Secondary Storage - General Knowledge

Secondary Storage A BRIEF REVIEW OF HARD DISK DRIVES AND STORAGE TECHNOLOGIES

Medical Infromation Management - Autumn 2013 2

Internal Structure


Internal Structure (Cont.)

• Head flies above platters

• Platter are divided into circular tracks and tracks which are subdivided into sectors. The set of tracks at one arm position make cylinder.

• Logical blocks, the smallest

unit of transfer (512 bytes)

that maps to the sectors.


Internal Structure (Speed)

• Disk speed has two parts:• Transfer Rate, the rate at each data flow between the drive and computer.

• Efficient Transfer Rate• Position Time (Random-Access Time),

• Seek time, the time necessary to move disk arm to the desired cylinder.• Rotational Latency, time necessary for the desired sector to rotate to disk head.


Internal Structure (Head Crash - Connection)

• Head Crash, head may contact the surface

• I/O bus connects Disk Drive to the computer.• EIDE• ATA, PATA, SATA• USB• FC• SCSI• FireWire! (Developed by Apple, IEEE 1934 standard) (400 Mbps)

• A disk controller is built into each disk drive that has a cache ...

SATA

SCSI



Magnetic Tapes

• Early secondary storages with very slow access time (1000 times slower than HDD)

• Can be used for back up or non-frequently used data.


Disk Attachment

• Host-Attachment Storage (via I/O) common for small systems.

• Network-Attach Storage, remote host in a distributed file system• Remote-procedure call interface (NFS for UNIX, CIFS for windows)• NAS is implemented as a RAID (redundant array of independent disks) array

with software that implements RPC interface.


NAS


Server


Storage Area Network

• Drawback in NAS: Storage I/O operations consume Bandwith


SAN

• Oracle

• Google

• Hp

• …


RAID

• Disks get smaller and cheaper

• Redundant Arrays of Independent (not completely) Disks

• Inexpensive -> Independent

• Increasing read and write rate (Parallel) - increase the throughput (load balancing) / reduce response time of large accesses

• Increasing Reliability (Redundant) – Mirroring

• RAID levels (Redundancy to Striping)


RAID Levels – 0

• Nothing


RAID Levels – 1

• Mirroring


RAID Levels – 2

• Parity bits - ECC


RAID Levels – 3

• Parity bits

• Less disks

• Much Speed

• fewer I/O

• Dedicated Parity Hardware


RAID Levels – 4

• Like 3 but blocks in each disk

• Higher level of I/O rate


RAID Levels – 5

• Spreading the parity blocks

• Safer - Most Common


RAID Levels – 6

• Like 5, Parity + Reed Solomon ECC Code


RAID Levels – 0+1 , 1+0


Choosing a RAID Level

• Continues supply of data is needed• Rebuilding is easiest in RAID level 1

• Level 0 for high performance where data loss is not so important

• Level 1+0 and 0+1 for both Performance and reliability (ex. Small Databases)

• Level 5 can be used instead of 1

• Level 6 is not supported commonly, but it should be more reliable that level 5


WHAT SHOULD WE THINK ABOUT?

• How many disks should be in a given RAID set? • More disks, More Data-Transfer Rate,

More Expensive

• How many bits should be protected by each parity bit?• Less bits each parity, More Chance to

Modify the Failure, More Overhead


SSD VS. HDDSolid State Drive

Hard Disk Drive


THE END

Secondary Storage - General Knowledge

Technology

raid disks

disk head

disk arm

disk controller

time necessary

storage io operations

storage technologies

io common