INFS2200 Chap.4 Storage Blocks and Access Time
Storage Blocks and Access Time
Logical queries meet physical storage through database blocks. This chapter follows sectors, tracks, surfaces and blocks, then computes capacity, rotation, transfer and average access time with explicit units. Blocking factor and unused space connect record layout to page count.
Each calculation ends with a database consequence, such as locality, sequential access, buffer use or the trade-off created by larger transfer units.
Capacity questions use a multiplicative chain: bytes per sector become bytes per track, tracks become one surface and recording surfaces become the disk total. Access questions use an additive model after every term is converted to one time unit.
Average seek positions the head, average rotational delay brings the target under it, and transfer time moves the block. The assumptions and units are as important as the final number.
Block layout turns hardware arithmetic into database cost. Fixed-length records yield a blocking factor through floor division, leaving unused space when another record cannot fit.
Rows clustered into nearby pages support locality and buffer reuse, while scattered matches can force many reads. This prepares index analysis, where the relevant benefit is the number and pattern of pages avoided.
The worked methods keep quantities separate. Capacity is measured in bytes, access is measured in time, and blocking factor is counted in records per page.
Combining only compatible units makes each step checkable. Every solution ends by identifying the dominant delay or layout trade-off so the result informs a query or storage decision.
Capacity arithmetic should expose every multiplicative factor. Sectors per track multiplied by bytes per sector gives track capacity; tracks per surface and the number of recording surfaces extend the chain.
A double-sided platter contributes two recording surfaces, not one. Report exact bytes first, then state whether any later conversion uses decimal or binary prefixes so a correct method is not obscured by an unstated convention.
Average rotational delay comes from a stated distribution assumption. If the target position is uniformly distributed around a rotating surface, the average wait is half a revolution.
Revolutions per minute must be inverted and converted into a duration before it can be added to seek and transfer time. The resulting classroom model isolates components; it does not claim to include queueing, controller behaviour or every modern device effect.
Record layout connects tuple size to pages.
With fixed-length, non-spanning records, floor division gives the blocking factor and the remainder gives unused space. Allowing records to span blocks changes that interpretation and can require two page reads for one record.
Variable-length records add slot directories and free-space management, so an answer should never apply the simplest formula without naming its layout assumptions.
Locality explains why equal output cardinalities can have unequal I/O costs. Matches packed into a short adjacent run can benefit from sequential transfer and buffer reuse, while one match on each of many pages requires far more page visits.
Estimate the lower packing bound and a scattered upper case before choosing an index or scan. The database consequence follows from blocks touched, access order and expected reuse, not tuple count alone.
What this chapter covers
- 01
Storage hierarchy and blocks
- 02
Track, surface and disk capacity
- 03
Seek and rotational delay
- 04
Block transfer time
- 05
Blocking factor and locality
Worked example · free
Estimate one block access
- +1Convert rpm into time per rotation.
- +1Take one-half rotation for the average delay.
- +1Convert block transfer to milliseconds.
- +1Add the compatible time terms.
- +1Identify the dominant component.
Key terms
- Database block
- The physical unit transferred between storage and memory.
- Seek time
- The time required to position a disk head at the target track.
- Rotational delay
- The wait for the required sector to rotate under the head.
- Blocking factor
- The number of fixed-length records that fit in one block.
Storage Blocks and Access Time FAQ
Why do database costs count blocks?
Storage devices and buffer managers transfer pages rather than isolated logical tuples. Several rows can share one block, while scattered rows can require many block reads even when the result set is small. Page count therefore connects logical predicates to physical work.
How is average rotational delay calculated?
Convert revolutions per minute into time for one revolution, then take one-half under the uniform-position assumption. Convert that duration to the same unit as seek and transfer time. State the assumption because it belongs to the model.
Why does locality matter?
Rows packed into nearby blocks can be read with fewer positioning operations and better buffer reuse. The same row count scattered across many pages can cost substantially more physical I/O. Locality can therefore change the preferred access plan.
Exam move
Separate capacity, time and layout questions. Write the multiplicative capacity chain, the additive access-time relationship and the floor operation for blocking factor. After calculating, state which component or layout property drives the database decision. Practise one complete device problem with bytes, tracks, surfaces, rpm and transfer rate, then change a single parameter and predict the direction before recalculating.
Use an order-of-magnitude estimate to catch missing factors. For layout questions, sketch several records inside a block and compute the remainder explicitly. Finish by explaining whether the result favours sequential access, clustering, caching or a different block choice. The interpretation connects hardware arithmetic to database behaviour.