Transcript Features Of Sprite Operating System

THE SPRITE NETWORK
OPERATING SYSTEM
PRESENTED BY,
PRASHANTHI NARAYAN NETTEM.
WHAT IS SPRITE?
• Sprite is an operating system implemented at the
University of California, Berkeley as part of
development of SPUR a high-performance multiprocessor
Workstation.
• Sprite is a distributed operating system that provides a single
system image to a cluster of workstations.
Features Of Sprite Operating System
• Process Migration
• Virtual Memory
• File Management
• Remote Procedure Calls
• Mutual Exclusion and
synchronization
Process Migration
• Sprite presents an illusion of a single fast time-sharing
system, rather than distributed system with many
independent hosts.
• Process migration facility moves a process execution
site between two machines of the same architecture.
• Process migration is both transparent and automatic.
• Migration involves two phases. They are extracting and
executing.
• Extracting includes installing the process state from
source to the target. It depends on the state associated
with the process.
Process Migration (Contd…)
• The second phase, process execution, depends not only
on the way in which state is transferred, but also the
degree to which migration is intended to be transparent.
• Process transfer includes files transfer, virtual memory
transfer and migrating the open file.
• Sprite uses a common high performance file system.
• File server keeps track of which host has a file open for
reading or writing.
• If a file is open for writing then caching is disabled and
all hosts must forward their read and write requests for
that file to the server so that they can be serialized.
Process Migration (Contd…)
• In Sprite backing storage for virtual memory is implemented on the
ordinary files.
• These backing files stored on the network file system, are accessible
throughout the network.
• The backing files are stored on network file servers, which cache
recently used file data in memory.
Process Migration (contd…)
• When migrating an open file to a new host, access to the file is
managed using standard mechanisms in the Sprite file system.
• The server that stores a file is responsible for keeping file’s
contents and processes consistent.
Virtual Memory Management
• Sprite’s virtual memory is very much similar to UNIX and has
been redesigned to support three features. They are
multiprocessing, networks and large physical memory.
• The large physical memories in Sprite offer the opportunity for
speeding program startup by using free memory as a cache for
recently used programs.
• Sprite uses the clock algorithm for page replacement and
provides shared read write data segments.
SHARED MEMORY
• The processes working together in a multi-processor
environment need to provide mechanisms for shared memory
and inter process communication (IPC) for synchronization and
sharing data.
• Sprite uses the shared writable memory and messages for IPC.
A virtual memory must be able to load pages into
memory and provide backing storage when a process
faults on a page.
Sprite uses file system for both demand loading and
backing store.
DEMAND LOADING OF CODE
• UNIX allows processes to share only code but Sprite allows
processes to share both code and heap of a process address space.
• Sprite initially loads code and initialized heap pages from object files
in the file system.
• Because of using high-performance file servers, Sprite file system is
fast enough for the virtual memory system to use it for all demand
loading.
• The Virtual memory is simplified because it does not have to worry
about the physical location on disk and the file’s server cache can be
used to increase performance as page reads can be serviced out of
cache instead of going to disk.
BACKING STORE MANAGEMENT
•
Backing store is used to store dirty pages when they are taken
away from a segment. In Sprite each segment has its own file in
the file system that it uses for backing store instead of a special
disk partition.
•
When a segment needs to write a dirty page out it is written to
the file using normal file write operation and all the read
operations by normal read operation by using the virtual address
of the page to be read.
•
Advantages :
1. no preallocated partition of disk space for backing
storage.
2. Migration is made easier as only the pointer to the
file need to be transferred.
FILE MANAGEMENT
• Sprite’s file system is implemented by a distributed set of computers
and its internal state is distributed among the operating system kernels at
the different sites.
• Sprite supports pseudo file system which is a facility that includes the
foreign file systems and arbitrary user services.
• It is a file system that allows further extensions to the system to be
implemented at the user level server processes instead of inside the
kernel.
• The pseudo file server does the recovery and the caching mechanism.
• Thus the Sprite’s file system provides support for user implemented
services rather than a message based kernel.
CACHING OF FILE SYSTEM
• Caches increase the file system performance by storing the
memory which is repeatedly accessed without the use of disk which
reduces the delay in going to disk.
• In Sprite caching is done at both the main memories of both server
and client.
• Caching at client reduces the communication delay caused by
fetching blocks from server. This speeds up the program and
increases the number of clients supported by the server.
• The size of cache can change dynamically. The virtual memory and
the file system negotiate for the machine’s physical memory as the
needs of both change.
CACHE CONSISTENCY
• Many clients can cache the same file simultaneously as
long as none of them is writing the file.
• In Sprite sequential write sharing and concurrent write
sharing cause the consistency problems.
• Sequential write sharing occurs when the file is not open
for reading or writing at the same time on different clients.
• Concurrent write sharing occurs when multiple clients can
open a file and atleast one of them has it open for writing.
• Sprite uses file servers as centralized points for cache
consistency. There are no client to client interactions.
Sequential write sharing
C1 READS FILE
C2 WRITES TO FILE
C1 READS FILE
TIME
Concurrent write sharing
C1 READS FILE
C2 WRITES TO FILE
C1 WRITES TO FILE
TIME
REMOTE PROCEDURE CALLS
• A remote procedure call is a procedure executed on a foreign
host and the result is sent back to the calling program.
• The foreign host is called the server and the calling program a
client.
• Sprite uses network RPC protocol that is used for
communication among the Sprite kernels.
• The protocol used is the extension of Birell-Nelson protocol
and is used to optimize bulk transfer of data.
• An RPC request or reply consists of two buffers plus a header.
The first buffer is used for marshalling small segments and the
other one refers to the large unintegrated block of data.
• The header contains boot time-stamp to easily detect crashes
and reboots.
Args
Calling Procedure
Called Procedure
Results
Request msg
Request msg
Args
Args
Calling
procedure
Network
Client
Stub
RPC
Transport
RPC
Transport
Server
Stub
Called
Procedure
Results
Result msg
Result msg
Results
MULTI-PROCESSOR SPRITE KERNEL
• Sprite operating system is designed to run on multiprocessors.
• The kernel is multi-threaded to allow more than one processor to
execute kernel code simultaneously.
• It has to provide means for synchronization and mutual exclusion
between threads.
• It uses condition variables and monitor style locks for this
purpose.
• There are two types of locks.
• Monitor lock is used to implement monitor. It is acquired
at the beginning and released at the end. When the lock is
already acquired and another process attempts to acquire
lock then it is put to sleep. When lock is released all the
waiting threads are awakened.
• The another type of lock is master lock. It is used to
provide mutual exclusion between processes and interrupt
handlers.
If the lock is already existing then the processor retries the
locking operation until it succeeds.
It is a spin lock acquired with the interrupts disabled.
Interrupts are disabled because if an interrupt is taken by a
master lock then interrupt routine has to spin forever to
check whether the lock is released.
• Condition variables are used for synchronization. The process has
to wait for interesting conditions to occur. A process waits on the
conditional variable until it is signaled by another process.
CONCLUSIONS
Sprite is an operating system designed to provide high
performance, consistency and simplicity.
High performance is achieved using server and client caches.
Sprite provides non-write through file cache. The cache
consistency mechanism permits files to be shared without
danger of stale data.
Virtual memory uses ordinary files as backing storage for
simplicity, easy implementation of process migration and
dynamic usage of disk storage.