On Sun, May 3, 2020 at 1:44 AM Caveman Al Toraboran <[email protected]> wrote: > > * RAID 1: fails to satisfy points (1) and (3)... > this leaves me with RAID 10
Two things: 1. RAID 10 doesn't satisfy point 1 (read and write performance are identical). No RAID implementation I'm aware of does. 2. Some RAID1 implementations can satisfy point 3 (expandability to additional space and replication multiplicities), particular when combined with LVM. I'd stop and think about your requirements a bit. You seem really concerned about having identical read and write performance. RAID implementations all have their pros in cons both in comparison with each other, in comparison with non-RAID, and in comparison between read and write within any particular RAID implementation. I don't think you should focus so much on whether read=write in your RAID. I'd focus more on whether read and write both meet your requirements. And on that note, what are your requirements? You haven't mentioned what you plan to store on it or how this data will be stored or accessed. It is hard to say whether any design will meet your performance requirements when you haven't provided any, other than a fairly arbitrary read=write one. In general most RAID1 implementations aren't going to lag regular non-RAID disk by much and will often exceed it (especially for reading). I'm not saying RAID1 is the best option for you - I'm just suggesting that you don't toss it out just because it reads faster than it writes, especially in favor of RAID 10 which also reads faster than it writes but has the additional caveat that small writes may necessitate an additional read before write. Not knowing your requirements it is hard to make more specific recommendations but I'd also consider ZFS and distributed filesystems. They have some pros and cons around flexibility and if you're operating at a small scale - it might not be appropriate for your use case, but you should consider them. -- Rich
