Rensselaer Polytechnic Institute CSC 432 – Operating Systems David Goldschmidt, Ph.D.

 A virtual filesystem provides transparent access to different filesystem types on multiple device types and disk partitions

 A virtual machine is a layered approach that logically combines the kernel operating system and hardware  Creates the illusion of multiple processes, each executing on its own virtual processor with its own virtual memory

Non-virtual machine Virtual machine

 Java programs execute on a native Java Virtual Machine (JVM)

 Virtual machines provide complete protection of system resources  Each virtual machine is isolated from all other virtual machines ▪ which prohibits direct sharing of system resources  Virtual machines can be difficult to implement due to the effort required to provide an exact duplicate of each underlying machine

 The Input/Output (I/O) System has two primary objectives:  Handle application I/O requests ▪ Map logical address to physical disk or device address ▪ Send response back to the application  Optimize I/O performance ▪ Depends on request type and device type

 Disks and other devices operate in parallel to the CPU (but are much slower)  Typical disk drive mechanism:  Arm seeks to the appropriate track  Disk rotates until the desired sector is accessed

 Disk access time is the sum of the seek time and the rotational latency  Cache surrounding sectors or entire track to improve performance  Principle of locality (again!)

 Disk access times are orders of magnitude slower than CPU execution times  Improve I/O performance by:  Reducing the number of I/O requests  Implementing buffering  Implementing caching  Efficiently scheduling I/O requests do this at the application layer

 Use buffering to make physical I/O requests as large as possible  This reduces the number of I/O requests  Space-time tradeoff  Misleads programmers?  Other disadvantages?

 Use caching to keep retrieved data in fast memory for potential future access  Eliminates one or more I/O requests  Space-time tradeoff  Principle of locality (yet again!)

 A disk context switch occurs when switching from one I/O request to another  Disk context switch time is substantially higher than process context switch  Disk context switch time is substantially lower than disk read/write operation  The time to complete the nth I/O operation depends on where the (n-1)th operation finished

 Goal: optimize disk performance  Scheduling algorithm determines which pending disk I/O request to select next:  First-Come-First-Served (FCFS)  Shortest Seek Time First (SSTF)  Elevator (SCAN) and Circular SCAN (C-SCAN) maximize throughput, ensure fairness, etc.

 Request reference string specifies requested tracks: 44, 20, 95, 4, 50, 52, 47, 61, 87, 25

 Request reference string specifies requested tracks: 98, 183, 37, 122, 14, 124, 65, 67

 Request reference string specifies requested tracks: 44, 20, 95, 4, 50, 52, 47, 61, 87, 25

 Request reference string specifies requested tracks: 98, 183, 37, 122, 14, 124, 65, 67

 Request reference string specifies requested tracks: 44, 20, 95, 4, 50, 52, 47, 61, 87, 25 repeated end-to-end scans

 Request reference string specifies requested tracks: 98, 183, 37, 122, 14, 124, 65, 67

 Circular SCAN (C-SCAN) scans in one direction  When it reaches one end of the disk, it returns to the beginning of the disk without servicing any requests on the return trip  LOOK (and C-LOOK) algorithms  Disk arm moves in one direction as long as there are pending requests in that direction  Otherwise, it reverses direction immediately