Call For Testing BSD Fund

Weak Links and Slow Bison

#CrowdFunding #OpenZFS #ARM #virtualization

December 19th, 2018

Version 1.0

© Michael Dexter

Fifteen years in, I am seeking my first funding which I will use for critical testing of OpenZFS, ARM and hypervisors


I am raising money through BSD Fund in support of three distinct goals:

Just right, reading entirely:

Long before the reality TV series and the questionable drinking theory came along, "the weakest link" and "slowest buffalo" referred to the observations that "a chain is only as strong as its weakest link" and that "a herd of buffalo can only move as fast as the slowest buffalo". In my world, the most glaring weak links and slow bison are the experimental OpenZFS platforms and OpenZFS on ARM. I have demonstrated that testing is the secret to strengthening these weak links and slow creatures, and am asking you to help with this work.

Unsafe at Any Speed, Until Safe

OpenZFS on macOS developer Jörgen Lundman blew our minds with a demonstration of OpenZFS on Windows at the 2017 OpenZFS Developer Summit. Knowing I would meet Jörgen at AsiaBSDCon 2018, I configured my multi-spindle ThinkPad to include Windows 10, and proceded to install his work. This was met with more or less, "Are you insane trying this on real hardware?", and he has been diligently squashing bugs as quickly as I can find them. I have learned how to analyze the core dump that is saved during a blue screen of death and I am slowly building up a series of cross-platform shell scripts to exercise OpenZFS on every supported platform. In nine or so months, OpenZFS on Windows, a.k.a. ZFSin has gone from usable for a few minutes to usable for a few hours and occasionally days. This will not become a matter of weeks, months and years without continued testing and development.

Before I type another word, I invite you to consider becoming a Patreon sponsor of Jörgen (and possibly Michael W. Lucas) because after fifteen years at this, I always think about others first.

That said, I have been slowing building a lab that includes 11 identical Xeon E3 small form factor workstations that are each equipped with two identical drives. These machines allow me to test OpenZFS in parallel on various versions of Windows, NetBSD, GNU/Linux, FreeBSD, FreeNAS and Illumos. I also have a Mac Pro for OpenZFS on macOS testing, though hopefully a Hackintosh E3 will provide truly identical coverage. I have obtained all of these machines out-of-pocket but sincerely would like to spend thirty to 120 minutes a day for three to six months on rigorous testing. Currently I do this work long after I should have gone to bed and while these hours match Jörgen's time zone nicely, plenty of overlap will remain.

"Xeon E3 Power for Far Fewer Watts"

Speaking of Xeon E3's, I will never forget a comment I heard at the annual Supercomputing conference a few years ago regarding the promise that 64-bit ARM CPUs are expected to deliver Xeon E3 performance for a fraction of the power draw. It seems that not emulating a 1980 PC gives the ARM (and every other non-x86) architecture an advantage when the target operating system is not legacy Windows. My lab is also equipped with several ARM boards and I am pleased to report that FreeBSD 12.0 not only supports many of these with pre-built images but also include OpenZFS. I cannot tell if any GNU/Linux distributions for ARM support OpenZFS but it's only a matter of time before they appear. That said...

My PROFOUND thanks to the ARM Limited for donating an OverDrive 1000 64-bit ARM system to my lab!

I am working with Alexandru Elisei and Gigabyte to identify exactly which ThunderX desktop system would accelerate Alexandru's bhyve on ARM work and you'll be first to know what we come up with because we will probably seek your help.

bhyve, Xen, vmm, nvmm, KVM, Hyper-V, VMware, Oh My!

Finally, not everyone has a lab but they really should be testing OpenZFS in order to make it the world's cross-platform file system. There literally are no promising contenders at this time. I have some affordable colleagues who are available to help document how to run every operating system that supports OpenZFS on every available hypervisor. Much of this work is straight-forward but it is critical for further testing each OpenZFS platform in parallel to guarantee that no slow bison get eaten.

The end is within sight for many great things and I want your help realizing them. Having helped others for fifteen years, I am finally brave enough to ask for help for my own work. I know exactly how to spend a few thousand dollars and Euro but am happy to discuss how to invest more money should it become available.


Q: Are you hoping to do this full time?

A: I have heard that dream from many of my colleagues but the single most valuable thing I have done in my career is to provide FreeNAS and OpenZFS support for several years. It critical to see how free software is being used in the wild to minimize the disconnect between developers and users. I think we should all balance our itch scratching with time in the trences and the work I propose here is exactly that.

Q: Is it awkward when your colleagues support you?

A: It can be, but fortunately they are best equipped to explain this type of work to others such as their employers. Spread the word rather than donate! Especially if you have a limited budget.

Q: Shouldn't you be working through a non-profit organization?

A: I have worked with just about every free software funding structure and explain the advantages and disadvantages of "D2D" direct to developer funding in the BSD Fund F.A.Q.. In practice, this often turns out to be developer to developer funding because developers are best equipped to understand the issues at hand.

Q: Who are you?

A: Michael "Rolodexter" Dexter. "He knows everyone" – Thank you Allan Jude for that nickname. I sat down at 4.3BSD Unix for my CS101 course in 1991, a few months before the announcement of Linux, and pretty much can't stop thinking about it. I grew up behind the camera in Hollywood and instantly recognized the "internals" of the computer that Unix represented. To make a long story short, I have done everything in my power to foster free software with an extremely-limited budget.


Copyright © 2018 Michael Dexter unless specified otherwise. Feedback and corrections welcome.