It’s been a while since I posted about my unhappiness with NetApp’s FAS2020. And while we’ve replaced our 2020 with a 2050, I still get a lot of traffic to those pages, and a few emailed questions from prospective or new 2020 owners asking about the so-called 8TB limit. Here’s a summary of what I know:
- The aggregate limit appears to actually be 7TB, not 8TB as is commonly stated. I’m not sure what the source of this confusion is, but error messages on the 2020 and NetApp’s docs on NOW seem to bear this out.
- If you have 1TB drives, your maximum aggregate size is 4.36TB from a single DP RAID group of 8 drives. This is before aggregate and volume snap reserves, which you can adjust to fit your needs, but are configured by default to be five and twenty percent, respectively.
- If you have 750GB drives, your maximum aggregate size is 4.91TB, from a single DP RAID group of 11 drives - again, before aggregate and volume snap reserves.
- If you have other drives that you’ll be using, you can calculate the maximum aggregate size based on this NOW document and drive Rightsizes.
- The FAS2020 currently supports four shelves of fourteen drives each, plus twelve internal drives for a total of sixty-eight drives.
- I never got to the point of pushing our 2020’s performance limits while we had it, but in general, the more spindles you have, the faster. I suspect that the 2020’s performance might be hamstrung by it’s limited NVRAM (128MB with a single controller), but I can’t confirm that. Our 2050 has performed adequately as second-tier storage for two 3040s.
Updated: Fixed my math on the maximum aggregate sizes - I left out the 10% WAFL overhead!
2 Comments
Thanks a million Andrew, I am a potential buyer so I could not access this NOW document yet. Please let me make sure I get this right
- When I use a 8×1TB drives, the aggregate limit is 4.85TB, does this mean that the maximum filesystem size I get is that (75% * 4.85TB) ? Is that aggregate like Linux’s “volume group” ?
- When I install a new shelf with similar 8×1TB, I get “another” filesystem of the same size, so the limitation in reality is that those 2 file systems are separate and not one continuous storage space ?
- If I fill the internal 12×1TB disks, doesn’t that give the maximum aggregate size limit?
- How much did the 2050 raise the limits
Thanks for all the help
Hi Ahmed,
Please note that I left out the 10% WAFL overhead initially - the figures above are updated to be correct.
- An aggregate is similar to a volume group under Linux in that it contains the actual volumes that you use (FlexVols). With the 8×1TB drives, you’d get an aggregate of 4.36TB; if you left the aggregate and volume snapshot reserves at their default, you could make one volume of 3.31TB; however, you could set both reserves to zero - just understand the drawbacks of this before you do it - and create a 4.36TB FlexVol. (You could also, of course, create multiple smaller FlexVols in one containing aggregate.)
- You are correct that the limitation is that you would not have continuous storage space when you create a second aggregate - and that you’re spending more money on parity drives than you would with a Filer that could handle larger aggregates.
- With 12×1TB, you could create an aggregate using up to 8 disks in one RAID group; you could use the four leftover drives for spares, but it would be exceedingly inefficient to try to create a usable volume from them, particularly if you try to create a RAID-DP group with them.
-The 2050 has the same limits as the rest of the NetApp FAS product line - for instance, you can make an 11.4TB aggregate out of 26 750GB drives (two RAID-DP groups, so four drives are used for parity), after WAFL overhead and with the default 5% aggregate snapshot on.
Post a Comment