#### Storage stuff

### fundamental terminology

* sector			= smallest unit of storage, usually 512 bytes; some disks have 4096 bytes
* multisector transfer		= ability of disk to parallelize sector transfers in single operation

* chunk, stride, stripe unit	= segment of sequential data written/read from a disk before moving to next disk
				  aka chunk size, stride sie, stripe size, stripe depth, or stripe length
				  large files should have smaller chunks for more parallel I/O; CONFIRM
				  small files should have larger chunks; CONFIRM
* strip, stripe			= logical formation over array for single striped operations
* stripe width			= number of data disks in array
* wide striping			= chunks of data spread across multiple arrays, maybe all disks in system
* narrow striping		= chunks of data spread across drives in a single array
* superblock			= a record of the file system characteristics: size, block size, empty/filled blocks, size and location of inode tables, etc.

### RAID levels

* level 0 			= stipe set/striped volume; splits data evenly across 2 or more disks without parity. performance boost
* level 1 			= mirror; 2 drives
* level 2 			= stripes at bit rather than block; hammering code for error correction.  defunct
* level 3 			= byte level striping with parity disk; best for applications with high transfer rates in long, sequential reads/writes; 3 drives
* level 4 			= block level striping with parity disk; rare
* level 5 			= block level striping with parity distributed across all disks; can tolerate single disk failure; 3 drives
* level 6 			= extends level 5 by an additional parity block; can tolerate two disk failure; 4 drives
* level 10 			= (1+0 mirror over stripes; 0+1 stripe over mirrors); 4 drives
* level 50			= (5+0); 6 drives 
* level 60			= (6+0); 8 drives

* read performance		= for levels 0, 5, 6, and 10 figure is # of disks in array (-1 for r5 and -2 for r6)


### RAID performance

* Write penalty			= Associated with stripe count(s) and/or parity levels 
				= RAID 5 is 4 (read data, read parity, write data, write parity) 
				= RAID 6 is 6 (read data, read parity 1 & 2, write data , write parity 1  & 2)
				
# Standard IOPS calculation	= NX/P; Number of disks/spindles * IOPS rate/penalty; 8 disk RAID5 write penalty 50% IOPS ((8*125)/4).5
* IOPS types			= Read IOPS (RIOPS), Write IOPS (WIOPS), Blended IOPS (mixture of read and write IOPS)
* calculate BIOPS		= (read iops * 2) + (write iops * 2); 2 is the percentage, 50/50

* level 0			= No penalties
* level 1			= Write penalty of 2 (write twice)
* level 5 			= Write penalty of 4
* level 6			= Write penalty of 6
* level 10			= Write penalty of 1 but read performance of 2

# Capacity performance		= (R(+Parity)X)/Penalty; RAID5 is 1 parity, RAID6 is 2 parity

* Determine spindle count for raw capacity as R
* Each drive IOPS is represented as X
* Calculate as normal