• Cache
• Main Memory
• Secondary Storage
• Tertiary Storage

• Volatile Storage
• Very Fast, Very Expensive, Limited Capacity
• Hierarchical
• Typical capacities and access times
• Registers - 10 bytes 1 cycle
• L1 - ~10^4 bytes < 5 cycles
• L2 - ~10^5 bytes 5-10 cycles

• Volatile storage
• Fast affordable, medium capacity
• Typical Capacity 10^11 - 10^12 bytes
• Typical access time 10^-8s (20-30 cycles)

• Non Volatyle storage
• Slow, cheap, large capacity
• Typical capacity: 10^11 - 10^12 bytes
• Typical access time: 10^-3s (10^6 cycles)

• Non-volatile storage
• Very slow, very cheap, very large capacity
• Typical capacity: 10^13 - 10^17 bytes
• Typical access time: 10 - 100 seconds

• A - Track
• B - Geometrical Sector
• C - Track Sector
• D - Cluster

[image]

• Tracks closer to the disc edge are longer than those closer to the axis
• Bit densities vary in order to ensure a constant number of bits per sector
• Instead we can vary the number of sectors per track (Depending on track location)
• Improves overall storage density

Access time =

Seek Time + Rotational Delay + Transfer Time

Time taken for head assembly to move to a given track

Avg: 4ms for high end drives; 15ms for mobile devices

Time required for addresed area of disk to rotate into a position where it is accessible by the read/write head.

Transfer time = block size/ transfer rate

Transfer rate ranges from 

• up to 1000 Mbit/sec
• 432 Mbit/sec 12x Blu-Ray disk
• 1.23 Mbits/sec 1x CD
• for SSDs, limited interface. e.f. SATA 3000 Mbit/s

• Sequential i/o is much less expensive than random i/o
• RAM = what about reading "next" block?
• Access time = (block size/transfer rate) + negligible costs
• negligible costs: skip inter-block gap, switch track, switch to adjacent cylinder

• Costs similar to those for reading, unless we wish to verify data
• Verifying requires that we read the block we've just writen, so:

Access Time =

Seek Time +

Rotational Delay (1/2 rotation) +

Transfer Time (for writing) +

Rotational Delay (full rotation) +

Transfer time (For verifying)

1. Read block
2. Modify in memory
3. Write Block
4. Verify Block (optional)

Access Time =

Seek Time +

Rotational Delay (1/2 rotation) +

Transfer Time (for writing) +

Rotational Delay (full writing) +

Transfer time (For verifying) +

[ Rotational delay (full rotation) +

Transfer time (for verifying) ]

• Cylinder-head-sector
• Physical location of data on disk
• ZBR causes problems (sectors vary by tracks)
• Logical Block addressing
• Blocks located by integer index
• HDD firmware maps LBA addresses to physical locations on disk
• Allows remapping of bad blocks

• The size of blocks affects i/o efficiency:
• Big blocks reduce costs of access
• fewer seeks (seek tiem + rotational delay) for the same amount of data
• Big blocks also increase the amount of irrelevant data read
• If you're trying to read a single record in a block, larger blocks force you to read more data

Say a page is accessed every X seconds

CD = cost if we keep that page on disk

• D = cost of disk unit
• I = numbers IOs that unit can perform
• in X seconds, unit can do I*X IOs so:
• CD = D/(X*I)

Say a page is accessed every X seconds

CM = cost if we keep that page on RAM

• M = cost of 1MB of RAM
• P = numbers of pages in 1MB RAM
• CM = M/P

Say page is accessed every X seconds - if CD < CM, then keep page on disk, else keep in memory.

Break even point when CD = CM, or X = (D*P)/(I*M) 

P = 128 pages/MB (8KB pages)

I = 64 accesses/sec/disk

D = 2000 dollars/disk (9GB + controller)

M = 15 dollars/MB of DRAM

X = 266 seconds(about 5 minutes)

• They are either fixed length or variable length
• May also include type of data item (Tells how to interpret, size, etc)
• The object to store (Salary, name, date, etc)
• Stored in bytes
• Numbers can be represented as Integers (2bytes) Real Numbers (IEEE754 floating point), etc
• Characters can be represented with various coding schemes: ASCII, utf-8, etc
• Booleans represented by one byte or one bit
• Dates/Times representation: Integer, ISO8601 dates, etc
• Representation of strings (Null terminated, Length given, Fixed length)

• Collection of related data items/fields (employee record contains name field, salary field, etc)
• Records may have fixed or variable formats and lengths
• Contains record headers
• Data at begining of record that describes it
• Type (points to schema)
• length
• Timestamp
• Intermediate between fixed and variable format

• Fixed Formats - schema describes structure of records
• Number of fields
• Types of fields
• Order in record
• Meaning of each field

[image]

• Schema-less format
• Record itself contains format: "Self-describing"
• Useful for sparse records, repeating fields, evolving formats
• May waste space compared to a fixed format records

[image]