On Jun 12, 2014, at 11:27 AM, Warren Young <warren at etr-usa.com> wrote: > [*] The absolute XFS filesystem size limit is about 8 million terabytes, which requires about 500 cubic meters of the densest HDDs available today. I’ve been wondering what 500 TB looks like, so I worked it out. It requires a mere 100 x 6 TB disks for 20% redundancy. Viewed that way, 500 TB looks a little on the low side. You can get a 9U server chassis[*] with its face almost covered with 50 hot-swap 3.5 inch drive trays. That puts us only one size doubling from being able to achieve a max-size array in a single server. Even if we assume SAS drives, we’re still only about 3 doublings away from filling that 9U chassis with a 500 TB array. RHEL7 will be in production 1 level support for another 5 years, enough time for those 3 doublings. I assume we’re climbing out of the doubling doldrums brought on by the Taiwan floods by now. Even if not, we’ve got another *10* years before RHEL 7 leaves production level 3 support. Apparently Red Hat picked this number by doing similar projections, and set it fairly conservatively. What this means is that some of us will be DIYing petabyte scale arrays in a single commodity chassis by the time RHEL 8 ships. I’m not talking about high-dollar SAN or Big Iron stuff here; we’ll be making them from commodity parts you can buy off NewEgg without a special order. Wow. [*] http://goo.gl/IjSdHz