Mass Storage Structure - Lecture Slides | COMP 310, Study notes of Operating Systems

Download Mass Storage Structure - Lecture Slides | COMP 310 and more Study notes Operating Systems in PDF only on Docsity!

COMP 310:

Operating Systems

Lecture 27:

Mass Storage Structure

November 29, 2004

Christine Alvarado

Overview

 Disk Management

 RAID

 Back-up storage

Review

 What does the disk look like?

Using the Disk

 Formatting

 Booting from disk

 Dealing with Bad Blocks

Logical Formatting

 After low-level formatting, OS must record its

own disk management data structure on disk

 Partitioning:

 Logical Formatting

Boot Block

 Computer needs code to start running when

it is powered up

Bad Blocks

 Disks are flakey—some even come from factory

with bad blocks

 Handling bad blocks:

Swap-Space Management

 Swap-space size

 Must be determined in advance

 Better for overestimate size (Why?)

 Swap-space location

 2 options:

Reliability via Redundancy

 Mean time to failure—average time until a

specific disk fails

 Say mttf for a disk is 100,000 hours

 If we have 100 disks…

Solution: Redundancy

Performance via Parallelism

 Reliability slows thing down… but other

things can be done to speed it up.

 Bit striping:

 write each byte across 8 disks (1 bit each)

 Block striping:

 Blocks striped across disks

 Why does this help?

RAID in general

Common RAID “Levels”

 Reliability vs. Performance: Different “levels”

perform different tasks

 RAID Level 0: Bit-level striping  Good for random access, no redundancy  RAID 1: Mirroring  Two disks, write data to both (expensive)  RAID 3: Bit parity  Use bit parity to recover from disk failure  RAID 5: Floating parity  Parity blocks for different stripes written to different disks  No single parity disk, hence no bottleneck  RAID “ 10 ” (1+0): Striping plus mirroring  Higher bandwidth, but still large overhead

Assignment

 Reading for Wednesday:

 A Case for RAID

 Reading for Friday: The Google Filesystem