Storage by hand
One of my favorite books of all time is Paul Bertolli’s “Cooking by hand”. Paul, former Executive chef at Oliveto in Oakland, CA and Chez Panisse prior to that, wrote what I consider the best technical cookbook out there.
The thing that sets Cooking by Hand apart from all the other cooking books on my shelf (and there are a fair number of them) is the importance laid on the picking the right ingredients and the science that brings those ingredients together. You are taught to make your own pasta and sausage ab initio — from first principles. As you learn the art of making these fundamental things, the highest quality pasta, sausage, cheeses and other ingredients, you also learn to be creative and imaginative in your recipes… Bertolli wants you to learn to combine the science and technology of cooking with the art.
I believe that this same care and passion has a strong place in enterprise IT, and that it is particularly necessary in architecting storage solutions. Today’s storage market is awash with offerings, new and old, in both software and physical appliance form factors, all of which seem to have little differentiation from each other. The old, established systems are based on software that has been around for 30 years in many cases, designed entirely around the idea of cabinets full of spinning disks. Meanwhile, many of the newer systems have been slapped together using off-the-shelf file systems that are burdened with similar assumptions.
In storage today, there doesn’t seem to be a proper appreciation for the fact that the ingredients have fundamentally changed. These new ingredients, things like PCIe Flash, programmable networks, and the increasing density of datacenter environments, are so different than anything we’ve seen before that they demand entirely new recipes for storage — not just the peppering of the same old thing with a few SSDs. When my co-founder (and a pretty darn good cook), Andy Warfield and his team started working with PCIe flash, they made a surprising realization about performance: these new flash memories are incredibly demanding. Andy likes to say that PCIe flash is problematically fast, he observes that in order to take maximum advantage of flash as a storage device you have to be creative with the rest of the ingredients, especially the cpu and the network. Taking something so problematically fast and packaging it up into a traditional array form factor made no sense, because piling up a mountain of flash behind a small amount of CPU and connectivity, as has always been the approach with disks, is a failure to recognize it’s capabilities as a storage medium.
The Coho Data team has re-architected enterprise storage ab initio. Andy and the team have built a system that respects these new ingredients. It is based entirely on commodity components, because we know that the commodity hardware available to enterprise environments has, and continues to change and evolve at an astounding pace. Our team has written a software stack that unifies these commodity parts. It is optimized for PCIe flash and leverages software-defined networking in a unique and differentiated manner. The result, I am proud to announce, is the Coho DataStream 1000, a scale-out storage system that you can evolve over time, and one that beats every competitive solution out there on a price-performance basis. If Paul did storage, I like to think that this is the sort of thing that he might have come up with.
It also makes perfect sense that Frank Artale, the person who introduced me to Paul Bertolli’s book, should join Coho’s Board of Directors. There are few in the IT world who understand the challenges and opportunities facing enterprise systems as well as Frank. I’ve had the pleasure of working with him for fifteen years now and can confidently say Frank knows and cares passionately about building things properly. From scratch.
3,541 total views, 4 views today