If you are looking for data protection then go with Raid-6. Yes, Raid-6 can take a two disk hit failure and Raid-10 can only take a single drive hit failure.
Raid-6 creates two parity with each data block and spread it across all the drives and shifts it across the drive set. This method is more algorithm intensive than 10 (striped and mirror).
Raid-10, data is striped across a set number of drives and then it is mirror across to a different set of equal number of drives. So in your case with 24 drives... yes... half will be data and the other half will become the mirror of that data. One set will be the primary and the other is the backup... But no matter how many drives you have, if you loose 1 drive from one set, that set is toast but your data will be running on the mirror set. Once you replace the bad drive in the set that had the hit, that whole set get rebuilt from the mirror. But, during the time that you have a bad drive in one set, you are in a compromised condition in that until you get that drive replace and the set rebuilt, if you so happened to loose a drive in the backup mirror, your whole data is a gonner. Bottomline is that if you loose a drive in either set, that whole set has to be rebuilt before you are back to redundant.
Back to Raid-6, as mentioned, the data and two parity gets laid down across the drive set in a rotational manner... when you have a drive failure, your drive set continues to operate on all its drives less the bad one.... once the bad drive is replace, that drive gets rebuilt using the parity data that is stored on the other drives that belongs to that drive. Since there are two parity data being created, you are able to take a two drive hit....and both drives will be rebuilt by the parity data that is stored on the working drives once the bad drives is replaced. But if you take a third drive hit..... um... the whole kabultoos... is down the drain..
If you got a decent hardware raid .... I'd go for raid-6.
Raid-6 creates two parity with each data block and spread it across all the drives and shifts it across the drive set. This method is more algorithm intensive than 10 (striped and mirror).
Raid-10, data is striped across a set number of drives and then it is mirror across to a different set of equal number of drives. So in your case with 24 drives... yes... half will be data and the other half will become the mirror of that data. One set will be the primary and the other is the backup... But no matter how many drives you have, if you loose 1 drive from one set, that set is toast but your data will be running on the mirror set. Once you replace the bad drive in the set that had the hit, that whole set get rebuilt from the mirror. But, during the time that you have a bad drive in one set, you are in a compromised condition in that until you get that drive replace and the set rebuilt, if you so happened to loose a drive in the backup mirror, your whole data is a gonner. Bottomline is that if you loose a drive in either set, that whole set has to be rebuilt before you are back to redundant.
Back to Raid-6, as mentioned, the data and two parity gets laid down across the drive set in a rotational manner... when you have a drive failure, your drive set continues to operate on all its drives less the bad one.... once the bad drive is replace, that drive gets rebuilt using the parity data that is stored on the other drives that belongs to that drive. Since there are two parity data being created, you are able to take a two drive hit....and both drives will be rebuilt by the parity data that is stored on the working drives once the bad drives is replaced. But if you take a third drive hit..... um... the whole kabultoos... is down the drain..
If you got a decent hardware raid .... I'd go for raid-6.
