1. Prof. Leonardo Mostarda University of Camerino
Distributed Systems – processes
Prof. Leonardo Mostarda
University of Camerino
Prof. Leonardo Mostarda-- Camerino,

2. Last lecture Decentralised architectures
Structured peer-to-peer architectures
Unstructured peer-to-peer architectures
Hybrid architectures
Superpeer architectures
Collaborative Distributed Systems

3. Outline
Processes
Threads
Multithreaded clients
Multithreaded servers
Virtualisation

4. Learning outcomes
Understand threads and processes
Critically compare threads and processes
Understand the use of threads in distributes systems
Understand the role of virtualisation in distributes systems

5. Process A process is a program in execution
Processes are like human beings:
they are generated
they have a life
they optionally generate one or more child processes
and eventually they die
A small difference
Each process has exactly one parent

6. Processes: The Process Model
A) Conceptual model of 4 independent, sequential processes
B) Multiprogramming of four programs
C) Only one program active at any instant

7. Process States Possible process states
running
blocked
ready
Transitions between states shown

8. Processes
Creation of processes incurs in high overhead (concurrency transparency)
Independent address space
Initialising memory
Zeroing a data segment
Copying the program into the data segment ……
Switching between processes incurs in high overhead
Saving CPU context (register values, program counter, open files, stack pointer, …)
Swap processes between memory and disk
Load a new process
……..

9. The Thread Model (a) Three processes each with one thread
(b) One process with three threads

10. The Thread Model Items shared by all threads in a process
Items private to each thread

11. Threads vs Processes
Multithreading leads to performance gain(creation and switching are faster)
A multithread system maintains minimum information in order to allow different threads to share the same CPU
Switching can take place in the user space
Processes require Interprocess Communication (IPC) (pipes, message queues, shared memory) which require kernel intervention
Advantages of processes?
Thread are not automatically protected against each other (no concurrency transparency)
Developing multithreading application requires additional intellectual effort

12. Thread Usage in non-distributed systems
A word processor with three threads
The user can still write while the text is saved, and the spell check is performed

13. Threads in Distributed systems
Threads allow blocking system calls without blocking the entire process
Multiple connection can take place at the same time
We describe multithreaded clients
We describe multithreaded servers

14. Multithreaded client Clients hide communication latencies with threads
After a thread initiates the communication a different thread can run
For instance Google chrome
When the html is fetched various threads are used
Display the text and GUI elements
Fetch images
Fetch data
When WEB servers are replicated connections are set up to different replicas

15. Multithreaded server
Multithreading simplifies the server code and makes easier to develop server with better performance even on uniprocessor systems
For instance in a file server
A dispatcher reads the incoming request for a file operation
An idle worker thread is selected
The worker thread can block until data are fetched from the disk
The dispatcher can acquire new works
A single thread solution would waste CPU (3-4 cannot run in parallel)

16. Multithreaded server
What if threads are not available? How do we implement our file server?
Finite state machine
The server runs as a big state machine
There is a single thread that does not block. It records a request and its state (e.g., waiting from disk) and gets the next message
The next message may be a replay from the disk
Any important comments?
The server will have to make use of no blocking calls to send and receive (better performance but hard to program)

17. The Role of Virtualization in Distributed Systems
Every distributed computer system offers an interface to higher level software (Figure a)
Virtualisation modifies the existing interface so as to mimic the behaviour of another system (Figure b)
Introduced in 1970s to allow legacy software to run on expensive mainframe
Different OSs were ported on the IBM 370 mainframe

18. The use of virtualisation today
Hardware and low-level system change fast while higher level applications are more stable. Virtualisation help by porting a legacy interface to new platforms.
System administrators can manage a single platform type. An applications runs on its own virtual machine that in turn runs on the platform.
Other example?

19. Architectures of Virtual Machines
In order to understand how virtualisation can be realised we need to understand the different levels at which interfaces are provided:
An interface between the hardware and software consisting of machine instructions
that can be invoked by any program.
An interface between the hardware and software, consisting of machine instructions
that can be invoked only by privileged programs, such as an operating system.

20. Architectures of Virtual Machines
Interfaces at different levels
An interface consisting of system calls as offered by an operating system.
An interface consisting of library calls
generally forming what is known as an application programming interface (API).
In many cases, the aforementioned system calls are hidden by an API.

21. Architectures of Virtual Machines
First way of implement virtualisation
A runtime system can provide an abstract instruction set that can be used by applications
The runtime system can be an interpreter or the emulation of an OS

22. Architectures of Virtual Machines
Second way of implement virtualisation
A layer completely shields the original hardware
This is referred to as Virtual Machine Monitor (VMM)
Decoupling between hardware and software

23. Summary
Processes
Threads
Multithreaded client
Multithreaded server
Virtualisation