Operating Systems 18CS43
MODULE 2
MULTITHREADED PROGRAMMING
A thread is a basic unit of CPUutilization.
It consists of
thread ID
PC
register-set and
stack.
It shares with other threads belonging to the same process its code-section &data-section.
A traditional (or heavy weight) process has a single thread of control.
If a process has multiple threads of control, it can perform more than one task at a time.
such a process is called multithreaded process
Fig: Single-threaded and multithreaded processes
Motivation for Multithreaded Programming
1. The software-packages that run on modern PCs are multithreaded. An application is
implemented as a separate process with several threads of control. For ex: A word processor
may have
first thread for displaying graphics
second thread for responding to keystrokes and
Third thread for performing grammar checking.
1
Operating Systems 18CS43
2. In some situations, a single application may be required to perform several similar tasks. For
ex: A web-server may create a separate thread for each client requests. This allows the server to
service several concurrent requests.
3. RPC servers are multithreaded.
When a server receives a message, it services the message using separate concurrent
threads.
4. Most OS kernels are multithreaded;
Several threads operate in kernel, and each thread performs a specific task,
such as managing devices or interrupt handling.
Benefits of Multithreaded Programming
Responsiveness A program may be allowed to continue running even if part of it is
blocked. Thus, increasing responsiveness to the user.
Resource Sharing by default, threads share the memory (and resources) of the
process to which they belong. Thus, an application is allowed to have several
different threads of activity within the same address-space.
Economy Allocating memory and resources for process-creation is costly. Thus, it is
more economical to create and context-switch threads.
Utilization of Multiprocessor Architectures in a multiprocessor architecture,
threads may be running in parallel on different processors. Thus, parallelism will be
increased.
MULTITHREADING MODELS
Support for threads may be provided at either
1. The user level, for user threads or
2. By the kernel, for kernel threads.
User-threads are supported above the kernel and are managed without kernel support.
Kernel-threads are supported and managed directly by the OS.
Three ways of establishing relationship between user-threads &kernel-threads:
1. Many-to-one model
2. One-to-one model and
3. Many-to-many model.
2
Operating Systems 18CS43
Many-to-One Model
Many user-level threads are mapped to one kernel thread.
Advantages:
Thread management is done by the thread library in user space, so it is efficient.
Disadvantages:
The entire process will block if a thread makes a blocking system-call.
Multiple threads are unable to run in parallel on multiprocessors.
For example:
Solaris green threads
GNU portable threads.
Fig: Many-to-one model
One-to-One Model
Each user thread is mapped to a kernel thread.
Advantages:
It provides more concurrency by allowing another thread to run when a thread
makes a blocking system-call.
Multiple threads can run in parallel on multiprocessors.
Disadvantage:
Creating a user thread requires creating the corresponding kernel thread.
For example:
Windows NT/XP/2000, Linux
Fig: one-to-one model
3
Operating Systems 18CS43
Many-to-Many Model
Many user-level threads are multiplexed to a smaller number of kernel threads.
Advantages:
Developers can create as many user threads as necessary
The kernel threads can run in parallel on a multiprocessor.
When a thread performs a blocking system-call, kernel can schedule another thread
for execution.
Two Level Model
A variation on the many-to-many model is the two level-model
Similar to M: N, except that it allows a user thread to be bound to kernel thread.
for example:
HP-UX
Tru64 UNIX
Fig: Many-to-many model Fig: Two-level model
Thread Libraries
It provides the programmer with an API for the creation and management of threads.
Two ways of implementation:
1. First Approach:
Provides a library entirely in user space with no kernel support. All code and data structures
for the library exist in the user space.
2. Second Approach
Provides a library entirely in user space with no kernel support. All code and data structures
for the library exist in the user space.
4
Operating Systems 18CS43
Three main thread libraries:
1. POSIXP threads
2. Win32 and
3. Java.
Pthreads
This is a POSIX standard API for thread creation and synchronization.
This is a specification for thread-behavior, not an implementation.
OS designers may implement the specification in any way they wish.
Commonly used in: UNIX and Solaris.
5
Operating Systems 18CS43
Win32 threads
Implements the one-to-one mapping
Each thread contains
A threaded
Register set
Separate user and kernel stacks
Private data storage area
The register set, stacks, and private storage area are known as the context of the
threads The primary data structures of a thread include:
ETHREAD (executive thread block)
KTHREAD (kernel thread block)
TEB (thread environment block)
Java Threads
Threads are the basic model of program-execution in
Java program and
Java language.
The API provides a rich set of features for the creation and management of threads.
All Java programs comprise at least a single thread of control.
Two techniques for creating threads:
1. Create a new class that is derived from the Thread class and override its run () method.
2. Define a class that implements the Runnable interface. The Runnable interface is
defined as follows:
6
Operating Systems 18CS43
THREADING ISSUES
fork () and exec () System-calls
fork () is used to create a separate, duplicate process.
If one thread in a program calls fork (), then
1. Some systems duplicates all threads and
2. Other systems duplicate only the thread that invoked the forkO.
If a thread invokes the exec(), the program specified in the parameter to exec() will
replace the entire process including all threads.
Thread Cancellation
This is the task of terminating a thread before it has completed.
Target thread is the thread that is to be cancelled
Thread cancellation occurs in two different cases:
1. Asynchronous cancellation: One thread immediately terminates the target thread.
2. Deferred cancellation: The target thread periodically checks whether it should be
terminated.
Signal Handling
In UNIX, a signal is used to notify a process that a particular event has occurred.
All signals follow this pattern:
1. A signal is generated by the occurrence of a certain event.
2. A generated signal is delivered to process.
3. Once delivered, the signal must be handled.
A signal handler is used to process signals.
A signal may be received either synchronously or asynchronously, depending on the source.
1. Synchronoussignals
Delivered to the same process that performed the operation causing the signal.
E.g. illegal memory access and division by 0.
2. Asynchronous signals
Generated by an event external to a running process.
E.g. user terminating a process with specific keystrokes<ctrl><c>.
7
Operating Systems 18CS43
Every signal can be handled by one of two possible handlers:
1. A Default Signal Handler
Run by the kernel when handling the signal.
2. A User-defined Signal Handler
Overrides the default signal handler.
In single-threaded programs, delivering signals is simple (since signals are always
delivered to a process).
In multithreaded programs, delivering signals is more complex. Then, the following
options exist:
1. Deliver the signal to the thread to which the signal applies.
2. Deliver the signal to every thread in process
3. Deliver the signal to certain threads in the process.
4. Assign a specific thread to receive all signals for the process.
THREAD POOLS
The basic idea is to
create a no. of threads at process-startup and
place the threads into a pool (where they sit and wait for work).
Procedure:
1. When a server receives a request, it awakens a thread from the pool.
2. If any thread is available, the request is passed to it for service.
3. Once the service is completed, the thread returns to the pool.
Advantages:
Servicing a request with an existing thread is usually faster than waiting to
create a thread.
The pool limits the no. of threads that exist at any one point.
No. of threads in the pool can be based on actors such as
no. of CPUs
amount of memory and
expected no. of concurrent client-requests.
THREAD SPECIFIC DATA
Threads belonging to a process share the data of the process.
this sharing of data provides one of the benefits of multithreaded programming.
In some circumstances, each thread might need its own copy of certain data. We will call such
data thread-specific data.
For example, in a transaction-processing system, we might service each transaction in a
separate thread.
8
Operating Systems
Furthermore, each transaction may be assigned a unique identifier. To associate
each thread with its unique identifier, we could use thread-specific data.
SCHEDULER ACTIVATIONS
Both M:M and Two-level models require communication to maintain the
appropriate number of kernel threads allocated to the application.
Scheduler activations provide up calls a communication mechanism from
thekernel to the thread library
This communication allows an application to maintain the correct number kernel
threads
One scheme for communication between the user-thread library and the kernel is
known as scheduler activation.
9
Operating Systems
PROCESS SCHEDULING
Basic Concepts
In a single-processor system,
Only one process may run at a time.
Other processes must wait until the CPU is rescheduled.
Objective of multiprogramming:
To have some process running at all times, in order to maximize CPU
utilization.
CPU-I/0 Burst Cycle
Process execution consists of a cycle of
CPU execution and
I/O wait
Process execution begins with a CPU burst, followed by an I/O burst, then
another CPU burst, etc…
Finally, a CPU burst ends with a request to terminate execution.
An I/O-bound program typically has many short CPUbursts.
A CPU-bound program might have a few long CPU bursts.
Fig Alternating sequence of CPU and I/O bursts
10
Operating Systems
Fig: Histogram of CPU-burst durations
CPU Scheduler
This scheduler
selects a waiting-process from the ready-queue and
allocates CPU to the waiting-process.
The ready-queue could be a FIFO, priority queue, tree and list.
The records in the queues are generally process control blocks (PCBs) of the processes.
CPU Scheduling
Four situations under which CPU scheduling decisions take place:
1. When a process switches from the running state to the waiting state. For ex; I/O
request.
2. When a process switches from the running state to the ready state. For ex:
when an interrupt occurs.
3. When a process switches from the waiting state to the ready state. For ex:
completion of I/O.
4. When a process terminates.
Scheduling under 1 and 4 is non- preemptive. Scheduling under 2 and 3 is preemptive.
Non Preemptive Scheduling
Once the CPU has been allocated to a process, the process keeps the CPU until it
releases the CPU either
by terminating or
by switching to the waiting state.
Preemptive Scheduling
This is driven by the idea of prioritized computation.
Processes that are runnable may be temporarily suspended
Disadvantages:
1. Incurs a cost associated with access to shared-data.
2. Affects the design of the OSkernel.
11
Operating Systems
Dispatcher
It gives control of the CPU to the process selected by the short-term scheduler.
The function involves:
1. Switching context
2. Switching to user mode&
3. Jumping to the proper location in the user program to restart that program
It should be as fast as possible, since it is invoked during every process switch.
Dispatch latency means the time taken by the dispatcher to
stop one process and
start another running.
SCHEDULING CRITERIA:
In choosing which algorithm to use in a particular situation, depends upon the properties
of the various algorithms. Many criteria have been suggested for comparing CPU-
scheduling algorithms. The criteria include the following:
1. CPU utilization: We want to keep the CPU as busy as possible. Conceptually,
CPU utilization can range from 0 to 100 percent. In a real system, it should range
from 40 percent (for a lightly loaded system) to 90 percent (for a heavily used
system).
2. Throughput: If the CPU is busy executing processes, then work is being done.
One measure of work is the number of processes that are completed per time unit,
called throughput. For long processes, this rate may be one process per hour; for
short transactions, it may be ten processes per second.
3. Turnaround time. This is the important criterion which tells how long it takes to
execute that process. The interval from the time of submission of a process to the
time of completion is the turnaround time. Turnaround time is the sum of the
periods spent waiting to get into memory, waiting in the ready queue, executing on
the CPU, and doing I/0.
4. Waiting time: The CPU-scheduling algorithm does not affect the amount of time
during which a process executes or does I/0, it affects only the amount of time that
a process spends waiting in the ready queue. Waiting time is the sum of the periods
spent waiting in the ready queue.
5. Response time: In an interactive system, turnaround time may not be the best
criterion. Often, a process can produce some output fairly early and can continue
computing new results while previous results are being output to the user. Thus,
another measure is the time from the submission of a request until the first response
is produced. This measure, called response time, is the time it takes to start
responding, not the time it takes to output the response. The turnaround time is
generally limited by the speed of the output device.
12
Operating Systems
SCHEDULING ALGORITHMS
CPU scheduling deals with the problem of deciding which of the processes in
the ready-queue is to be allocated the CPU.
Following are some scheduling algorithms:
1. FCFS scheduling (First Come First Served)
2. Round Robin scheduling
3. SJF scheduling (Shortest Job First)
4. SRT scheduling
5. Priority scheduling
6. Multilevel Queue scheduling and
7. Multilevel Feedback Queue scheduling
FCFS Scheduling
The process that requests the CPU first is allocated the CPU first.
The implementation is easily done using a FIFO queue.
Procedure:
1. When a process enters the ready-queue, its PCB is linked onto the tail of
the queue.
2. When the CPU is free, the CPU is allocated to the process at the queue’s head.
3. The running process is then removed from the queue.
Advantage:
1. Code is simple to write & understand.
Disadvantages:
1. Convoy effect: All other processes wait for one big process to get off the CPU.
2. Non-preemptive (a process keeps the CPU until it releases it).
3. Not good for time-sharing systems.
4. The average waiting time is generally not minimal.
Example: Suppose that the processes arrive in the order P1, P2,P3.
The Gantt Chart for the schedule is as follows:
Waiting time for P1 = 0; P2 = 24; P3 =27
Average waiting time: (0 + 24 + 27)/3 = 17ms
Suppose that the processes arrive in the order P2, P3, P1.
13
Operating Systems
The Gantt chart for the schedule is as follows:
Waiting time for P1 = 6; P2 = 0; P3 =3
Average waiting time: (6 + 0 + 3)/3 = 3ms
SJF Scheduling
The CPU is assigned to the process that has the smallest next CPU burst.
If two processes have the same length CPU burst, FCFS scheduling is used to break
the tie.
For long-term scheduling in a batch system, we can use the process time limit
specified by the user, as the ‘length’
SJF can't be implemented at the level of short-term scheduling, because there is
no way to know the length of the next CPU burst
Advantage:
1. The SJF is optimal, i.e. it gives the minimum average waiting time for a
given set of processes.
Disadvantage:
1. Determining the length of the next CPU burst.
SJF algorithm may be either 1) non-preemptive or 2) preemptive.
1. Non preemptive SJF
The current process is allowed to finish its CPU burst.
2. Preemptive SJF
If the new process has a shorter next CPU burst than what is left of the
executing process, that process is preempted. It is also known as SRTF
scheduling (Shortest-Remaining-Time-First).
Example (for non-preemptive SJF): Consider the following set of processes,
with the length of the CPU-burst time given in milliseconds.
For non-preemptive SJF, the Gantt Chart is as follows:
14
Operating Systems
Waiting time for P1 = 3; P2 = 16; P3 = 9; 4=0 Average waiting time: (3 + 16 + 9 +
0)/4= 7
preemptive SJF/SRTF: Consider the following set of processes, with the length
of the CPU- burst time given in milliseconds.
For preemptive SJF, the Gantt Chart is as follows:
The average waiting time is ((10 - 1) + (1 - 1) + (17 - 2) + (5 - 3))/4 = 26/4 =6.5.
Priority Scheduling
A priority is associated with each process.
The CPU is allocated to the process with the highest priority.
Equal-priority processes are scheduled in FCFS order.
Priorities can be defined either internally or externally.
1. Internally-defined priorities.
Use some measurable quantity to compute the priority of a process.
For example: time limits, memory requirements, no. f open files.
2. Externally-defined priorities.
Set by criteria that are external to the OS for
example:
importance of the process, political factors
Priority scheduling can be either preemptive or non-preemptive.
1.Preemptive
The CPU is preempted if the priority of the newly arrived process is
higher than the priority of the currently running process.
2. Non Preemptive
The new process is put at the head of the ready-queue
Advantage:
Higher priority processes can be executed first.
Disadvantage:
Indefinite blocking, where low-priority processes are left waiting
indefinitely for CPU. Solution: Aging is a technique of increasing
priority of processes that wait in system for a long time.
15
Operating Systems
Example: Consider the following set of processes, assumed to have arrived at
time 0, in the order PI, P2, ..., P5, with the length of the CPU-burst time given in
milliseconds.
The Gantt chart for the schedule is as follows:
The average waiting time is 8.2milliseconds.
Round Robin Scheduling
Designed especially for timesharing systems.
It is similar to FCFS scheduling, but with preemption.
A small unit of time is called a time quantum (or time slice).
Time quantum is ranges from 10 to 100ms.
The ready-queue is treated as a circular queue.
The CPU scheduler
goes around the ready-queue and
allocates the CPU to each process for a time interval of up to 1-time
quantum.
To implement:
The ready-queue is kept as a FIFO queue of processes
CPU scheduler
1. Picks the first process from the ready-queue.
2. Sets a timer to interrupt after 1 time quantum and
3. Dispatches the process.
One of two things will then happen.
1. The process may have a CPU burst of less than 1 time quantum. In this case,
the process itself will release the CPU voluntarily.
2. If the CPU burst of the currently running process is longer than 1 time
quantum, the timer will go off and will cause an interrupt to the OS. The
process will be put at the tail of the ready-queue.
Advantage:
Higher average turnaround than SJF.
Disadvantage:
Better response time than SJF.
Example: Consider the following set of processes that arrive at time 0, with the
length of the CPU-burst time given in milliseconds.
16
Operating Systems
The Gantt chart for the schedule is as follows:
The average waiting time is 17/3 = 5.66milliseconds.
The RR scheduling algorithm is preemptive.
No process is allocated the CPU for more than 1 time quantum in a row.
If a process' CPU burst exceeds 1 time quantum, that process is
preempted and is put back in the ready- queue.
The performance of algorithm depends heavily on the size of the time quantum.
1. If time quantum=very large, RR policy is the same as the FCFS policy.
2. If time quantum=very small, RR approach appears to the users as though each
of n processes has its own processor running at l/n the speed of the real
processor.
In software, we need to consider the effect of context switching on the
performance of RR scheduling
1. Larger the time quantum for a specific process time, less time is spend on
context switching.
2. The smaller the time quantum, more overhead is added for the purpose of
context- switching.
Fig: How a smaller time quantum increases context switches
17
Operating Systems
Fig: How turnaround time varies with the time quantum
Multilevel Queue Scheduling
Useful for situations in which processes are easily classified into different groups.
For example, a common division is made between
foreground (or interactive) processes and
background (or batch) processes.
The ready-queue is partitioned into several separate queues (Figure2.19).
The processes are permanently assigned to one queue based on some property like
memory size
process priority or
process type.
Each queue has its own scheduling algorithm.
For example, separate queues might be used for foreground and background
processes.
Fig Multilevel queue scheduling
18
Operating Systems
There must be scheduling among the queues, which is commonly implemented
as fixed-priority preemptive scheduling.
For example, the foreground queue may have absolute priority over the
background queue.
Time slice: each queue gets a certain amount of CPU time which it can
schedule amongst its processes; i.e., 80% to foreground in RR 20% to background
in FCFS
Multilevel Feedback Queue Scheduling
A process may move between queues
The basic idea: Separate processes according to the features of their CPU bursts. For
example
1. If a process uses too much CPU time, it will be moved to a lower-priority queue.
This scheme leaves I/O-bound and interactive processes in the higher-priority
queues.
2. If a process waits too long in a lower-priority queue, it may be moved to a
higher-priority queue This form of aging prevents starvation.
Figure 2.20 Multilevel feedback queues
In general, a multilevel feedback queue scheduler is defined by the following parameters:
1. The number of queues.
2. The scheduling algorithm for each queue.
3. The method used to determine when to upgrade a process to a higher priority queue.
4. The method used to determine when to demote a process to a lower priority queue.
5. The method used to determine which queue a process will enter when that
process needs service
19
Operating Systems
MULTIPLE PROCESSOR SCHEDULING
If multiple CPUs are available, the scheduling problem becomes more complex.
Two approaches:
Asymmetric
MultiprocessingThe basic
idea is:
A master server is a single processor responsible for all scheduling decisions, I/O
processing and other system activities.
The other processors execute only user code.
Advantage: This is simple because only one processor accesses the system data
structures, reducing the need for data sharing.
Symmetric Multiprocessing
The basic idea is:
Each processor is self-scheduling.
To do scheduling, the scheduler for each processor
Examines the ready-queue and
Selects a process to execute.
Restriction: We must ensure that two processors do not choose the same process and that
processes are not lost from the queue.
Processor Affinity
In SMP systems,
1. Migration of processes from one processor to another are avoided and
2. Instead processes are kept running on same processor. This is known as
processor affinity.
Two forms:
1. Soft Affinity
When an OS try to keep a process on one processor because of
policy, but cannot guarantee it will happen.
It is possible for a process to migrate between processors.
2. Hard Affinity
When an OS have the ability to allow a process to specify that it is not to
migrate to other processors. Eg: Solaris OS
Load Balancing
This attempts to keep the workload evenly distributed across all processors in an
SMP system.
Two approaches:
1. Push Migration
A specific task periodically checks the load on each processor and if it finds an
imbalance, it evenly distributes the load to idle processors.
2. Pull Migration
An idle processor pulls a waiting task from a busy processor.
20
Operating Systems
Symmetric Multithreading
The basic idea:
1. Create multiple logical processors on the same physical processor.
2. Present a view of several logical processors to the OS.
Each logical processor has its own architecture state, which includes general-
purpose and machine-state registers.
Each logical processor is responsible for its own interrupt handling.
SMT is a feature provided in hardware, not software.
THREAD SCHEDULING
On OSs, it is kernel-level threads but not processes that are being scheduled by the OS.
User-level threads are managed by a thread library, and the kernel is unaware of them.
To run on a CPU, user-level threads must be mapped to an associated kernel-level
thread.
Contention Scope
Two approaches:
1. Process-Contention scope
On systems implementing the many-to-one and many-to-many models, the
thread library schedules user-level threads to run on an available LWP.
Competition for the CPU takes place among threads belonging to thesame
process.
2. System-Contention scope
The process of deciding which kernel thread to schedule on the CPU.
Competition for the CPU takes place among all threads in the system.
Systems using the one-to-one model schedule threads using only SCS.
Pthread Scheduling
Pthread API that allows specifying either PCS or SCS during thread creation.
Pthreads identifies the following contention scope values:
1. PTHREAD_SCOPEJPROCESS schedules threads using PCS scheduling.
2. PTHREAD-SCOPE_SYSTEM schedules threads using SC scheduling.
Pthread IPC provides following two functions for getting and setting the contention
scope policy:
1. pthread_attr_setscope(pthread_attr_t *attr, intscope)
2. pthread_attr_getscope(pthread_attr_t *attr, int*scop)
21