Contributed by pitrh on from the stop complaining and do something about it dept.
Now Portland, OR based hardware reseller Gainframe is about to address that problem in the best way thinkable for OpenBSD users. The company has started offering downloadable OpenBSD dmesg(8), pcidump(8) and usbdevs(8) output for every system they build.
To seasoned OpenBSD users the advantages of this approach are obvious. Undeadly sat down with Gainframe CTO Michael Dexter (also known to regular readers as the editor of Callfortesting.org) to hear more about the company and its relationship with the OpenBSD project:
Why are you doing this ?
A healthy BSD marketplace isn't going to build itself and while BSD is hiding in countless products, you have to invert your whole software relationship to make any real progress. Windows-oriented vendors, often with preferential licensing discount schedules will at best dip their toe into Linux and rarely, if ever a BSD; even if you mail them bootable OpenBSD USB keys to test with their white box notebooks, which I've done! With most vendors having a harder and harder time leaving their well-established comfort zones, my comfort zone happens to be in the BSD community and that's where I am starting.
How many different kinds of machines do you build ?
I am focusing on one primary platform that is available as NAS and PRO variants depending on the CPU and if it includes a flash boot module. Both models include 16GB of RAM because given memory modules generally go up in price over time. From there I have some complimentary products and plan to offer two proven ThinkPad models that will include dual-boot instructions and eventually instructions on how to boot Windows in an emulator or hypervisor. I looked into white box PC's but they didn't stand out on quality, price or compatibility.
What did it take to come up with supported configurations ?
The first step is to disregard 95% or so of the available products and carefully research what's left. Intel's recent move to the Ivy Bridge architecture has further narrowed the selection of candidate hardware but top-tier motherboard manufacturers like Intel and Supermicro are offering some very good models. They appear to finally be accepting that that wide choice for choice's sake is not always economical and doesn't make for a better product.
From there I learned how to navigate Intel's bewildering selection of processors and how to obtain their more-interesting low power consumption models. We all have the gaming community to thank for fanatically reviewing certain components such as cooling fans and the rest is a question of diligent research.
What sorts of issues have you run into ?
Well, let's just say I'm not convinced that many manufacturers never actually use the hardware they produce. One motherboard had a capacitor that collided with the large copper CPU heat sink I use and a newer board moved the USB headers to cramped locations that caused flash module installation issues. Quality SLC-based industrial flash modules themselves are a challenge and worst of all, serial headers are now history. Intel offers the "LPC" port for "legacy" interfaces but call it the "LPC Debug" port on newer boards in recognition of the fact that there are very few LCP adapters available. Another unfortunate trend is the move from standard to proprietary CMOS batteries.
What can you tell us about 'the quest for supported hardware' ?
As you well know, the marketplace is a frustrating balance between hardware quality and documentation quality: a great, poorly-documented product might never be usable in a free software environment and poor-quality alternative should probably not have been manufactured in the first place. Good hardware does exist but you have to work hard to find it and above all, to find a stable supply of it. The community is the best resource for recognizing good hardware because of the sheer diversity of products they have access to. The community is also where virtually all drivers and other forms of hardware support come from. I am working with several OpenBSD and FreeBSD developers to resolve some minor hardware issues and fortunately, resolving an issue in one BSD makes it far easier on the others. In one case, the variables have different names but the problem is identical.
Have you spoken with any hardware vendors about this ?
Absolutely. You have to or the situation will not get any better. One flash storage vendor wants sales projections to fix a specification compliance bug and slim down their form factor. Another will hopefully bump up a wire gauge on a fan. At some point however you have to take charge: I have a design for a more robust USB-A to header adapter for programming flash memory modules and am working with an architect to design a case that is rack-friendly and does away with proprietary drive trays. A device to not only daisy chain but create failover redundancy with two standard power supplies would be nice but I suspect the engineering and manufacturing costs would be prohibitively high. There's never been a better time to design new hardware but you also have to know when to compromise.
How can the community help ?
If anything, they can simply keep up the good work. As I mentioned, it is great to know what hardware stands out in terms of quality and driver support because I will gladly investigate sources of it. I am currently working with a developer and an organization to get a ThinkPad x230 to the upcoming hackathon and hope to eventually have a budget for strategic software development. I certainly also welcome people to browse my catalog and I am happy to quote special items from the channel distributors.
In the big picture, I would love to offer a selection of OpenBSD-based appliances but as you know, relatively few people realize that OpenBSD itself is an enlightened appliance. I plan to investigate related projects like nsh and flashrd but it's hard to gauge how proven they are or how active a community surrounds them.
(Comments are closed)