I just ordered a new T630. It had been a while since I had to order a new server, so I fell back on my old RAID1/RAID5 plan. Despite the fact that drives have come down in price, I could not quite push my client (cost-wise) to RAID10. So now I have OS RAID1 and data RAID5 (three drives).
That is not a bad solution, but reflecting now upon the fact that I ordered drives sized that even RAID 1 would provide plenty of data storage, I wonder if I would be better off with RAID1 + host spare.
And, while I am at it, is there any such thing as three-way RAID1 to allow double-redundancy?
It isn't going to be possible to migrate from a 3 drive Raid 5 to a 2 drive Raid 1. From a Raid 5 you would only be able to migrate to a Raid 0 or 6. The only option you would have to do so would be to backup the data, then delete and recreate the Virtual Disk as the Raid 1, then configure the last drive for the hostspare.
To answer your other questions.
The benefit of a Raid 1 isn't the simplicity of recovery, as both allow for a single drive failure, but in IOPS (input/output operations per second) performance. The table will show you that a Raid 5 is actually double the IOPS of a Raid 1. Now this won't be very noticeable if the server doesn't operate under heavy load. So take that into consideration.
|RAID level||Backend Write IOPS per incoming write request|
You can see the Raid 0 actually has the best performance, but has no redundancy.
I hope this helps clarify everything.
You will have to explain what you mean by a 3 way Raid 1, as I am not certain what you are referring to,
Let me know.
Dell | Social Outreach Services - Enterprise
Get Support on Twitter @DellCaresPro
Download the Dell Quick Resource Locator app today to access PowerEdge support content on your mobile device! (iOS, Android, Windows)
Thank you, Chris.
By 3-way RAID 1, I mean double-redundancy, where there are simply two mirrors instead of one (i.e. three identical drives w/o parity). I have seen various posts indicating it MAY be possible on some systems, but none clearly stating that it IS possible (nothing Dell-specific, for sure). If true, one could still have redundancy after a single drive failure. Thought it was worth asking here.
My reading also seems to indicate that RAID1 recovery time may be faster than RAID5 recovery time, so I guess it is not technically "simplicity of recovery", but perhaps "time for recovery" that I intended to say.
Since I first posted, I have also seen some warnings against using HS with RAID 1 due to the auto-rebuild potentially putting extra load on the 1 good drive before a human can interact and backup (before re-building RAID). Thoughts on that? I am simply stretched too thinly with non-server issues that I will never be an expert on the constantly-improving world of data storage--which is why I post here where, presumably, eyes far more expert than my own are on the issues.
It is just that right now, before I actually configure the server live, is my best opportunity to do any reconfiguration, since the data array has nothing on it as yet.
Robert: thank you for the link. I have use OM in the past, but it has been several years since I added any new servers. The video should help immensely. No data to be lost here, since the data array has nothing on it as yet.
Robert: FYI. I got your post with the link, etc, but when I look at the thread here, your post says only, "Thanks!." Not sure why, but I think you link answered my quest for specific instructions.
No, a 3-disk RAID 1 is not possible on Dell controllers (or any LSI controllers that I'm aware of). You can still get that same level of redundancy, but it will cost you an extra disk: It's a RAID 6. There is also RAID 10.
RAID 6 operates like a RAID 5, but distributes extra parity across the disks to allow for 2-disk failures. Performance is the downside here.
RAID 10 is essentially two mirrors working in tandem to increase performance. It will allow 1 to 2 disks to fail; but it depends on which disks fail. If both disks in the same mirror fail, then the array is down. RAID 10 performance is much higher than RAID 5 or 6 and has become standard in most server storage scenarios because of the balance of speed and redundancy.
Yes, RAID 1 (including RAID 10) rebuild times are lower than RAID 5 build times, and the performance hit the server takes during the rebuild is lower than RAID 5/6 where missing data/parity bits must be calculated for the rebuild.
As far as the load on the remaining drive(s) during a rebuild ... it would not be limited to RAID 1, but this is probably nothing to worry about either. One thing to keep in mind though, especially when working with really large drives, is anytime you have a degraded array, with no protective redundancy, there is no way to error check what is read/written during the process. There are some error checks, but if there are errors during the process, the data being written to the replacement can be compromised, then any discrepancies must be addressed and choices made by the controller when they are encountered. Multi-disk redundancy can be a check for that - just because a RAID 6 can lose two drives doesn't mean you can wait longer before replacing the failed disk; it simply means that constant error-checking is going on even with a failed disk, to protect the array until a disk can be replaced and rebuilt into the array.