D-Link Forums
The Graveyard - Products No Longer Supported => D-Link Storage => DNS-343 => Topic started by: treestump on January 21, 2009, 11:17:35 AM
-
My original 343 went into an infinite reboot mode so I have an RMA replacement here in front of me (currently climatizing from it's chilly journey).
I cannot risk the data that is on this device, will I be able to just plug the drives into the new unit and it will recognize the existing RAID5 array? I'm afraid of having the DNS-343 automatically build a new array over my existing data.
Thoughts?
-
I don't know if it will recognize the RAID, but if it doesn't it won't reformat it or anything. When you log into the admin, it will tell you it sees new drives and *ask* you to format it again.
Make sure you insert the drives in the same order. Although linux RAIDs usually can recover from out of order RAIDs, it's best not to tempt fate. I usually stick a label on my drives to know the order.
If the RAID is not recognized, "skip" and then go to Maintenance/System/Shutdown.
-
Thanks. After your post, I just plugged them in (I too record the numbers) and the RAID setup was just as if it was the old unit.
Now if only I had taken a backup of the settings, then I wouldn't be scratching my head while trying to recreate all the media server/ftp users & groups...
-
Whilst you're about it - consider backing up the data too - what would you have done if the new unit prompted you to format the drives?
-
Whilst you're about it - consider backing up the data too - what would you have done if the new unit prompted you to format the drives?
That's a good point, but kind of negates the point of me going RAID5.
What is done in the enterprise world?
-
That's a good point, but kind of negates the point of me going RAID5.
What is done in the enterprise world?
In the enterprise world it depends on the usage of the array.
For a transactional (IO intensive) purpose such as a large database, then it would be a bad idea to go with RAID 5 because you'll take a hit in performance. Therefore you would want to use RAID10 so you have a strip and a mirror for data security.
For a file server in enterprise it would be fine using RAID 5 array.
http://www.miracleas.com/BAARF/RAID5_versus_RAID10.txt
Here is some of the calculations from some research that has been going on at my work place on IO performance.
IOPS = (1 / [Rotation Latency + Average Seektime]) x 1000
A 7200upm drive has a rotation latency of round about 4m. Lets assume a average seek time of 5ms.
IOPS = (1 / [4 ms + 5 ms]) x 1000
IOPS = (1 / 9) x 1000
IOPS = 0,11 x 1000
IOPS = 110
But real world test seem to put the drives at about 65 IOPs.
P is the performance of a spindle and N is the number of spindles that comprise the RAID set.
For RAID 1 or 0+1, or 10,
Read performance = P * N
Write performance = P * (N/2)
For RAID 5,
Read performance = P * (N-1)
Write performance = P * (N-1)/N
So;
A 3 disk raid 5 config;
Reads=65*(3-1)=130
Writes=65*((3-1)/3)=43.3
A 4 disk raid 5 config;
Reads=65*(4-1)=195
Writes=65*((4-1)/4)=48.75
A 4 disk raid 10 config;
Reads=65*4=260
Writes=65*(4/2)=130
-
Excellent info NeoNZ!
treestump,
RAID is not a replacement for a backup; the intention of RAID is to provide 'real-time' redundancy. Most enterprise environments use RAID configurations to ensure that IT personnel have sufficient time to address HDD related failures.
Backups are typically performed on a daily basis; in some organizations it's 'real-time' and in others, weekly - in short, it depends on the requirements.
Additionally, keep in mind that backups are usually rotated; this ensures that should corrupted data be discovered on the RAID (HDD) or previous backup, IT personnel can go back to previous backup(s) in the hopes of recovering a non-corrupted version of the data.
HTH,
-
That's a good point, but kind of negates the point of me going RAID5.
What is done in the enterprise world?
Take a look at my signature - where it says RAID1, think RAID5 - RAID5 is not a backup, it's about disk redundancy, if a disk fails, your data will continue to be available - it protects you from data loss due to failure of a single disk ONLY.
You also need protection from data loss in cases of equipment failure (as happened to you), deletion of data - whether accidental or deliberate, corruption of data due to network failure or virus, etc.
Q1 - what happens to your data when one disk fails?
A1 - it should continue to remain available.
Q2 - what happens to your data if a second disk fails before the first one is replaced and the rebuild completed?
A2 - the data is gone - completely & permanently.
What is done in the enterprise world? Enterprise administrators who value their jobs back up their data.