Database hardware

Memory Hierarchy(from primary, faster to secondary and slower memory)


Main memory

Flash memory

Non volatile

Flash memory will store data even it is powered off.


Cheaper for a given storage capacity. Used a lot in cameras, phones.

USB(Universal serial bus)

A lot of USB sticks use flash drive.


Solid state drives.


Optical storage(CD, DVD)

Use laser to read.


Can only be read sequentially.

Magnetic disk

magnetic disks

Each magnetic disk will contain several platter. They are flat, circular. Their surfaces are covered with magnetic material. Plates are usually made of glass or metal.

units on disk

Each disk platter is divided into tracks, and then divided into sectors.


Usually a platter contains 50,000 – 100,000 tracks. Inner tracks usually have smaller length(smaller number of sector).


Sectors are usually 512 bytes. Each track usually have 500 to 1000 sectors in the inner track. sectors are the smallest “addressable” unit on the disk

Read-write head:

The motor spins the disk at 60, 90, 120 rps(revolution per second).

How read-write head stores information through reversing the direction of the magnetic field.

The head floats above the disk

The head doesn’t touch the plate. The spinning creates breeze that lift the head.

Head crash

Head crash would remove the storage medium. Further more, the removed medium will fly around, causing more crash.

Disk controller

Interface between the disk and the computer system. It is usually with in the disk drive. It accepts read-write command over a sector. controller also moves heads around.

Remapping of bad sectors.

If controller sees a sector become bad, an attempt to write to this sector will let controller logically map the sector to a different physical sector.

Interfaces to OS

SATA(serial ATA)

ATA is a connection standard from 80’s

SCSI(smaller-computer-system interconnect)

SAS(serial attached SCSI

Fibre channel

SAN(Storage area network

Large numbers of disks are connected by a high speed network. The disks are usually organized using RAID (redundant arrays of independent disks). to give a logical view of a very large disk

NAS(network attached storage

It provides a file system interface using networked file system protocols such as NFS

Optimizing disk-block access

Requests for Disk I/O are from vm manager and file system


A block is a logical unit consisted of fixed number of contiguous sectors. Sometimes page is used to refer to blocks.

Access disks


For successive block numbers. The disk will seek the first block. Then successive request will not require seek. It is a lot faster.

Random access

Just as the name imply


Blocks read are stored in an in-memory buffer for future requests.

Read ahead

When reading a block, read in consecutive blocks from the same track.(great for sequential)


When requesting different block, we might benefit from ordering the block requests.

Elevator algorithm

File organization

We can organize blocks such that it corresponds closely to the way the data will be accessed. OS will often allocate multiple consecutive blocks at a time to a file to increase the chance of sequential write.


A file that have multiple small appends might become fragmented. The system can make a backup copy of the data and restore the entire disk. This restoration will write back the file sequentially to decrease the fragmentation.

Log disk

Use a disk devoted to writing sequential log. Several consecutive blocks can be written at once. Since log disk can do writes later, the DB don’t have to wait for the write to complete.

Leave a Reply

Your email address will not be published. Required fields are marked *

To create code blocks or other preformatted text, indent by four spaces:

    This will be displayed in a monospaced font. The first four 
    spaces will be stripped off, but all other whitespace
    will be preserved.
    Markdown is turned off in code blocks:
     [This is not a link](

To create not a block, but an inline code span, use backticks:

Here is some inline `code`.

For more help see