Having integrated, architected, scrutinized, acquired or dismantled dozens of scale-out architectures in my career as a CTO inside some of the biggest technology companies in the world, I can tell you that scale-out is not new, it is not easy and it is not perfect – but it is valuable.
In concept scale-out is just an incremental way to build a system.
Add as you go, start small and get as big as you want – useful stuff. We have all heard someone, including me right now, espouse the benefits of scale-out.
There is just one problem, delivering on the benefit is really, really hard and incredibly rare. I have worked with many brilliant people – and the problem, as one of my managers use to say, is that the PowerPoint-to-Code Converter is on the fritz.
What typically happens is that in the race to get a product out, time-to-market tradeoffs result in shortcuts that undermine the very scale-out properties they sought to deliver. “We’ll go back and add them back in” is the phrase heard most often. By the time these words are spoken it is too late.
Scale-out is an approach, not a product – if you compromise the approach to get too quickly to a product it is nearly impossible to go back.
In my experience the two most important scale-out properties are:
- The ability to change the properties of the system in response to better information
- The ability to incorporate and exploit new technology
You see, scale-out systems, like any system in data rich environments, are long lived or at least the data is long lived. This means that the system will need to respond to changes in the applications that consume the data or changes in the way the data is used and valued. Service level attributes such as size, performance and reliability will vary over time. If the implementation embraces this flexibility it will allow the customer to extract more value from their data. If not, it will constrain the latent value of the customers data.
Incorporating new technology is critical because, well, there is new technology…
If two years after your system is operational you still have to buy the same components to grow a scale-out system then the system is under-delivering on why scale-out matters. Being able to incorporate the latest technology means the system will naturally improve over time, even for the early tenants. Customers will benefit from Moore’s Law.
Datera got it right. They architected and implemented the system to deliver on the two key scale-out attributes – and more!
By being able to specify what is needed, it frees the consumer from having to know how it works.
Let them be experts in things that matter to their business. By being able to deploy now and change later, when better informed, it accelerates deployment and informed optimization. By managing to an SLO instead of just best-effort, it enables predictable multi-tenant operation. Finally, by being able to incorporate and exploit the latest technology into a live system it delivers on the scale-out promise of a system that gets better – even re-invents itself – over time.
In case you can’t tell, I am really excited to be part of the Datera team!