should i use raid 5 or raid 10 for a data server?
we are getting a new dataserver for an engineering office. it will have approximately 70 users accessing the data on it. it will be our only data server and i am wondering which raid setup is better. the datasets consist of word and excel spreadsheets, cad drawings and some large gis projects.
ASKER CERTIFIED SOLUTION
membership
This solution is only available to members.
To access this solution, you must be a member of Experts Exchange.
No. Use RAID6 for your data and RAID1 for your O/S, swap, and scratch table space (TempDB). You sure as heck better have a decent caching controller as well and stay away from SATA drives if you like performance.
You can use the RAID6 for local backup of the RAID1. Do NOT confuse backup with archiving. You need to always make offline copies of data for protection against catastrophic failure / fire, etc...
You can use the RAID6 for local backup of the RAID1. Do NOT confuse backup with archiving. You need to always make offline copies of data for protection against catastrophic failure / fire, etc...
Raid 10, Raid 10, Raid 10.
Raid 5 and 6, upon reintroducing a replacement drive for a failed drive will rewrite ALL of the data across the spanned disks. since all of your disks in an array were purchased at the same time they will all reach MTBM at the same time. the extra stress placed on the drives during a rebuild has a very good chance of revealing the next weakest drive by failing during a rebuild.
Raid 0 is the fastest disk configuration for both reads and writes. Raid 1 is the safest way to protect your data.
Raid 10 is the perfect marriage. - my standard setup for a server with 8-12 drives is to configure raid 10 for all drives, partition the drive for the host OS HyperVisor, use the rest of the drives for OS VHD's and Data VHD's.
Attach a cheap NAS (that does raid 10) for backups.
Raid 5 and 6, upon reintroducing a replacement drive for a failed drive will rewrite ALL of the data across the spanned disks. since all of your disks in an array were purchased at the same time they will all reach MTBM at the same time. the extra stress placed on the drives during a rebuild has a very good chance of revealing the next weakest drive by failing during a rebuild.
Raid 0 is the fastest disk configuration for both reads and writes. Raid 1 is the safest way to protect your data.
Raid 10 is the perfect marriage. - my standard setup for a server with 8-12 drives is to configure raid 10 for all drives, partition the drive for the host OS HyperVisor, use the rest of the drives for OS VHD's and Data VHD's.
Attach a cheap NAS (that does raid 10) for backups.
RAID0 is not faster than RAID10 for reads by any stretch of the imagination. With RAID10 every byte of data is in two places, so in perfect world you get twice the read performance due to load balancing. When it comes to writes, then both writes are done in parallel. If you have a caching controller then there is little difference in the real world.
Now if you have a data logging app where nearly 100% of the data is writes then that is different.
Now if you have a data logging app where nearly 100% of the data is writes then that is different.
Additional drives (spindles) can increase performance, so multiple raid1 pairs for transaction logs and the like also helps. If going solid state, stick to reputable makes and their enterprise ranges.
Avoid using raid on motherboard, as implementations can be poor.