If there is a company that I seriously liked from my last trip to the US it is Purestorage. I always try to be impartial in my writings but sometimes style counts and the guys from Purestorage were smart and disclosed what they are doing very well.
Pure isn’t new to me. I had already met met them for the first time few months ago and I was left with a really good impression. From the strategic point of view they involved Samsung among the investors: The Korean corporation brings them money, flash technology to build the arrays and access to the firmware code of flash devices. These firmwares are modified and optimized to get the maximum advantage from the underlaying (cheap MLC) technology. On the other side, if you look at Pure from a business standpoint, you’ll find the classic well financed Silicon Valley startup, with a lot of money in the bank for early stages of their business development.
The array is still in beta, and not publicly available, but there are some customers that have already useed it in production environments.
The product gestation has been long but the goal is to come out directly with a stable and mature technical proposition. The company is positioning its product at the medium to large enterprise level and they claim a sustained 200K IOps at the maximum configuration. It’s important to notice that Purestorage always talks about real world workloads and not theoretical ones, something that I like very much!
The product is in constant development and it already has all the features that you can expect from a next generation SSD array: data reduction capabilities (thin provisioning, dedupe, compression, etc.), snapshots (redirect-on-write unlimited snapshots that you can schedule every few seconds for each volume!), and replication soon. They are also working on integration with the upper layers like VMware VAAI or Microsoft VSS.
Controllers are now Active/Active while the supported protocols are FC and iSCSI. FCoE will come shortly when customers ask for it and, if I’ve heard correctly, Infiniband is also a possible option.
The overall architecture is simple (based on standard x86 hardware) and it has already been designed for a future scale-out growth. The use of NVRAM on every disk tray is also very interesting for caching and writes optimization, aimed to avoid SAS channels saturation.
Economically speaking, when we go through $/GB, the positioning will be similar to traditional HDD based arrays , but with an enormous advantage in terms of IOps and low latency (indeed, many SSD array vendors are claiming the same). It’s clear to me that now, until they release a price list, we can’t do more than have faith and trust their words.
Probably, one of the most visible flaws (also to Pure representatives) is the lack of a NAS option, but they have decided that it isn’t a top prioritie at the moment.
Well, it’s clear that I like the company and the product. They both seemed to have their feet on the ground, I touched the array with my hands and attended some impressive demo sessions.
Based on the informations I gathered, the product is aligned, with other competitors who have similar ambitions: this means that a good part of the success will derive from good business/technical development and a lot of marketing.
In any case, when the product is available in Europe, it wouldn’t be a bad idea to consider it as a viable option. On the other hand, as for every new startup, it will be very important to carefully evaluate support quality and presence… those two could be very weak points.
Disclaimer: I was invited at this meeting by Condor Consulting Group and they paid for travel and accommodation, I have not been compensated for my time and am not obliged to blog. Furthermore, the content is not reviewed, approved or published by any other person than the Juku’s team.