Greetings everyone – it’s nice to be back on the Micron Storage blog. I hope you’ve enjoyed my series on the positive impact of Micron SSDs via my series "In the Homelab". Even though the homelab is cooking up some new experiments (never a dull moment), in this blog post I want to share some of my experiences out in the field. Talking with customers, which I do on behalf of Micron Storage, takes a good portion of my time. I have said for many years now that “customers are where it’s at” – and I use that as a framework for my discussions. Yes, lab work is great, internal projects come to fruition – but there is nothing like getting into discussions with customers who are trying to solve real workloads with your technology.
Over the past year, I’ve had discussions with literally hundreds of customers, ranging from the world’s largest cloud and service providers to small businesses of less than 100 employees. While the scale is enormously different, their challenges are remarkably similar – save money, save time, become more efficient, reduce complexity. But as the cliché goes, the devil is in the details, and no single customer is exactly the same as the next. It’s what makes my job as a VP of Advanced Storage hugely interesting and always fresh.
It is those details which I want to expand upon today. Without having to write a 5,000 word blog entry – here are two key points I’ve distilled from my hundreds of customer interactions.
- Customers understand that flash in general and SSDs in particular are fast, and make many workloads more efficient by reducing I/O latency. But they aren’t quite sure how to implement SSD technology, so they have to proceed with caution. Should they use all-flash arrays – replacements for their existing hybrid or all-rotating arrays? Should they use a hyperconverged approach, with compute and SSD storage in the same server, with some form of clustering? Should they transition to scale-out bare-metal servers, and ‘roll their own’ infrastructure, much as some of the world’s largest cloud providers do?
- Customers want to know “what’s coming” – what to expect next from their vendors regarding compute, networking and storage, and how it all works together. From Micron, they want to hear what’s next in the world of flash memory and all the other memory technologies that we design and manufacture – especially 3D XPoint – because they need to plan “down the road” in order to stay competitive in their own industry.
With regards to the first point, I spend quite a bit of time listening to companies who are in transition. By this, I am referring to their choice of on-premises or hosted architecture to run their workloads. While traditional infrastructure – servers interconnected via adapters to a storage area network in turn connected to storage arrays – is still in wide use, most of the customers I visit are quite intent on moving towards a different approach, one involving servers as the storage platform of choice. Whether they also wish to converge compute into these same servers – i.e. hyperconvergence – is a mixed bag. Some do, others don’t. Some love the concept of hyperconvergence, especially for virtual workloads. Others eschew that approach and want to keep compute separate from storage, and rely on high-bandwidth LAN technology such as 40 or 100 Gb/s Ethernet or (in some cases) FDR or EDR Infiniband.
In either case, the ‘serverfication’ of storage – a phrase Micron coined several months ago – is happening, and I see it happening quite rapidly. Nearly every customer I see is in some stage of adoption of this – be it widespread and mainstream affecting the majority of their workloads, or starting out with a specific workload or two. The advent of sophisticated, powerful virtualization and software-defined-storage (SDS) stacks is propelling this, along with extremely efficient and cost-effective SSDs such as those from Micron.
Put these two elements together – which many customers are doing for certain scalable workloads, especially those constructed from open-source (e.g. Apache on Linux) components – and you get a strong shift away from the time, complexity and expense of traditional architecture. Now, don’t get me wrong – there is nothing ‘wrong’ with SANs and arrays. Nothing at all. I’m “a SAN guy” from way back. The key difference is that today, there are highly effective, significantly cost-reducing, order-of-magnitude-time-reducing alternative approaches to that architecture, one that the customers I see are asking for help to construct. At Micron, we are helping them do just that.
In closing, what I see in my travels is a rapidly changing world – for the better. Some customers are just beginning to shift, others have already done so and are looking for help to optimize their approach. In either case, I can tell you it’s great fun, and I enjoy the interaction of combining business and technology. Best of all, I get to either introduce customers to wonderful new technology and methods to leverage it. Or I get to help them optimize their resources and spend by thinking of new and different ways to solve workloads in the near future given their new server-based, memory-intensive, data-close-to-compute architecture –by hitting the whiteboard and sketching it out. There’s nothing better than that if you’re a nerd like me!
As I said before, customers are where it’s at. I hope to see you in the field – out and about – and help you reach your goals with Micron as your partner.