IMPLEMENTING FAULT TOLERENCE IN WINDOWS NT
HOW TO IMPLEMENT FAULT TOLERANCE IN WINDOWS NT.A vital tool for protecting a network's data is the use of a REDUNDANT ARRAY OF INEXPENSIVE DISKS (RAID).
A vital tool for protecting a network's data is the use of a REDUNDANT ARRAY OF INEXPENSIVE DISKS (RAID). Using a RAID system enables you to set up the best disk array design to protect your system. A RAID system combines two or more disks to create a large virtual disk structure that enables you to store redundant copies of the data. In a disk array, the drives are coordinated into different levels of RAID, to which the controller card distributes the data. RAID uses a format of splitting data among drives at the bit, byte or block level, the term data striping refers to the capability of arranging data in different sequences across drives. According to meet your user's requirements you need to implement the best level of RAID. RAID comes in seven levels; they are 0,1,2,3,4,5,6 and 10.Windows NT supports 3 levels they are RAID 0,RAID 1,and RAID 5.
Let us see what these levels does for us.
Stripe Set is created by combining free spaces over two or more hard disks. RAID 0 uses data striping and block interleaving, a process that involves distributing the data block by block across the disk array in the same location across each disk. Data can be read or written to these same sectors from either disks, thus improving performance. To implement RAID 0 you need at least 2 Hard disks and important point to be noted that the striped partitions is to be of same size. RAID 0 will not provide redundancy of data; means that the failure of any one disk can bring down the entire system and loss of data contained in it. This level is supported by WINDOWS NT Server and WINDOWS NT Workstation and this will not provide any fault tolerance mind you. But it provide faster disk performance.
Note: The System Partition and Boot Partition of the system cannot be a part of stripe set. You can combine areas from different types of hard disks that included IDE, SCSI, ESDI.
In RAID 1 the disks are mirrored. Each data which is written into one disk is written to another also at same time. It can be defined as two hardisks; one primary and one secondary that use same controller and cable. RAID 1 can also be called as disk duplexing when you add separate controller for each drive. Disk duplexing is more reliable than disk mirroring. I'll tell you about this later. Disk mirroring will not provide much performance because it need to write data two times into the disk. The advantage of disk mirroring is it copies the boot loader also. So if one drive in the pair fails to boot the another can take control and continue to operate. It is a bit expensive too. Suppose you have two 10GB hard disk you can use only 10GB because the second disk is used as mirrored disk. Windows NT will support disk mirroring but Windows NT workstation will not provide. disk mirroring.
Note: Even if you create another partition of same size on the same hard disk, you cannot create a mirror set. It requires two separate hard disks. With disk mirroring you cannot recover the data if more than one disk fails. In such a situation, you have to use the emergency boot disk and boot from the mirrored disk and mirror the information on mirrored disk to a new disk. Disk mirroring system runs if the original disk fails. However if the hard disk controller fails, disk mirroring cannot continue the operation. In such a case you have to replace the controller and then the system will be back to normal. System Partition and Boot Partition can be a part of disk mirroring. You cannot implement disk mirroring using different types of hard disk. .
Stripe Set with Parity
Stripe set with parity is the enhanced version of stripe set. It works in a similar fashion an it provides a great degree of fault tolerance. RAID 5 uses striping with parity information written across multiple drives to enable fault tolerance with a minimum of wasted disk space. Striping with parity is based on the principle that all data is written to the hard drive is binary code. To implement RAID 5 you need at least 3 drives because the technique writes data across two of them and then creates the parity block on the third. This process would continues as the next parity bit is written to the first drive, and the data to the second and third. On the third round the parity is written to the second disk and data to the third and first disk. This cycle will go on. Suppose if any one disk fails, the data can be retained from the data an parity bits on the other two disks. You can connect up to 32 disk in WINDOWS NT to implement RAID 5.
Note: If more than one disk fails you cannot recover the data. In such a situation, you have to recreate a stripe set with parity and restore backups. The system partition and boot partition cannot be a part of stripe set with parity. You can create stripe set with parity and format it as FAT or NTFS..