COMP90015 Distributed Systems
Indirect Communication
School of Computing and Information Systems © The University of Melbourne
Copyright By PowCoder代写 加微信 powcoder
2022 Semester II
(School of Computing and InformaCtiOonMSPy9s0te0m15s ©DistTrihbeutUednivSeyrsstietymos f Melbourne)
2022 Semester II
1 Message Queues
2 Group Communication
3 Publish/Subscribe
4 Distributed Shared Memory Tuple Spaces
5 Comparison
(School of Computing and InformaCtiOonMSPy9s0te0m15s ©DistTrihbeutUednivSeyrsstietymos f Melbourne)
2022 Semester II
Indirect Communication
• Whereas direct communication is communication that takes place directly between the communicating processes, indirect communication is defined as communication between entities in a distributed system through an intermediary with no direct coupling between the sender and the receiver(s).
• Space uncoupling – sender does not know or need to know the identity of the receiver(s)
• Time uncoupling – sender and receiver can have independent lifetimes, they do not need
to exist at the same time
• Time uncoupling is not synonymous with asynchronous communication. Asynchronous communication doesn’t imply that the receiver has an independent lifetime, in other words we could consider a time coupled asynchronous system.
• Indirect communication paradigms tend to be described in terms of a metaphor that aids in understanding the expectations of the paradigm and for what kinds of distributed applications it is useful:
• Message Queues
• Group Communication
• Publish/Subscribe
• Shared Memory
• Tuple Spaces
(School of Computing and InformaCtiOonMSPy9s0te0m15s ©DistTrihbeutUednivSeyrsstietymos f Melbourne)
2022 Semester II
Message Queues
Message Queues
• Message queues provide a point-to-point service using the concepts of a message for data encapsulation and queue as an indirection. They are point-to-point in that each message is sent by a single process – producer – and received by a single process – consumer.
• Since communication uses messages, the message queue paradigm may not be suitable for applications that require streaming data or bulk data transfer.
• Good for distributing units of work to processes and command/control type operations.
(School of Computing and InformaCtiOonMSPy9s0te0m15s ©DistTrihbeutUednivSeyrsstietymos f Melbourne) 2022 Semester II 4 / 25
Message Queues
Programming model
Usually the message queue system is expected to provide reliability in that messages are not dropped or lost, and since its a queue, messages are received in the order sent. The API is very much the same as a blocking queue that is used in concurrent programming.
• send – producers put a message on a particular queue, may block the sender if the queue has finite capacity.
• blocking receive – a consumer waits for at least one message on a queue and then returns.
• non-blocking receive – or poll, a consumer will check and get a message if there, otherwise it returns without a message.
• notify – a signal is sent to the consumer when messages are available on the queue for consumption.
It is useful to consider this API in terms of actual processes and low-level exchange protocols. E.g. the implementation may use TCP for producers and consumers to connect to the queueing system.
(School of Computing and InformaCtiOonMSPy9s0te0m15s ©DistTrihbeutUednivSeyrsstietymos f Melbourne) 2022 Semester II 5 / 25
Message Queues
• A message queueing system typically provides a library for the programming to build a client, either a producer or a consumer, and a server implementation that implements the queue manager itself. The server implementation will typically run a process that allows producers and consumers to connect.
• Modern examples include RabbitMQ and ZeroMQ.
(School of Computing and InformaCtiOonMSPy9s0te0m15s ©DistTrihbeutUednivSeyrsstietymos f Melbourne) 2022 Semester II 6 / 25
Message Queues
Discussion questions
Question (1): Discuss how the message queue paradigm could be used to implement a chat room application.
(School of Computing and InformaCtiOonMSPy9s0te0m15s ©DistTrihbeutUednivSeyrsstietymos f Melbourne) 2022 Semester II 7 / 25
Group Communication
Group Communication
Group communication offers a space uncoupled service whereby a message is sent to a group and then this message is delivered to all members of the group. It provides more than a primitive IP multicast:
• manages group membership
• detects failures and provides reliability and ordering guarantees
Typical aspects of a Group API: • group creation
• create/delete a group
• list/search available groups
• group membership
• join/leave a group
• list members of a group
• multicast to selected members of a group, broadcast to all members
Efficient sending to multiple receivers, instead of multiple independent send operations, is an essential feature of group communication.
(School of Computing and InformaCtiOonMSPy9s0te0m15s ©DistTrihbeutUednivSeyrsstietymos f Melbourne) 2022 Semester II 8 / 25
Group Communication
(School of Computing and InformaCtiOonMSPy9s0te0m15s ©DistTrihbeutUednivSeyrsstietymos f Melbourne) 2022 Semester II 9 / 25
Group Communication
Group services
• closed groups only allow group members to multicast to it
• overlapping groups allows entities to be members of multiple groups • synchronous and asynchronous variations can be considered
(School of Computing and InformaCtiOonMSPy9s0te0m15s ©DistTrihbeutUednivSeyrsstietymos f Melbourne) 2022 Semester II 10 / 25
Group Communication
Implementation issues
• reliability and ordering in multicast
• FIFO (first in first out) ordering is concerned with preserving the order from the perspective of a sender process
• causal ordering, a message that happens before another message will be preserved in that order in the delivery at all processes
• total ordering, if a message is delivered before another message at one process then this is preserved at all processes
• group membership management
• group members leave and join
• failed members
• notifying members of group membership changes
• changes to the group address
(School of Computing and InformaCtiOonMSPy9s0te0m15s ©DistTrihbeutUednivSeyrsstietymos f Melbourne)
2022 Semester II
Group Communication
Discussion questions
Question (2): Discuss how the group communication paradigm could be used to implement a chat room application. Would this be better or worse than using the message queue paradigm?
(School of Computing and InformaCtiOonMSPy9s0te0m15s ©DistTrihbeutUednivSeyrsstietymos f Melbourne) 2022 Semester II 12 / 25
Publish/Subscribe
Publish/Subscribe Systems
• Publish/subscribe systems are sometimes referred to as distributed event-based systems. A publish/subscribe system is a system where publishers (event sources) publish structured events to an event service and subscribers express interest in particular events through subscriptions which can be arbitrary patterns or query expressions over the structured events.
• financial information systems
• live feeds of real-time data, e.g. RSS feeds
• support for cooperative working, where a number of participants need to be informed of
events of shared interest
• support for ubiquitous computing, including management of events emanating from the
ubiquitous infrastructure, e.g. location events
• a broad set of monitoring applications, including network monitoring in the Internet
• Types of pub-sub systems include:
• Channel Based – Publishers publish to named channels and subscribers subscribe to all events on a named channel.
• Type Based – Subscribers register interest in types of events and notifications occur when particular types of events occur.
• Topic Based – Subscribers register interest in particular topics and notifications occur when any information related to the topic arrives.
• Content Based – This is the most flexible of the schemes. Subscribers can specify interest is particular values or ranges of values for multiple attributes. Notifications are based on matching the attribute specification criteria.
• When and event matches a subscriber’s subscription then the system sends a notification that contains the event to the subscriber.
(School of Computing and InformaCtiOonMSPy9s0te0m15s ©DistTrihbeutUednivSeyrsstietymos f Melbourne) 2022 Semester II 13 / 25
Publish/Subscribe
Programming model
Advertise provides an additional mechanism for publishers to declare the nature of future events, i.e. the types of events of interest that may occur.
(School of Computing and InformaCtiOonMSPy9s0te0m15s ©DistTrihbeutUednivSeyrsstietymos f Melbourne) 2022 Semester II 14 / 25
Publish/Subscribe
Multi-server architecture
The Broker exchanges or routes information from publishers to subscribers.
(School of Computing and InformaCtiOonMSPy9s0te0m15s ©DistTrihbeutUednivSeyrsstietymos f Melbourne) 2022 Semester II 15 / 25
Publish/Subscribe
Overall System Architecture
(School of Computing and InformaCtiOonMSPy9s0te0m15s ©DistTrihbeutUednivSeyrsstietymos f Melbourne) 2022 Semester II 16 / 25
Publish/Subscribe
Examples of pub/sub systems
A modern example of a pub/sub system is Apache Kafka. Others are shown below.
(School of Computing and InformaCtiOonMSPy9s0te0m15s ©DistTrihbeutUednivSeyrsstietymos f Melbourne) 2022 Semester II 17 / 25
Publish/Subscribe
Discussion questions
Question (3): Discuss how the publish/subscribe paradigm could be used to implement a chat room application. Would this be better or worse than the message queue and or group communication paradigm?
(School of Computing and InformaCtiOonMSPy9s0te0m15s ©DistTrihbeutUednivSeyrsstietymos f Melbourne) 2022 Semester II 18 / 25
Distributed Shared Memory
Shared memory approaches
Distributed shared memory is an abstraction for sharing data between computers that do not share physical memory. Processes access DSM by reads and updates to what appears to be ordinary memory within their address space.
(School of Computing and InformaCtiOonMSPy9s0te0m15s ©DistTrihbeutUednivSeyrsstietymos f Melbourne) 2022 Semester II 19 / 25
Distributed Shared Memory Tuple Spaces
Tuple Spaces
The tuple space is a more abstract form of shared memory, compared to DSM.
(School of Computing and InformaCtiOonMSPy9s0te0m15s ©DistTrihbeutUednivSeyrsstietymos f Melbourne) 2022 Semester II 20 / 25
add(IField field); set(IField field, int index); get(int index);
Distributed Shared Memory Tuple Spaces
Example: The LighTS interface
Picco, Balzarotti, et. al., “LighTS: A Lightweight, Customizable Tuple Space Supporting Context-Aware Applications”
public interface ITupleSpace {
String getName(); // name of tuple space
void out(ITuple tuple); // put to tuple space void outg(ITuple[] tuples); // put to tuple space ITuple in(ITuple template); // blocking take ITuple inp(ITuple template); // non-blocking take ITuple[] ing(ITuple template); // blocking take ITuple rd(ITuple template); // blocking read ITuple rdp(ITuple template); // non-blocking read ITuple[] rdg(ITuple template); // blocking read int count(ITuple template); // count tuples
public interface ITuple {
ITuple insertAt(IField field, int index); ITuple removeAt(int index);
IField[] getFields();
int length();
boolean matches(ITuple tuple); }
public interface IField {
Class getType();
IField setType(Class classObj); boolean matches(IField field);
public interface IValuedField extends IField {
boolean isFormal(); // formal is a wildcard java.io.Serializable getValue();
IValuedField setValue(java.io.Serializable obj);
ITuple ITuple IField
ITupleSpace ts = new TupleSpace(“SAC05”);
IField f1 = new Field().setValue(“Paolo”);
IField f2 = new Field().setValue( new Integer (10)); ITuple t1 = new Tuple().add(f1).add(f2); ts.out(t1);
(School of Computing and InformaCtiOonMSPy9s0te0m15s ©DistTrihbeutUednivSeyrsstietymos f Melbourne)
2022 Semester II
Distributed Shared Memory Tuple Spaces Example York Linda Kernel
The implementation uses multiple Tuple Space Servers.
(School of Computing and InformaCtiOonMSPy9s0te0m15s ©DistTrihbeutUednivSeyrsstietymos f Melbourne) 2022 Semester II 22 / 25
Distributed Shared Memory Tuple Spaces
Discussion questions
Question (4): Discuss how the tuple space paradigm could be used to implement a chat room application. Compare this to message queue, group communication, and publish/subscribe.
(School of Computing and InformaCtiOonMSPy9s0te0m15s ©DistTrihbeutUednivSeyrsstietymos f Melbourne) 2022 Semester II 23 / 25
Comparison
(School of Computing and InformaCtiOonMSPy9s0te0m15s ©DistTrihbeutUednivSeyrsstietymos f Melbourne) 2022 Semester II 24 / 25
Comparison
Discussion questions
Question (5): Considering all of these indirect communication paradigms, which one, if any, would be more suitable for implementing a video conferencing system like Zoom? If none of them are suitable then what kind of paradigm/metaphor would be more suitable?
(School of Computing and InformaCtiOonMSPy9s0te0m15s ©DistTrihbeutUednivSeyrsstietymos f Melbourne) 2022 Semester II 25 / 25
程序代写 CS代考 加微信: powcoder QQ: 1823890830 Email: powcoder@163.com