This paper was sponsored by Pure Storage.
Prediction is very difficult, particularly about the future, but we can make some educated guesses. We can also be fairly confident that customers will want FlashBlade to be bigger, faster, cheaper, and use less power. This is easy to predict because this is what customers always want.
How to do that is the challenge that Pure needs to navigate. Let’s look at the pieces that Pure has at its disposal that it could use to adapt FlashBlade to meet future needs.
Hardware and Software, Together
One of the strengths of Pure Storage’s approach is the close relationship between the hardware and the software. While there are claims in the industry about the pre-eminence of software, there’s plenty of evidence for the importance of hardware.
Apple’s investment in hardware has paid off substantially with the M-series chips. Customers love the performance increase, and with lower power. Nvidia is driving a lot of the AI/ML market with its GPUs. AWS has announced the third iteration of its Graviton-based instance type.
Hardware matters. But so, of course, does the software that runs on that hardware.
Pure has already invested in hardware with its DirectFlash modules. Having greater control over the pathway to the flash is a major contributor to removing performance bottlenecks. By talking directly to the flash chips, Pure can bypass a lot of the abstractions and translation layers of more commodity flash modules that add latency and restrict the options available to the software.
The greater control over the hardware stack enables Pure to write their software in a different way to other vendors. They can do things in software to prolong the life of QLC flash, for example, which enables greater densities of flash. Software updates enable Pure to address these larger pools of flash, and to deliver additional functionality to existing customers. Sometimes this might mean taking advantage of hardware architecture that was built specifically to be used in the future, as software capabilities caught up with the hardware.
One challenge with hardware is that it’s harder to change than software. While it’d be nice if all possible future needs could be met with hardware that was built into an array you already have, sometimes physical upgrades of the hardware are necessary.
This is where Pure’s online upgrade capabilities combine with the Evergreen financial model.
Upgrading Hardware Like Software
If Pure makes changes to its existing hardware offerings, as it has in the past, customers are faced with a dilemma: keeping their existing hardware or buying new hardware to get the new features that can’t be delivered in software. With Pure’s Evergreen approach, that becomes a lot easier. You simply subscribe to hardware updates the same way you do with software.
If you always want to keep up with the latest, subscribing to updates makes a lot of sense. It also matches a mindset that views storage as a constantly changing set of options that provides a service to applications. We know that new arrays will come out in future, and upgrading will definitely happen. It’s up to you when that happens.
Storage as a service needs to live for a long time. That means decoupling the storage service from 3-5 year infrastructure depreciation cycles. Storage as a service needs to both last and also be changeable, probably multiple times, over a 10-year-plus time horizon.
Storage stops being something you acquire and starts being something you maintain. Making change easier is the key here, and again we see the benefits of Pure’s understanding of both the hardware and the software side of what they sell to customers.
The challenge for Pure is to provide the right mix of options for customers to choose from. Too few, and we feel unduly constrained. Too many, and it’s too difficult to decide between the baffling array of options (as anyone who’s tried to navigate the bewildering array of AWS instance types can attest).
Pure can help here by reducing the cost of changing your mind. If you can easily change your mind, mistakes are cheap, and you can afford to try more things before settling on the option that works best for you.
A more flexible FlashBlade helps customers experiment more with high-performance file and object storage, and that takes a combination of both hardware and software to achieve.