The SAM-SD Model 1
The first real SAM-SD, the one famous for acquiring the name, was nothing but a specification, originally presented in Manhattan at a meeting of the SpiceCorps NYC user group where I presented on the topic of “open storage” and provided an example spec for building a practical, real world storage device.
It was around 2009 and the presentation was in Manhattan. The original idea what talking about how black box storage appliances were nothing more than servers, set up for storage and with dedicated GUIs applied to them and so, because of that, we had the option of acquiring enterprise class hardware, installing an appropriate operating system and building enterprise class storage without the cost and constraints that the storage vendors create.
The specification in this first talk on the subject used the low cost Hewlett-Packard Proliant DL185 G5 high storage capacity server which, at the time, famously was one of the few 2U servers on the market that would not only hold twelve hot swappable 3.5″ LFF hard drives accessible in the front of the chassis, but also made two additional 3.5″ LFF form factor hot swabble bays availabe on the rear of the chassis for a total of fourteen hot swap bays: unheard of in 2009. And the P400 RAID controller that was available with the DL180 G5 had 256MB of cache and would support up to 2TB for each drive, which included both SAS and SATA support making for an extremely flexible system at a low price point and in a very compact chassis. In many ways, the DL185 G5 remains today, nearly a decade later, a high point in storage chassis technology with probably the best cost to storage capability ratio available if we account for its age. The DL185 G5 also had no requirements for using primary OEM vendor drives, but would allow for the use of third party drive trays and drive firmware.
The first SAM-SD specification was for making a server with hardware RAID, hot swappable drives (SAS or SATA, but typically SATA), third party drives and up to 28TB of raw capacity for very low cost and in 2009. Today these numbers seem almost pedestrian, but in 2009 those attending conferences where this was presented were quite surprised that such a feat was possible using products actually on the market. The specification was based on a real fourteen bay Proliant DL185 G5 that had been built to demonstrate the design. One of the strengths of the hardware portion of the design was that any operating system could be used with it such as CentOS, openSuse, FreeBSD or Windows. Flexible to meet nearly any need, very low cost with good performance and very high capacity for the era. The price point was unthinkably low for 2009.
In the specification and presentations was also the idea that building high reliability into the system was a matter of customer choice with options such as DRBD on Linux, which was already mature at the time. The idea that replicated local storage was available, for free, and was mature and enterprise deployment ready was also generally unknown at the time and warranted a lot of rethinking of how storage could be approached.
In the years since, I have spoken to and met many companies that used the original SAM-SD Model 1 specification and implemented it for production storage. I’ve been brought in to configure or support many, with openSuse and DRBD being key components of the majority that I have deployed and, surprisingly, NFS being the principle communications protocol making NAS the more popular use case.
One of the great things about the SAM-SD approach was that it was so flexible and in an era when unified storage handling both NAS and SAN functions at the same time was quite rare, it handled it with ease and was able to work with exotic storage types such as Gluster and CEPH, the later of which would not find popularity for a very long time to come.
The SAM-SD, meaning “Scott Alan Miller Storage Device”, did not come from me and was a quasi-joke from one of the attendees at the original presentation in New York City who, when a day or two later was writing about the presentation, gave that name to the specification and it stuck.
This is not really the beginning of the SAM-SD story, for that we have to dig back several years further and will do so in an upcoming blog post. But the first specification and open storage design presentation in Manhattan circa 2009 really began the idea of community, design and support around open storage in the way that we think of the SAM-SD today and certainly created the first specification of a well known SAM-SD device. For many years, the DL185 was certainly heavily identified with SAM-SD deployments.