, I attended the HPE Discover conference. As a technology nerd, I was looking forward to the launch of HPE ProLiant RL300 Gen11 servers running on Ampere Altra ARM processors. I mess around with ARM-based computers like Raspberry Pi computers in my home lab and was excited to see the release of a server using that technology. Apple already had ARM-based Macs, and Google was using some ARM-based black boxes internally. However, this was the first time I had seen ARM processors in data center-grade technology rather than products targeting makers and consumers.

I expressed my enthusiasm about the new servers to the HPE team. Five months later, they asked if I’d like to run a proof of concept (POC) using this new technology. I jumped at the chance to trial our internal services on Ampere Altra-based servers. I wanted to test the touted benefits in real-world applications and position ScaleMatrix as a knowledge leader in this new technology.

After an initial setup, I started running experiments to see how well HPE ProLiant RL300 Gen11 servers worked within the context of our operations. First, I gauged the consistency of the user experience. I wanted to ensure that any of our engineers could set up an ARM-based server without tons of workarounds and specialized knowledge.

I asked our developer team to migrate our corporate website to an HPE ProLiant RL300 Gen11 server without telling them it was a new platform using a new processor technology. They successfully ported the site to the new machine using our existing software stack and returned it to me. It was a significant milestone because the solution behaved like any other Linux server. The Ampere Altra architecture was completely transparent to my engineers, which made me confident we could migrate more applications to this solution.

Next, I tested the consistency of server operations. I wanted to see how consistently pages loaded, so we refreshed the corporate website 50,000 times and found that 99.997% of the results loaded within a millisecond of each other. This metric is an important performance indicator we can share with customers with sensitive workloads.

I also ran some general compute and video transcoding benchmarks. The latter is not mission-critical for us, but it’s a great way to benchmark raw performance. The initial results were neck-and-neck, but after fine-tuning the performance of our HPE ProLiant RL300 Gen11 server, the new Ampere Altra-based servers consistently outperformed servers using x86 chips by 10%.

Making Strides in Power Efficiency

These results were very promising, but our most important benchmark was power efficiency. Our HPE ProLiant RL300 Gen11 consumed 30–40% less electricity while running the same workloads as x86-based servers. That’s a massive win for us because we run data centers in five cities in the lower 48 states, including San Diego, which has some of the highest electricity prices in the country. More compute per watt is also a win for customers, whether they’re using on-prem managed infrastructure or colocated servers at one of our data centers.

Lowering our power consumption demonstrates our commitment to our values, as green computing is one of ScaleMatrix’s top priorities. One way we reduce energy usage is by putting our server racks in sealed metal cabinets with water coolers and heat exchangers on top instead of the oft-used approach of blasting air conditioning and hoping it circles through the racks effectively. Our green cabinet technology lets us take off-the-shelf servers, place them into these sealed cabinets, and reduce cooling costs by 40%.

Our cabinets also free up space because we can boost rack limits from 14,000 watts to 70,000 watts, allowing us to offer more density than other data centers. Adding HPE ProLiant RL300 Gen11 servers to that mix will further improve efficiency. 

A Successful POC Opens Doors to Much More

Although we haven’t put HPE ProLiant RL300 Gen11 servers into our production environment yet, I’ve secured buy-in from the company. It took a minute for people to understand my ask because it’s not a big adoption or a paradigm shift. Our applications run like they used to, but everything is different and more efficient under the hood. My boss was confident enough to let me run with the project, but he really started to pay attention when I showed him the numbers from the POC. After that, we eagerly entered discussions with HPE about next steps.

We hope to migrate some internal services to HPE ProLiant RL300 Gen11 servers. Anything that runs on Linux can potentially move to the new platform, including our CRM, Nagios monitoring platform, and VMware. We are also planning to replace our hardware load balancers, which are very expensive, with software-based ones running on general-purpose servers. Although I don’t want to generalize, I have not yet run into a piece of software that doesn’t run on our HPE ProLiant RL300 servers, which is very promising.

Our initial goal is to start with a quarter rack of HPE ProLiant RL300 Gen11 servers purely for internal use, continue to test it, scale up to a full rack, and start offering this technology as a service to our clients. Ampere is working on a multi-tenancy solution to support service providers like us, and I eventually foresee setting up virtual machines for our customers similar to AWS and our own cloud offering instead of leasing the HPE ProLiant RL300 servers alone.

The Technology Will Blow You Away 

My experience with HPE ProLiant RL300 Gen11 servers blew me away. I expected efficiency gains, but I didn’t expect it to be as well put together and as seamless as it is. ARM compute can involve a lot of customization and weird workarounds. That’s what I expected with the HPE ProLiant RL300 servers because I’ve seen it before. Adopting new platforms and solutions often comes with some benefits, a lot of promise, and a ton of “gotchas”—sometimes critical ones that are hard to get around as a service provider with hundreds of diverse clients.

There were no “gotchas” when transitioning from x86 to Ampere Altra technology. The servers just work, and they’re faster, greener, more efficient, and more compatible than I ever imagined. ARM-based technology is the future of Linux-based workflows, servers, and data centers, and they’re here today.