OS LU8

Device Management

• The operating system manages devices within a computer system.
• Devices include monitors, applications, disk drives, keyboards, mice, and printers.

Learning Objectives

• Describe features of dedicated, shared, and virtual devices.
• Explain differences between sequential and direct access media.
• Understand concepts of blocking and buffering and their impact on I/O performance.
• Understand the roles of seek time, search time, and transfer time in calculating access time.
• Explain the differences in access times in several types of devices.
• Describes critical components of the input/output subsystem and how they interact.
• Discuss the strengths and weaknesses of common seek strategies like FCFS, SSTF, SCAN/LOOK, and C-SCAN/C-LOOK.
• Explain what different levels of RAID are and what sets each apart from the others

Types of Devices

• Types of devices include dedicated, shared, and virtual
• Storage media can be either sequential access or direct access
• Device management is about optimizing the roles of each device by the operating system

Dedicated Devices

• Assigned to one job at a time and are active for the entire job duration.
• Tape drives, printers, and plotters can be examples of dedicated devices.
• A disadvantage is the inefficiency if the device isn't used 100% of the time as it is allocated for the job's entire execution.

Shared Devices

• Assigned to several processes, as exemplified by a Direct Access Storage Device (DASD).
• Processes can share a DASD simultaneously and feature interleaved requests.
• Management involves device manager supervision, controls interleaving, and predetermined policies to resolve conflicts.

Virtual Devices

• These include combinations of both dedicated and shared devices.
• Dedicated devices may be transformed into a shared device.
• A printer being converted by spooling program is an example.
• Spooling speeds up slow dedicated I/O devices.
• Universal Serial Bus (USB) controllers act as interfaces between the operating system, device drivers, applications, and devices attached via the USB host.

Storage Media

• Sequential access media records information sequentially
• Direct access storage devices (DASD) records information sequentially or uses direct access files.
• Significant differences exist, especially in speed and sharing capabilities.

Sequential Access Storage Media

• Magnetic tape was in early computer systems for routine secondary storage
• Used today for routine archiving and data backup.
• Magnetic tape records are stored serially.
• The length of records is determined by the application program and records are identified by their position on the tape.
• Access involves tape mounts and forwarding to the needed record; time-consuming process.
• Tape density is the number of characters recorded per inch, dependent on storage method: individual vs. blocked.
• Tape reading/writing employs mechanics that move the tape when the head is ready.
• Interrecord gaps (IRG) is a 1/2 inch gap inserted between each record and same size regardless of records it separates.
• Blocking groups records into blocks.
• Transfer Rate = (tape density) x (transport speed).
• Interblock gap (IBG) is a 1/2 inch gap inserted between each record and is more efficient than individual records and IRG.
• There are fewer I/O operations needed and less wasted tape when using the blocking advantages
• Disadvantages include overhead and software routines needed for blocking and deblocking
• Also buffer space may be wasted.
• Advantages include low cost, compact storage capabilities, good for magnetic disk backup and long-term archival
• Disadvantages include access time being poor for routine secondary storage and interactive applications.

Direct Access Storage Devices (DASD)

• DASDs are read or written to a specific disk area and is random access
• Four categories: magnetic disks, optical discs, flash memory, magneto-optical disks
• Access time variance isn't as wide as magnetic tape.
• Record location directly affects access time.

Fixed-Head Magnetic Disk Storage

• Resembles a large CD or DVD covered with magnetic film.
• Usually formatted on both sides in concentric circles called tracks.
• Data are recorded serially on each track with a fixed read/write head positioned over the data.
• A fixed-head disk is faster than movable head.
• Disadvantages include a higher cost and reduced storage.

Movable-Head Magnetic Disk Storage

• One read/write head floats over the disk surface
• An example of this includes computer hard drives with their disks including single platters or part of a disk pack
• Access requirements include cylinder, surface, and record number.
• Generally disk pack platters have two recording surfaces, except for top and bottom platters.
• Surfaces are formatted with concentric tracks.
• Track number varies with 1000+ for high-capacity disks.
• Track zero is the outermost concentric circle.
• The center contains the highest-numbered track.
• All heads move in unison
• Filling approaches include surface-by-surface that is slower while track-by-track is faster.
• The virtual cylinder approach involves filling track zero

Optical Disc Storage

• Optical disks have a different design compared to magnetic disks
• Magnetic disks have concentric tracks and sectors and spins at a constant angular velocity (CAV), which wastes storage space but provides fast data retrieval.

Optical Disc

• Optical disks have a spiral track of same-sized sectors from center to disc rim.
• Rotate at constant linear velocity (CLV).
• More sectors and more disk data.
• Two performance metrics: sustained data-transfer rate and average access time.
• Sustained data-transfer rate is speed of reading massive data amounts from disc measured in megabytes per second (Mbps)
• This is crucial for sequential access applications.
• Average access time is the time to love head to a specific disc location and is expressed in milliseconds (ms).
• A third feature is cache size which includes a buffer to transfer data blocks from disk

CD and DVD Technology

• In CDs data is recorded as zeros and ones.
• Zeros are pits and ones are lands.
• Reads with a low-power laser.
• Light strikes land and reflects to photodetector or when a pit is scattered and absorbed.
• The photodetector converts light intensity into a digital signal.
• CD-Recordable (CD-R) technology requires an expensive disk controller and records data using a write-once technique.
• The disk contains several layers, including a gold reflective and dye layer.
• It records with high-power lasers with permanent marks on the dye layer.
• It cannot be erased after data is recorded and is read on a standard CD drive (low-power beam).
• CD-Rewritable (CD-RW) technology includes data written, changed, and erased
• Features: phase change technology, with amorphous and crystalline phase states.
• Data Recording: beam heats up the disc which changes state from crystalline to amorphous to record data
• Erasing is a use of low-energy beam to heat up pits to return the alloy to original crystalline state.
• Drives read standard CD-ROM, CD-R, CD-RW discs.
• Drives store large quantities of data, sound, graphics, multimedia.
• DVD (Digital Versatile Disc) is similar to CD-ROM in design, shape, and size.
• Differs in data capacity.
• Dual-layer, single-sided DVD holds 13 CDs; Single-layer, single-sided DVD holds 8.6 GB (MPEG video compression).
• There are differences in laser wavelengths. Uses red laser (smaller pits, tighter spiral).
• DVDs cannot be read by CD or CD-ROM drives.
• DVD-R and DVD-RW provide rewritable flexibility.

Blu-Ray Disc Technology

• Has same physical size as DVD/CD.
• Employs smaller pits, more tightly tracks, and uses blue-violet lasers to enable multiple layers.
• It features a data storage capacity ranging from 50GB-500GB and a data transfer rate of 432 mbps.
• Blu-ray Disc Formats include include BD-ROM, BD-R, BD-RE.

Flash Memory Storage

• Flash memory is electronically erasable programmable read-only memory (EEP).
• Flash memory nonvolatile, removable
• It emulates random access
• A key difference is data is stored securely (even if removed)
• Data is stored on a microchip card or "key" such as compact flash drives
• Flash memory can be connected through a USB port.
• Writing data: electric charge sent through a floating gate.
• Erasing data: apply a strong electrical field to flash storage.

Magnetic Disk Drive Access Times

• Factors affecting magnetic disk drive access times include seek time, which is the slowest

• The time it takes the read/write to position on the track but doesn't apply to fixed read/write head devices.

• Search time is rotational dealy.

• Transfer time is data transfer between secondary storage to main memory and the fastest out of all the access times

Fixed-Head Devices

• Record access requires track and record number
• Access time = search time + transfer time
• Total access time is rotational speed dependent.
• Rotation is continuous.
• Three basic positions for a requested record in relation to the read/write position.
• DASD has little access variance
• Good candidates: low activity files and applications with random access.
• Blocking can minimize access time.

Movable-Head Devices

• It takes seek time + search time + transfer time for record access.
• Has same search and transfer time calculation as fixed-head DASDs.
• Blocking is useful to minimize access time.

Components of the I/O Subsystem

• The I/O Channel includes programmable units positioned between the CPU and control unit.
• Synchronizes device speeds, fast to I/O device which is slow.
• Manages concurrent processing of CPU and I/O device requests.
• Allows overlap of the CPU and I/O operations.
• Expensive because channels are shared.
• I/O Channel Programs specify action performed by devices.
• It controls data transmission between main memory and control units.
• I/O Control Unit gets and interprets signal
• Disk controller links disk drive and system bus
• Entire path must be available when the I/O command begins
• Flexibility and reliability is increased through I/O subsystem configurations

Communication Among Devices

• The CPU needs to know with components are busy
• Units solves this by structuring interaction
• Accommodate requests during heavy processes by buffering and queing.
• It needs to accommodate speed disparity between the CPU and I/O devices by buffering and queuing.
• I/O subsystem units finish independently .
• The CPU processes data while I/O is performed.
• Success requires device completion knowledge.
• A hardware flag gets tested by CPU.
• Channel status word (CSW) contains flag
• Three bits in flag represent I/O system components
• Flag is tested by polling and interrupts
• Interrupts are a more efficient way to test flag
• Direct Memory Access (DMA) enables control unit access directly
• DMA Transfers data without intervention of CPU
• Buffers in main memory, channels, control units are temporary storage areas
• Double buffering: processes a record by the CPU while another is read or written by channel

Management of I/O Requests

• I/O traffic controllers watch status of devices and control units
• The scheduler allocates control units
• In many systems I/O cannot be preempted
• For some systems their is preemption with the I/O request being subdivided or preferential treatment for requests with higher priority
• The device handler performs data transfer.
• Has device dependency and unique algorithm.

Device Handler Seek Strategies

• Determine device processing order and minimize time spent seeking.
• Types of Strategies include FCFS, SSTF, SCAN, N-Step SCAN, and C-LOOK
• Goals:
  • Minimize arm movement
  • Minimize average response time
  • Reduce variance in response time
• FCFS: FCFS is not effective with three seek strategy goals; experiences extreme arm movement.
• Shortest Seek Time First (SSTF) Minimizes overall seek time but postpones access to out-of-the-way tracks.
• SCAN The directional bit tells if the arm is towards the center. Services requests from outer to inner until innermost track is reached.
• LOOK Arms don't go to the actual edge Unless there is an indefinite postponement It goes only as far as there are existing requests

Seek Strategies

• N-Step SCAN All request held until arm returns, then grouped together for next movement
• C-SCAN (Circular Scan) Picks up all request during inward sweeps
• C-LOOK: Inward sweeps are where it will stop at higher numbered requests
• FCFS works best under light loads otherwise the service time can become too long under heavy use.
• SSTF works best under moderate loads but becomes a localization problem under heavy load.
• SCAN works best for light to moderate loads where there is elimination of indefinite postponement that provides throughput and mean service times similar to SSTF.
• C-SCAN works best for heavy loads because of its small time variances.

Search Strategies: Rotational Ordering

• Aims to reduce wasted time due to rotational delay by ordering requests based on rotations per second and first sectors
• Once the controller determines the optimum order for read/write heads positioned
• Read/write time comes into play.
• Hardware Dependent
• First sector on second track requested is next higher than one just served.

RAID (Redundant Array of Independent Disks)

• Is a physical disk set that is seen as a single logical unit
• Preferable over few large disk drives
• Improves I/O and data recovery
• Introduces redundancy for hardware failure recovery
• Factors when selecting a RAID Level Includes costs and applications
• Can increase hardware costs

RAID Levels

• Level 0: Uses data striping without truth RAID, no parity or error corrections therefore no redundancy and recovery. The benefits are that the devise will act as one logical unit. Best used for non-critical data.
• Level 1: Uses data striping with mirror configurations. Provides increased reliability but with a more expensive set.
• Level 2: Uses smaller stripes which requires more detection and correction for a single array disk. The stripe size determines what disks it can access.
• Level 3: modifies level 2 by adding more disks which enables more for each strip.
• Level 4: Is the same as level 0 and 1 by computing parities for each strip but has a designated party disk
• Level 5: It distributes party strips
• Level 6: Provides extra protection. Does multiple parity calculations because it uses the same level four/five algorithm to get the parities and then stores them on separate strips to array so that you can restores data easily because it is more reliable with restoration because it is likely to restore data even if disk fails.
• Nested RAID means to combine multiple RAID levels (which can become complex) for redundancy
• Level 0+1 enables you o have multiple RAID levels for improved performance

Summary

• The Device Manager must effectively manage devices by synchronizing speed with its use of direct and sequential access.
• For Magnetic media with disk speed that is either fixed or movable across surface There is a different type of magnetic media with one to many read/write heads
• Flash memory used for device management has USB devices with safe data transfer
• I/O relies on communication channels that link the devices of data
• SCAN reduces loading on the device
• C-SCAN is best under moderate usage
• With RAID helps with recovery through costs, speed, and application
• FCFS makes things easy and is optimal for light loads
• SSTF improves movement with loading
• SCAN and LOOK eliminate starvation but needs a directory
• N-Step means to easily implement and make recent requests wait
• C-Scanis great when variance is implemented