PA160: Net-Centric Computing II. Distributed Systems
Luděk Matýska
Slides by: Tomáš Rebok Faculty of Informatics Masaryk University
Spring 2011
Luděk Matýska  (Fl MU)
3. Distributed Systems
Spring 2011        1 / 100
Lecture overview
Lecture overview
(l) Distributed Systems o Key characteristics o Challenges and Issues
Distributed System Architectures
Inter-process Communication
2J Middleware
o Remote Procedure Calls (RPC)
o Remote Method Invocation (RMI)
o Common Object Request Broker Architecture (CORBA)
(3^ Web Services
(4) Grid Services
Issues Examples
Scheduling/Load-balancing in Distributed Systems Fault Tolerance in Distributed Systems
Conclusion
Ludek Matyska  (FIMU) 3. Distributed Systems Spring 20ll 2/100
Distributed Systems
Lecture overview
(l) Distributed Systems o Key characteristics o Challenges and Issues
Distributed System Architectures
Inter-process Communication
[^1 Middleware
• Remote Procedure Calls (RPC)
• Remote Method Invocation (RMI)
• Common Object Request Broker Architecture (CORBA)
Web Services
V4f Grid Services
Fj>) Issues Examples
Scheduling/Load-balancing in Distributed Systems Fault Tolerance in Distributed Systems
[*6| Conclusion
Ludek Matyska  (FIMU) 3. Distributed Systems Spring 20ll 3/100
Distributed Systems
Distributed Systems - Definition
A system in which hardware and software components located at networked computers communicate and coordinate their actions only by message passing.
Distributed System by Tanenbaum and Steen
A collection of independent computers that appears to its uses as a single coherent system.
• the independent/autonomous machines are interconnected by communication networks and equipped with software systems designed to produce an integrated and consistent computing environment
Core objective of a distributed system: resource sharing
Luděk Matýska  (Fl MU)
3. Distributed Systems
Spring 2011        4 / 100
Distributed Systems      Key characteristics
Distributed Systems - Key characteristics
Key Characteristics of Distributed Systems:
o Autonomicity - there are several autonomous computational entities, each
of which has its own local memory o Heterogeneity - the entities may differ in many ways
• computer HW (different data types' representation), network
interconnection, operating systems (different APIs), programming languages (different data structures), implementations by different developers, etc. o Concurrency - concurrent (distributed) program execution and resource access
No global clock - programs (distributed components) coordinate their actions by exchanging messages
o message communication can be affected by delays, can suffer from variety of failures, and is vulnerable to security attacks Independent failures - each component of the system can fail independently, leaving the others still running (and possibly not informed about the failure)
o How to know/differ the states when a network has failed or became unusually slow?
o How to know if a remote server crashed immediately?
Ludek Matyska  (FIMU) 3. Distributed Systems Spring 2011 5/100
Distributed Systems     Challenges and Issues
Distributed Systems - Challenges and Issues
What do we want from a Distributed System (DS)? o Resource Sharing o Openness
Concurrency o Scalability o Fault Tolerance
Security
Transparency
3. Distributed Systems
Spring 2011        6 / 100
Distributed Systems     Challenges and Issues
Distributed Systems - Challenges and Issues
Resource Sharing and Openness
Resource Sharing
o main motivating factor for constructing DSs o is should be easy for the users (and applications) to access remote resources, and to share them in a controlled and efficient way
□ each resource must be managed by a software that provides interfaces
which enable the resource to be manipulated by clients o resource = anything you can imagine (e.g., storage facilities, data, files, Web pages, etc.) Openness
o whether the system can be extended and re-implemented in various ways and new resource-sharing services can be added and made available for use by a variety of client programs
o specification and documentation of key software interfaces must be published
o using an Interface Definition Language (IDL) involves HW extensibility as well
o i.e., the ability to add hardware from different vendors
Ludek Matyska  (FIMU) 3. Distributed Systems Spring 2011 7/100
Distributed Systems     Challenges and Issues
Distributed Systems - Challenges and Issues
Concurrency and Scalability Concurrency
o every resource in a DS must be designed to be safe in a concurrent environment
o applies not only to servers, but to objects in applications as well o ensured by standard techniques, like semaphores Scalability
o a DS is scalable if the cost of adding a user (or resource) is a constant amount in terms of resources that must be added
o and is able to utilize the extra hardware/software efficiently and remains manageable
Date	Comuuters Web	servers
1979, Dec.	188	0
1989, July	130,000	0
1999, July	56,218,000	5,560,866
2003,Jan	171,638,297	35,424,956
Ludek Matyska (FI MU) 3. Distributed Systems Spring 2011        8 / 100
Distributed Systems     Challenges and Issues
Distributed Systems - Challenges and Issues
Fault Tolerance and Security
Fault Tolerance
o a characteristic where a distributed system provides an appropriately handling of errors that occurred in the system
o the failures can be detected (sometimes hard or even impossible), masked (made hidden or less severe), or tolerated
o achieved by deploying two approaches: hardware redundancy and software recovery
Security
o involves confidentiality, integrity, authentication, and availability
3. Distributed Systems
Spring 2011        9 / 100
Distributed Systems     Challenges and Issues
Distributed Systems - Challenges and Issues
Transparency I.
Transparency
o certain aspects of the DS should be made invisible to the user / application programmer
o i.e., the system is perceived as a whole rather than a collection of independent components
o several forms of transparency.
o Access transparency - enables local and remote resources to be
accessed using identical operations o Location transparency - enables resources to be accessed without
knowledge of their location o Concurrency transparency - enables several processes to operate
concurrently using shared resources without interference between them o Replication transparency - enables multiple instances of resources to
be used to increase reliability and performance
o without knowledge of the replicas by users / application programmers
3. Distributed Systems
Spring 2011        10 / 100
Distributed Systems     Challenges and Issues
Distributed Systems - Challenges and Issues
Transparency II.
• forms of transparency cont'd.:
o Failure transparency - enables the concealment of faults, allowing users and application programs to complete their tasks despite of a failure of HW/SW components
Mobility/migration transparency - allows the movement of resources and clients within a system without affecting the operation of users or programs
o Performance transparency - allows the system to be reconfigured to improve performance as loads vary
Scaling transparency - allows the system and applications to expand in scale without changes to the system structure or application algorithms
3. Distributed Systems
Spring 2011        11 / 100
Distributed Systems      Distributed System Architectures
Distributed Systems - Architecture Models
An architecture model:
o defines the way in which the components of systems interact with one another, and
o defines the way in which the components are mapped onto an underlying network of computers
the overal goal is to ensure that the structure will meet present and possibly future demands
o the major concerns are to make system reliable, manageable, adaptable, and cost-effective o principal architecture models:
o client-server model - most important and most widely used o a service may be further provided by multiple servers o the servers may in turn be clients for another servers o proxy servers (caches) may be employed to increase availability and performance
o peer processes - all the processes play similar roles
o based either on structured (Chord, CAN, etc.), unstructured, or hybrid architectures
Ludek Matyska  (FIMU) 3. Distributed Systems Spring 2011 12/100
Distributed Systems     Distributed System Architectures
Distributed Systems - Architecture Models
Client-Server model
3. Distributed Systems
Spring 2011        13 / 100
Distributed Systems     Distributed System Architectures
Distributed Systems - Architecture Models
Client-Server model - A Service provided by Multiple Servers
3. Distributed Systems
Spring 2011        14 / 100
Distributed Systems      Distributed System Architectures
3. Distributed Systems
Spring 2011        15 / 100
Distributed Systems     Inter-process Communication
Distributed Systems - Inter-process Communication (IPC)
o the processes (components) need to communicate o the communication may be:
o synchronous - both send and receive are blocking operations o asynchronous - send is non-blocking and receive can have blocking (more common) and non-blocking variants
o the simplest forms of communication: UDP and TCP sockets
3. Distributed Systems
Spring 2011        16 / 100
Distributed Systems      Inter-process Communication
Distributed Systems - Inter-process Communication
UDP and TCP sockets
UDP/TCP sockets
o provide unreliable/reliable communication services
+ the complete control over the communication lies in the hands of applications
- too primitive to be used in developing a distributed system software
o higher-level facilities (marshalling/unmarshalling data, error detection, error recovery, etc.) must be built from scratch by developers on top of the existing socket primitive facilities
o force read/write mechanism instead of a procedure call
- another problem arises when the software needs to be used in a platform different from where it was developed
o the target platform may provide different socket implementation
== these issues are eliminated by the use of a Middleware
3. Distributed Systems
Spring 2011        17 / 100
Middleware
Lecture overview
Distributed Systems
• Key characteristics
• Challenges and Issues Distributed System Architectures Inter-process Communication
2) Middleware
o Remote Procedure Calls (RPC)
o Remote Method Invocation (RMI)
o Common Object Request Broker Architecture (CORBA)
3^ Web Services
4^ Grid Services
5^ Issues Examples
Scheduling/Load-balancing in Distributed Systems Fault Tolerance in Distributed Systems
6l Conclusion
3. Distributed Systems
Spring 2011        18 / 100
Middleware
Middleware Middleware
o a software layer that provides a programming abstraction as well as masks the heterogeneity of the underlying networks, hardware, operating systems, and programming languages o      provides transparency services
o represented by processes/objects that interact with each other to implement communication and resource sharing support provides building blocks for the construction of SW components that can work with one another
middleware examples: q Sun RPC (ONC RPC) o DCE RPC o MS COM/DCOM o Java RMI o CORBA o etc.
Ludek Matyska (FI MU)
3. Distributed Systems
Network Spring 2011
19 / 100
Middleware
Middleware
Basic Services
Basic services provided:
o Directory services - services required to locate application services and resources, and route messages o « service discovery o Data encoding services - uniform data representation services for dealing with incompatibility problems on remote systems o e.g., Sun XDR, ISO's ASN.1, CORBA's CDR, XML, etc. data marshalling/unmarshalling
o Security services - provide inter-application client-server security mechanisms
o Time services - provide a universal format for representing time on different platforms (possibly located in various time zones) in order to keep synchronisation among application processes
o Transaction services - provide transaction semantics to support commit, rollback, and recovery mechanisms
Ludek Matyska  (FIMU) 3. Distributed Systems Spring 2011        20 / 100
Middleware
Middleware
Basic Services - A Need for Data Encoding Services
Data encoding services are required, because remote machines may have:
o different byte ordering o different sizes of integers and other types o different floating point representations o different character sets alignment requirements
main() {
unsigned int n; char *a =  (char *)&n;
n = 0x11223344;
printf("%02x,   %02x, %02x,  %02x\n",
a[0],  a[l], a[2], a[3]);
}
Output on a Pentium: 44,  33,  22, 11
Output on a G4:
11,  22,  33, 44
3. Distributed Systems
Spring 2011       21 / 100
Middleware      Remote Procedure Calls (RPC)
Remote Procedure Calls (RPC)
Remote Procedure Calls (RPC)
o very simple idea similar to a well-known procedure call mechanism o a client sends a request and blocks until a remote server sends a response
o the goal is to allow distributed programs to be written in the same style as conventional programs for centralised computer systems o while being transparent - the programmer need not be aware that the called procedure is executing on a local or a remote computer o the idea:
o the remote procedure is represented as a stub on the client side
o behaves like a local procedure, but rather than placing the parameters into registers, it packs them into a message, issues a send primitive, and blocks itself waiting for a reply o the server passes the arrived message to a server stub (known as skeleton as well)
the skeleton unpacks the parameters and calls the procedure in a conventional manner
the results are returned to the skeleton, which packs them into a message directed to the client stub
Ludek Matyska  (FIMU) 3. Distributed Systems Spring 2011        22 / 100
Middleware     Remote Procedure Calls (RPC)
Remote Procedure Calls (RPC)
3. Distributed Systems
Spring 2011       23 / 100
Middleware     Remote Procedure Calls (RPC)
Remote Procedure Calls (RPC)
The remote procedure call in detail
Communication network
3. Distributed Systems
Spring 2011       24 / 100
Middleware     Remote Procedure Calls (RPC)
Remote Procedure Calls (RPC)
Components
module module dispatcher
O client program, client stub
O communication modules
O server stub, service procedure
O dispatcher - selects one of the server stub procedures according to the procedure identifier in the request message
Sun RPC: the procedures are identified by:
o program number - can be obtained from a central authority to allow every
program to have its own unique number o procedure number - the identifier of the particular procedure within the program o version number - changes when a procedure signature changes
Ludek Matyska  (FI MU) 3. Distributed Systems Spring 2011        25 / 100
Middleware     Remote Procedure Calls (RPC)
Remote Procedure Calls (RPC)
Location Services - Portmapper
Clients need to know the port number of a service provided by the server == Portmapper
o a server registers its program#, version#, and port# to the local portmapper
o a client finds out the port# by sending a request o the portmapper listens on a well-known port (111) o the particular procedure required is identified in the subsequent procedure call
3. Distributed Systems
Spring 2011       26 / 100
Middleware      Remote Procedure Calls (RPC)
Remote Procedure Calls (RPC)
Parameters passing
How to pass parameters to remote procedures?
o pass by value - easy: just copy data to the network message o pass by reference - makes no sense without shared memory
Pass by reference: the steps P> copy referenced items (marshalled) to a message buffer ^> ship them over, unmarshal data at server ^> pass local pointer to server stub function send new values back
o to support complex structures:
copy the structure into pointerless representation o transmit
reconstruct the structure with local pointers on the server
Ludek Matyska  (FIMU) 3. Distributed Systems Spring 2011        27 / 100
Middleware      Remote Procedure Calls (RPC)
Remote Procedure Calls (RPC)
Parameters passing - eXternal Data Representation (XDR)
Sun RPC: to avoid compatibility problems, the eXternal Data Representation (XDR) is used
o XDR primitive functions examples:
o xdr_int(), xdr_char(), xdr_u_short(), xdr_bool(), xdr_long(), xdr_u_int() , xdr_wrapstring() , xdr_short(), xdr_enum(), xdr_void()
XDR aggregation functions:
o xdr_array(), xdr_string(), xdr_union(), xdr_vector(), xdr_opaque()
o only a single input parameter is allowed in a procedure call
o == procedures requiring multiple parameters must include them as components of a single structure
3. Distributed Systems
Spring 2011       28 / 100
Middleware      Remote Procedure Calls (RPC)
Remote Procedure Calls (RPC)
When Things Go Wrong I.
o local procedure calls do not fail
o if they core dump, entire process dies o there are more opportunities for errors with RPC
server could generate an error o problems in network (lost/delayed requests/replies)
server crash
o client might crash while server is still executing code for it transparency breaks here
o applications should be prepared to deal with RPC failures
Ludek Matyska  (FIMU) 3. Distributed Systems Spring 2011        29 / 100
Middleware      Remote Procedure Calls (RPC)
Remote Procedure Calls (RPC)
When Things Go Wrong II.
Semantics of local procedure calls: exactly once
o difficult to achieve with RPC Four remote calls semantics available in RPC:
o at-least-once semantic
□ client keeps trying sending the message until a reply has been received
• failure is assumed after n re-sends • guarantees that the call has been made "at least once", but possibly multiple times
o ideal for idempotent operations at-most-once semantic
client gives up immediately and reports back a failure
guarantees that the call has been made "at most once", but possibly none at
all
exactly-once semantic
o the most desirable, but the most difficult to implement o maybe semantic
no message delivery guarantees are provided at all o (easy to implement)
Ludek Matyska  (FIMU) 3. Distributed Systems Spring 2011        30 / 100
Middleware     Remote Procedure Calls (RPC)
Remote Procedure Calls (RPC)
When Things Go Wrong III.
Figure: Message-passing semantics. (a) at-least-once; (b) exactly-once.
3. Distributed Systems
Spring 2011       31 / 100
Middleware    Remote Procedure Calls (RPC)
Remote Procedure Calls (RPC)
When Things Go Wrong IV. - Complications
Tí it is necessary to understand the application
o idempotent functions - in the case of a failure, the message may be
retransmitted and re-run without a harm q non-idempotent functions - has side-effects == the retransmission has to be controlled by the server
o the duplicity request (retransmission) has to be detected
• once detected, the server procedure is NOT re-run; just the results are resent (if available in a server cache)
y in the case of a server crash, the order of execution vs. crash matters
REG
Mo REP
	REQ	
Receive	-	Receive
Execute		1 Crash 1
. iCrashl	No REP	
in the case of a client crash, the procedure keeps running on the server
□ consumes resources (e.g., CPU time), possesess resources (e.g., locked files), etc.
o may be overcome by employing soft-state principles • keep-alive messages
Ludek Matyska  (FI MU)
S. Distributed Systems
Spring 2011
S2 / 100
Middleware      Remote Procedure Calls (RPC)
Remote Procedure Calls (RPC)
Code Generation I.
o RPC drawbacks.
o complex API, not easy to debug o the use of XDR is difficult
o but, it's often used in a similar way
o == the server/client code can be automatically generated
o assumes well-defined interfaces (IDL)
the application programmer has to supply the following.
o interface definition file - defines the interfaces (data structures, procedure names, and parameters) of the remote procedures that are offered by the server o client program - defines the user interfaces, the calls to the remote
procedures of the server, and the client side processing functions o server program - implements the calls offered by the server compilers.
o rpcgen for C/C++, jrpcgen for Java
3. Distributed Systems
Spring 2011     33 / 100
Middleware      Remote Procedure Calls (RPC)
Remote Procedure Calls (RPC)
Code Generation II.
3. Distributed Systems
Spring 2011        34 / 100
Middleware      Remote Method Invocation (RMI)
Remote Method Invocation (RMI)
o as the popularity of object technology increased, techniques were developed to allow calls to remote objects instead of remote procedures only
== Remote Method Invocation (RMI)
essentially the same as the RPC, except that it operates on objects instead of applications/procedures
o the RMI model represents a distributed object application
□ it allows an object inside a JVM (a client) to invoke a method on an object running on a remote JVM (a server) and have the results returned to the client o the server application creates an object and makes it accesible remotely (i.e., registers it)
the client application receives a reference to the object on the server and invokes methods on it
o the reference is obtained through looking up in the registry o important: a method invocation on a remote object has the same syntax as a method invocation on a local object
Ludek Matyska  (FIMU) 3. Distributed Systems Spring 2011        35 / 100
Middleware     Remote Method Invocation (RMI)
Remote Method Invocation (RMI)
Architecture I.
The interface, through which the client and server interact, is (similarly to RPC) provided by stubs and skeletons:
3. Distributed Systems
Spring 2011        36 / 100
Middleware     Remote Method Invocation (RMI)
Remote Method Invocation (RMI)
Architecture II.
Two fundamental concepts as the heart of distributed object model:
o remote object reference - an identifier that can be used throughout a distributed system to refer to a particular unique remote object its construction must ensure its uniqueness
32 bits	32 bits	32 bits	32 bits	
Internet address	port number	time	object number	interface of remote object
o remote interface - specifies, which methods of the particular object
can be invoked remotely
Spring 2011        37 / 100
Middleware      Remote Method Invocation (RMI)
Remote Method Invocation (RMI)
Architecture III.
o the remote objects can be accessed concurrently
the encapsulation allows objects to provide methods for protecting themselves against incorrect accesses
□ e.g., synchronization primitives (condition variables, semaphores, etc.)
o RMI transaction semantics similar to the RPC ones
o at-least-once, at-most-once, exactly-once, and maybe semantics data encoding services:
stubs use Object Serialization to marshal the arguments
• object arguments' values are rendered into a stream of bytes that can
be transmitted over a network o == the arguments must be primitive types or objects that implement Serializable interface
parameters passing:
local objects passed by value remote objects passed by reference
3. Distributed Systems
Spring 2011       38 / 100
Middleware      Common Object Request Broker Architecture (CORBA)
Common Object Request Broker Architecture (CORBA)
Common Object Request Broker Architecture (CORBA)
o an industry standard developed by the OMG (Object Management Group - a consortium of more than 700 companies) to aid in distributed objects programming
o OMG was established in 1988
o initial CORBA specification came out in 1992
o but significant revisions have taken place from that time
provides a platform-independent and language-independent architecture (framework) for writing distributed, object-oriented applications
o i.e., application programs can communicate without restrictions to: programming languages, hardware platforms, software platforms, networks they communicate over o but CORBA is just a specification for creating and using distributed objects; it is not a piece of software or a programming language • several implementations of the CORBA standard exist (e.g., IBM's SOM and DSOM architectures)
Ludek Matyska  (FIMU) 3. Distributed Systems Spring 2011        39 / 100
Middleware      Common Object Request Broker Architecture (CORBA)
CORBA Components
CORBA is composed of five major components: o Object Request Broker (ORB) o Interface Definition Language (IDL) o Dynamic Invocation Interface (DII)
Interface Repositories (IR) o Object Adapters (OA)
3. Distributed Systems
Spring 2011        40 / 100
Middleware      Common Object Request Broker Architecture (CORBA)
CORBA Components
Object Request Broker (ORB) I.
Object Request Broker (ORB)
o the heart of CORBA
• introduced as a part of OMG's Object Management Architecture (OMA), which the CORBA is based on
o a distributed service that implements all the requests to the remote object(s)
o it locates the remote object on the network, communicates the request to the object, waits for the results and (when available) communicates those results back to the client
o implements location transparency
o exactly the same request mechanism is used regardless of where the object is located
might be in the same process with the client or across the planet
o implements programming language independence
o the client issuing a request can be written in a different programming language from the implementation of the CORBA object both the client and the object implementation are isolated from the ORB by an IDL interface
o Internet Inter-ORB Protocol (IIOP) - the standard communication protocol between ORBs
Ludek Matyska  (FIMU) 3. Distributed Systems Spring 2011 41/100
Middleware      Common Object Request Broker Architecture (CORBA)
CORBA Components
Object Request Broker (ORB) II.
Client                         Object Implementation                            Client                      Object Implementation									
	;IDL	n nr	1DL		IDL	n		IDL*	
t-......._.      ORB ORB									
									
NETWORK									
Ludek Matyska  (FIMU) 3. Distributed Systems Spring 2011        42 / 100
Middleware      Common Object Request Broker Architecture (CORBA)
CORBA Components
Interface Definition Language (IDL) I.
Interface Definition Language (IDL)
o as with RMI, CORBA objects have to be specified with interfaces
o interface ~ a contract between the client (code using a object) and the server (code implementing the object)
o indicates a set of operations the object supports and how they should be invoked (but NOT how they are implemented)
o defines modules, interfaces, types, attributes, exceptions, and method signatures
o uses same lexical rules as C++
o with additional keywords to support distribution (e.g. interface, any, attribute, in, out, inout, readonly, raises)
o defines language bindings for many different programming languages (e.g., C/C++, Java, etc.)
o via language mappings, the IDL translates to different constructs in the different implementation languages
it allows an object implementor to choose the appropriate programming language for the object, and
it allows the developer of the client to choose the appropriate and possibly different programming language for the client
Ludek Matyska  (FIMU) 3. Distributed Systems Spring 2011        43 / 100
Middleware      Common Object Request Broker Architecture (CORBA)
CORBA Components
Interface Definition Language (IDL) II.
Interface Definition Language (IDL) example:
module StockObjects { struct Quote {
string symbol; long at_time; double price; long volume;
};
exception Unknown{}; interface Stock {
// Returns the current stock quote.
Quote get_quote() raises(Unknown);
// Sets the current stock quote.
void set_quote(in Quote stock_quote);
// // Provides the stock description, e.g. company name. readonly attribute string description;
};
interface StockFactory {
Stock create_stock(in string symbol, in string description);
};
};
Ludek Matyska  (FIMU) 3. Distributed Systems Spring 2011        44 / 100
Middleware      Common Object Request Broker Architecture (CORBA)
IDL compiler automatically compiles the IDL into client stubs and object skeletons:
3. Distributed Systems
Spring 2011        45 / 100
Middleware      Common Object Request Broker Architecture (CORBA)
CORBA Components
Interface Definition Language (IDL) IV. - Development Process Using IDL
3. Distributed Systems
Spring 2011       46 / 100
Middleware      Common Object Request Broker Architecture (CORBA)
CORBA Components
Dynamic Invocation Interface (DII) & Dynamic Skeleton Interface (DSI)
Dynamic Invocation Interface (DII)
o CORBA supports both the dynamic and the static invocation interfaces
o static invocation interfaces are determined at compile time o dynamic interfaces allow client applications to use server objects without knowing the type of those objects at compile time
o DII - an API which allows dynamic construction of CORBA object
invocations
Dynamic Skeleton Interface (DSI)
• DSI is the server side's analogue to the client side's DII
o allows an ORB to deliver requests to an object implementation that does not have compile-time knowledge of the type of the object it is implementing
3. Distributed Systems
Spring 2011       47 / 100
Middleware      Common Object Request Broker Architecture (CORBA)
CORBA Components
Interface Repository (IR)
Interface Repository (IR)
o a runtime component used to dynamically obtain information on IDL types (e.g. object interfaces)
using the IR, a client should be able to locate an object that is unknown at compile time, find information about its interface, and build a request to be forwarded through the ORB
this kind of information is necessary when a client wants to use the DII to construct requests dynamically
3. Distributed Systems
Spring 2011       48 / 100
Middleware      Common Object Request Broker Architecture (CORBA)
CORBA Components
Object Adapters (OAs)
Object Adapters (OAs)
o the interface between the ORB and the server process
o OAs listen for client connections/requests and map the inbound
requests to the desired target object instance o provide an API that object implementations use for:
generation and interpretation of object references
method invocation
security of interactions
object and implementation activation and deactivation mapping object references to the corresponding object implementations registration of implementations two basic kinds of OAs:
o basic object adapter (BOA) - leaves many features unsupported, requiring proprietary extensions
portable object adapter (POA) - intended to support multiple ORB implementations (of different vendors), allow persistent objects, etc.
Ludek Matyska  (FIMU) 3. Distributed Systems Spring 2011        49 / 100
Middleware      Common Object Request Broker Architecture (CORBA)
CORBA Object & Object Reference
CORBA Objects are fully encapsulated
o accessed through well-defined interfaces only
o interfaces & implementations are totally separate
for one interface, multiple implementations possible
one implementation may be supporting multiple interfaces
IDL Interface '"""^^^^
CORBA Object Reference is the distributed computing equivalent of a pointer o CORBA defines the Interoperable Object Reference (IOR)
o an IOR contains a fixed object key, containing:
• the object's fully qualified interface name (repository ID)
• user-defined data for the instance identifier o can also contain transient information:
the host and port of its server, metadata about the server's ORB (for potential optimizations), etc. == the IOR uniquely identifies one object instance
Ludek Matyska  (FI MU) 3. Distributed Systems Spring 2011
50 / 100
CORBA Services
Middleware      Common Object Request Broker Architecture (CORBA)
CORBA Services (COS)
o the OMG has defined a set of Common Object Services to support the integration and interoperation of distributed objects
o = frequently used components needed for building robust applications o typically supplied by vendors
o OMG defines interfaces to services to ensure interoperability
3. Distributed Systems
Spring 2011        51 / 100
Middleware      Common Object Request Broker Architecture (CORBA)
CORBA Services
Popular Services Example
Service	Description
Object life cycle	Defines how CORBA objects are created, removed, moved, and copied
Naming	Defines how CORBA objects can have friendly symbolic names
Events	Decouples the communication between distributed objects
Relationships	Provides arbitrary typed n-ary relationships between CORBA objects
Externalization	Coordinates the transformation of CORBA objects to and from external media
Transactions	Coordinates atomic access to CORBA objects
Concurrency Control	Provides a locking service for CORBA objects in order to ensure serializable access
Property	Supports the association of name-value pairs with CORBA objects
Trader	Supports the finding of CORBA objects based on properties describing the service offered by the object
Query	Supports queries on objects
Ludek Matyska  (FIMU) 3. Distributed Systems Spring 2011        52 / 100
Middleware      Common Object Request Broker Architecture (CORBA)
CORBA Architecture Summary
3. Distributed Systems
Spring 2011        53 / 100
Web Services
Lecture overview
Distributed Systems
• Key characteristics
• Challenges and Issues Distributed System Architectures Inter-process Communication
2f Middleware
• Remote Procedure Calls (RPC)
• Remote Method Invocation (RMI)
• Common Object Request Broker Architecture (CORBA)
3) Web Services
4^ Grid Services
5^ Issues Examples
Scheduling/Load-balancing in Distributed Systems Fault Tolerance in Distributed Systems
6l Conclusion
3. Distributed Systems
Spring 2011       54 / 100
Grid Services
Lecture overview
Distributed Systems
• Key characteristics
• Challenges and Issues Distributed System Architectures Inter-process Communication
Middleware
• Remote Procedure Calls (RPC)
• Remote Method Invocation (RMI)
• Common Object Request Broker Architecture (CORBA)
Web Services
(4) Grid Services
Issues Examples
Scheduling/Load-balancing in Distributed Systems Fault Tolerance in Distributed Systems
['6| Conclusion
3. Distributed Systems
Spring 2011       55 / 100
Issues Examples
Lecture overview
Distributed Systems
• Key characteristics
• Challenges and Issues Distributed System Architectures Inter-process Communication
2f Middleware
• Remote Procedure Calls (RPC)
• Remote Method Invocation (RMI)
• Common Object Request Broker Architecture (CORBA)
3^ Web Services
4^ Grid Services
5^ Issues Examples
Scheduling/Load-balancing in Distributed Systems Fault Tolerance in Distributed Systems
6l Conclusion
3. Distributed Systems
Spring 2011       56 / 100
Issues Examples     Scheduling/Load-balancing in Distributed Systems
Scheduling/Load-balancing in Distributed Systems I.
For concurrent execution of interacting processes:
o communication and synchronization between processes are the
two essential system components
Before the processes can execute, they need to be: scheduled and allocated with resources
Why scheduling in distributed systems is of special interest?
o because of the issues that are different from those in traditional multiprocessor systems:
o the communication overhead is significant o the effect of underlying architecture cannot be ignored the dynamic behaviour of the system must be addressed
o local scheduling (on each node) + global scheduling
3. Distributed Systems
Spring 2011        57 / 100
Issues Examples     Scheduling/Load-balancing in Distributed Systems
Scheduling/Load-balancing in Distributed Systems II.
Scheduling/Load-balancing in Distributed Systems
• let's have a pool of jobs
• there are some inter-dependencies among them
• let's have a set of nodes (processors), which are able to reciprocally communicate
Load-balancing
The term load-balancing means assigning the jobs to the processors in the way, which minimizes the time/communication overhead necessary to compute them.
• load-balancing - divides the jobs among the processors
• scheduling - defines, in which order the jobs have to be executed (on each processor)
• load-balancing and planning are tightly-coupled (usually considered as synonyms in DSs)
• objectives:
• enhance overall system performance metric
• process completion time and processor utilization
• location and performance transparency
atyska  (FIMU) 3. Distributed Systems Spring 2011        58 / 100
Issues Examples     Scheduling/Load-balancing in Distributed Systems
Scheduling/Load-balancing in Distributed Systems III.
o the scheduling/load-balancing task can be represented using graph theory:
the pool of N jobs with dependencies can be described as a graph G(V, U), where
o the nodes represent the jobs (processes)
o the edges represent the dependencies among the jobs/processes (e.g., an edge from i to j requires that the process i has to complete before j can start executing)
the graph G has to be splitted into p parts, so that:
o N = N1 U N2 U • • • U Np
o \N; | is the number of jobs assigned to the processor i, and op is the number of processors, and
the number/cost of the edges connecting the parts is minimal o the objectives:
• uniform jobs' load-balancing
o minimizing the communication (the minimal number of edges among
a
which satisfy the condition, that
where
the parts)
o the splitting problem is NP-complete
Ludek Matyska  (FI MU) 3. Distributed System
Spring 2011
59 / 100
Issues Examples     Scheduling/Load-balancing in Distributed Systems
Scheduling/Load-balancing in Distributed Systems III.
An illustration
(a) Precedence (b) Communication (c) Disjoint
process model process model process model
Figure: An illustration of splitting 4 jobs onto 2 processors.
3. Distributed Systems
Spring 2011        60 / 100
Issues Examples     Scheduling/Load-balancing in Distributed Systems
Scheduling/Load-balancing in Distributed Systems IV.
• the "proper" approach to the schedulign/load-balancing problem depends on the following criteria: o jobs' cost
o dependencies among the jobs o jobs' locality
3. Distributed Systems
Spring 2011        61 / 100
Issues Examples     Scheduling/Load-balancing in Distributed Systems
Scheduling/Load-balancing in Distributed Systems IV.
Jobs' Cost
• the job's cost may be known:
• before the whole problem set's execution
• during problem's execution, but before the particular job's execution o just after the particular job finishes
o cost's variability - all the jobs may have (more or less) the same cost or the costs may differ
• the problem classes based on jobs' cost:
o all the jobs have the same cost: easy
o the costs are variable, but, known: more complex
o the costs are unknown in advance: the most complex
3. Distributed Systems
Spring 2011       62 / 100
Issues Examples     Scheduling/Load-balancing in Distributed Systems
Scheduling/Load-balancing in Distributed Systems IV.
Dependencies Among the Jobs
• is the order of jobs' execution important?
o the dependencies among the jobs may be known:
• before the whole problem set's execution
• during problem's execution, but before the particular job's execution o are fully dynamic
• the problem classes based on jobs' dependencies:
o the jobs are fully independent on each other: easy o the dependencies are known or predictable: more complex o flooding
o in-trees, out-trees (balanced or unbalanced) o generic oriented trees (DAG) o the dependencies dynamically change: the most complex
e.g., searching/lookup problems
3. Distributed Systems
Spring 2011       63 / 100
Issues Examples     Scheduling/Load-balancing in Distributed Systems
Scheduling/Load-balancing in Distributed Systems IV.
Locality
o communicate all the jobs in the same/similar way?
• is it suitable/necessary to execute some jobs "close" to each other?
• when the job's communication dependencies are known?
• the problem classes based on jobs' locality:
o the jobs do not communicate (at most during initialization): easy o the communications are known/predictable: more complex
o regular (e.g., a grid) or irregular o the communications are unknown in advance: the most complex
• e.g., a discrete events' simulation
3. Distributed Systems
Spring 2011       64 / 100
Issues Examples     Scheduling/Load-balancing in Distributed Systems
Scheduling/Load-balancing in DSs - Solving Methods
• in general, the "proper" solving method depends on the time, when the particular information is known
• basic solving algorithms' classes:
o static - offline algorithms
□ semi-static - hybrid approaches
o dynamic - online algorithms
some (but not all) variants: o static load-balancing o semi-static load-balancing o self-scheduling o distributed queues o DAG planning
3. Distributed Systems
Spring 2011       65 / 100
Issues Examples     Scheduling/Load-balancing in Distributed Systems
Scheduling/Load-balancing in DSs - Solving Methods
Semi-static load-balancing
Semi-static load-balancing
o suitable for problem sets with slow changes in parameters, and with
locality importance o iterative approach
o uses static algorithm
o the result (from the static algorithm) is used for several steps (slight unbalance is accepted)
after the steps, the problem set is recalculated with the static algorithm again
often used for:
particle simulation o calculations of slowly-changing grids (but in a different sense than in the previous lectures)
3. Distributed Systems
Spring 2011     66 / 100
Issues Examples     Scheduling/Load-balancing in Distributed Systems
Scheduling/Load-balancing in DSs - Solving Methods
Self-scheduling I.
Self-scheduling
o a centralized pool of jobs
o idle processors pick the jobs from the pool
new (sub)jobs are added to the pool + ease of implementation o suitable for:
a set of independent jobs
jobs with unknown costs
jobs where locality does not matter
o unsuitable for too small jobs - due to the communication overhead o == coupling jobs into bulks
o fixed size
o controlled coupling
o tapering
weighted distribution
Ludek Matyska  (FIMU) 3. Distributed Systems Spring 2011        67 / 100
Issues Examples     Scheduling/Load-balancing in Distributed Systems
Scheduling/Load-balancing in DSs - Solving Methods
Self-scheduling II. - Fixed size & Controlled coupling
Fixed size
o typical offline algorithm
o requires much information (number and cost of each job, ...) o it is possible to find the optimal solution theoretically important, not suitable for practical solutions
Controlled coupling o uses bigger bulks in the beginning of the execution, smaller bulks in the end of the execution
o lower overhead in the beginning, finer coupling in the end
• the bulk's size is computed as:     = [R] where:
Ri . . . the number of remaining jobs op ... the number of processors
Ludek Matyska  (FIMU) 3. Distributed Systems Spring 2011        68 / 100
Issues Examples     Scheduling/Load-balancing in Distributed Systems
Scheduling/Load-balancing in DSs - Solving Methods
Self-scheduling II. - Tapering & Weighted distribution
Tapering
• analogical to the Controlled coupling, but the bulks' size is further a function of jobs' variation
o uses historical information
o low variance == bigger bulks o high variance == smaller bulks
Weighted distribution
• considers the nodes' computational power o suitable for heterogenous systems
o uses historical information as well
3. Distributed Systems
Spring 2011        69 / 100
Issues Examples     Scheduling/Load-balancing in Distributed Systems
Scheduling/Load-balancing in DSs - Solving Methods
Distributed Queues
Distributed Queues
o ?« self-scheduling for distributed memory
o instead of a centralized pool, a queue on each node is used (per-processor queues) suitable for:
distributed systems, where the locality does not matter for both static and dynamic dependencies for unknown costs
o an example: diffuse approach
o in every step, the cost of jobs remaining on each processor is computed o processors exchange this information and perform the balancing o locality must not be important
3. Distributed Systems
Spring 2011        70 / 100
Issues Examples     Scheduling/Load-balancing in Distributed Systems
Scheduling/Load-balancing in DSs - Solving Methods
Centralised Pool vs. Distributed Queues
Figure: Centralised Pool (left) vs. Distributed Queues (right).
3. Distributed Systems
Spring 2011        71 / 100
Issues Examples     Scheduling/Load-balancing in Distributed Systems
Scheduling/Load-balancing in DSs - Solving Methods
DAG Planning
DAG Planning
another graph model
the nodes represent the jobs (possibly weighted) o the edges represent the dependencies and/or the communication (may be also weighted)
o e.g., suitable for digital signal processing
o basic strategy - divide the DAG so that the communication and the processors' occupation (time) is minimized o NP-complete problem
o takes the dependencies among the jobs into account
3. Distributed Systems
Spring 2011       72 / 100
Issues Examples     Scheduling/Load-balancing in Distributed Systems
Scheduling/Load-balancing in DSs - Design Issues I.
When the scheduling/load-balancing is necessary? for middle-loaded systems
• lowly-loaded systems - rarely job waiting (there's always an idle processor)
o highly-loaded systems - little benefit (the load-balancing cannot help) What is the performance metric?
mean response time What is the measure of load? must be easy to measure o must reflect performance improvement o example: queue lengths at CPU, CPU utilization
3. Distributed Systems
Spring 2011     73 / 100
Issues Examples     Scheduling/Load-balancing in Distributed Systems
Scheduling/Load-balancing in DSs - Design Issues I.
Components
Types of policies:
o static (decisions hardwired into system), dynamic (uses load information), adaptive (policy varies according to load) Policies:
o Transfer policy: when to transfer a process?
threshold-based policies are common and easy Selection policy: which process to transfer? prefer new processes
transfer cost should be small compared to execution cost o == select processes with long execution times
Location policy: where to transfer the process?
polling, random, nearest neighbor, etc. Information policy: when and from where?
o demand driven (only a sender/receiver may ask for), time-driven
(periodic), state-change-driven (send update if load changes)
Ludek Matyska  (FIMU) 3. Distributed Systems Spring 2011        74 / 100
Issues Examples     Scheduling/Load-balancing in Distributed Systems
Scheduling/Load-balancing in DSs - Design Issues II.
Sender-initiated Policy
Sender-initiated Policy
o Transfer policy
o Selection policy: newly arrived process o Location policy: three variations
□ Random - may generate lots of transfers
o == necessary to limit max transfers o Threshold - probe n nodes sequentially
o transfer to the first node below the threshold, if none, keep job o Shortest - poll Np nodes in parallel o choose least loaded node below T if none, keep the job
Ludek Matyska  (FI MU)
3. Distributed Systems
Spring 2011
75 / 100
Issues Examples     Scheduling/Load-balancing in Distributed Systems
Scheduling/Load-balancing in DSs - Design Issues II.
Receiver-initiated Policy
Receiver-initiated Policy
o Transfer policy: if departing process causes load < T, find a process from elsewhere
o Selection policy: newly arrived or partially executed process
Location policy:
o Threshold - probe up to Np other nodes sequentially
o transfer from first one above the threshold; if none, do nothing o Shortest - poll n nodes in parallel
o choose the node with heaviest load above T
3. Distributed Systems
Spring 2011       76 / 100
Issues Examples     Scheduling/Load-balancing in Distributed Systems
Scheduling/Load-balancing in DSs - Design Issues II.
Symmetric Policy
Symmetric Policy
o combines previous two policies without change □ nodes act as both senders and receivers
uses average load as the threshold
3. Distributed Systems
Spring 2011     77 / 100
Issues Examples     Scheduling/Load-balancing in Distributed Systems
Scheduling/Load-balancing in DSs - Case study
V-System (Stanford)
V-System (Stanford)
state-change driven information policy
o significant change in CPU/memory utilization is broadcast to all other nodes
o M least loaded nodes are receivers, others are senders o sender-initiated with new job selection policy o Location policy:
probe random receiver
if still receiver (below the threshold), transfer the job otherwise try another
3. Distributed Systems
Spring 2011       78 / 100
Issues Examples     Scheduling/Load-balancing in Distributed Systems
Scheduling/Load-balancing in DSs - Case study
Sprite (Berkeley) I.
Sprite (Berkeley)
Centralized information policy: coordinator keeps info o state-change driven information policy
o Receiver: workstation with no keyboard/mouse activity for the defined time period (30 seconds) and below the limit (active processes < number of processors)
o Selection policy: manually done by user == workstation becomes sender
Location policy: sender queries coordinator
the workstation with the foreign process becomes sender if user becomes active
3. Distributed Systems
Spring 2011        79 / 100
Issues Examples     Scheduling/Load-balancing in Distributed Systems
Scheduling/Load-balancing in DSs - Case study
Sprite (Berkeley) II.
Sprite (Berkeley) cont'd, o Sprite process migration:
o facilitated by the Sprite file system o state transfer:
o swap everything out
o send page tables and file descriptors to the receiver o create/establish the process on the receiver and load the necessary pages
pass the control o the only problem: communication-dependencies
solution: redirect the communication from the workstation to the receiver
3. Distributed Systems
Spring 2011        80 / 100
Issues Examples     Scheduling/Load-balancing in Distributed Systems
Scheduling/Load-balancing in DSs - Code and Process Migration
Code and Process Migration
o key reasons: performance and flexibility o flexibility:
o dynamic configuration of distributed system • clients don't need preinstalled software (download on demand) o process migration (strong mobility) o process = code + data + stack o examples: Condor, DQS o code migration (weak mobility)
transferred program always starts from its initial state o migration in heterogeneous systems:
only weak mobility is supported in common systems (recompile code, no run time information)
the virtual machines may be used: interprets (scripts) or intermediate code (Java)
Ludek Matyska  (FIMU) 3. Distributed Systems Spring 2011 81/100
Issues Examples      Fault Tolerance in Distributed Systems
Fault Tolerance in Distributed Systems I.
o single machine systems
failures are all or nothing
o OS crash, disk failures, etc. o distributed systems: multiple independent nodes
partial failures are also possible (some nodes fail)
probability of failure grows with number of independent components
(nodes) in the system
o fault tolerance: system should provide services despite faults
o transient faults o ntermittent faults o ermanent faults
3. Distributed Systems
Spring 2011       82 / 100
Issues Examples     Fault Tolerance in Distributed Systems
Fault Tolerance in Distributed Systems I.
Failure Types
Type of failure	Description
Crash failure	A server halts, but is working correctly until it halts
Omission failure Receive omission Send omission	A server fails to respond to incoming requests A server fails to receive incoming messages A server fails to send messages
Timing failure	A server's response lies outside the specified time interval
Response failure Value failure State transition failure	The server's response is incorrect The value of the response is wrong The server deviates from the correct flow of control
Arbitrary failure	A server may produce arbitrary responses at arbitrary times
Luděk Matýska  (Fl MU)
3. Distributed Systems
Spring 2011        83 / 100
Issues Examples      Fault Tolerance in Distributed Systems
Fault Tolerance in Distributed Systems II.
o handling faulty processes: organize several processes into a group o all processes perform the same computation o all messages are sent to all the members of the particular group o majority needs to agree on results of a computation o ideally, multiple independent implementations of the application are desirable (to prevent identical bugs)
use process groups to organize such processes
3. Distributed Systems
Spring 2011        84 / 100
Issues Examples     Fault Tolerance in Distributed Systems
3. Distributed Systems
Spring 2011       85 / 100
Issues Examples      Fault Tolerance in Distributed Systems
Fault Tolerance in DSs - Agreement in Faulty Systems
o How should processes agree on results of a computation? o K-fault tolerant: system can survive k faults and yet function o assume processes fail silently
o == need (k + 1) redundancy to tolerant k faults
o Byzantine failures: processes run even if sick
o produce erroneous, random or malicious replies o byzantine failures are most difficult to deal with
3. Distributed Systems
Spring 2011       86 / 100
Issues Examples      Fault Tolerance in Distributed Systems
Fault Tolerance in DSs - Agreement in Faulty Systems
Byzantine Faults
Byzantine Generals Problem:
o four generals lead their divisions of an army
• the divisions camp on the mountains on the four sides of an enemy-occupied valley
the divisions can only communicate via messengers
o messengers are totally reliable, but may need an arbitrary amount of time to cross the valley
they may even be captured and never arrive o if the actions taken by each division is not consistent with that of the others, the army will be defeated
we need a scheme for the generals to agree on a common plan of action (attack or retreat)
o even if some of the generals are traitors who will do anything to prevent loyal generals from reaching the agreement
o the problem is nontrivial even if messengers are totally reliable with unreliable messengers, the problem is very complex o Fischer, Lynch, Paterson: in asynchronous systems, it is impossible to reach a consensus in a finite amount of time
Ludek Matyska  (FIMU) 3. Distributed Systems Spring 2011        87 / 100
Issues Examples      Fault Tolerance in Distributed Systems
Fault Tolerance in DSs - Agreement in Faulty Systems
Formal Definition I.
Formal definition of the agreement problem in DSs:
• let's have a set of distributed processes with initial states e 0,1 o the goal: all the processes have to agree on the same value
o additional requirement: it must be possible to agree on both 0 or 1 states
basic assumptions:
o system is asynchronous
• no bounds on processes' execution delays exist
• no bounds on messages' delivery delay exist o there are no synchronized clocks
o no communication failures - every process can communicate with its neighbors
o processes fail by crashing - we do not consider byzantine failures
3. Distributed Systems
Spring 2011       88 / 100
Issues Examples      Fault Tolerance in Distributed Systems
Fault Tolerance in DSs - Agreement in Faulty Systems
Formal Definition II.
Formal definition of the agreement problem in DSs: cont'd. o implications:
=> there is no deterministic algorithm which resolves the consensus problem in an asynchronous system with processes, which may fail o because it is impossible to distinguish the cases: o a process does not react, because it has failed o a process does not react, because it is slow o practically overcomed by establishing timeouts and by ignoring/killing too slow processes
o timeouts used in so-called Failure Detectors (see later)
3. Distributed Systems
Spring 2011       89 / 100
Issues Examples      Fault Tolerance in Distributed Systems
Fault Tolerance in DSs - Agreement in Faulty Systems
Fault-tolerant Broadcast
Fault-tolerant Broadcast:
o if there was a proper type of fault-tolerant broadcast, the agreement
problem would be solvable o various types of broadcasts:
o reliable broadcast o FIFO broadcast casual broadcast
o atomic broadcast - the broadcast, which would solve the agreement problem in asynchronous systems
3. Distributed Systems
Spring 2011        90 / 100
Issues Examples      Fault Tolerance in Distributed Systems
Fault Tolerance in DSs - Agreement in Faulty Systems
Fault-tolerant Broadcast - Reliable Broadcast
Reliable Broadcast:
o basic features:
o Validity - if a correct process broadcasts m, then it eventually delivers m
□ Agreement - if a correct process delivers m, then all correct processes
eventually deliver m o (Uniform) Integrity - m is delivered by a process at most once, and
only if it was previously broadcasted
o possible to implement using send/receive primitives:
o the process p sending the broadcast message marks the message by its identifier and sequence number o and sends it to all its neighbors o once a message is received:
• if the message has not been previously received (based in sender's ID and sequence number), the message is delivered
• if the particular process is not message's sender, it delivers it to all its neighbors
Ludek Matyska  (FIMU) 3. Distributed Systems Spring 2011 91/100
Issues Examples      Fault Tolerance in Distributed Systems
Fault Tolerance in DSs - Agreement in Faulty Systems
Fault-tolerant Broadcast - FIFO Broadcast
FIFO Broadcast:
• the reliable broadcast cannot assure the messages' ordering
• it is possible to receive a subsequent message (from the sender's view) before the previous one is received
o FIFO broadcast: the messages from a single sender have to be delivered in the same order as they were sent
• FIFO broadcast = Reliable broadcast + FIFO ordering
o if a process p broadcasts a message m before it broadcasts a message m', then no correct process delivers m' unless it has previously delivered m
o broadcastp(m) — broadcastp(m') == deliverq(m) — deliverq(m') o a simple extension of the reliable broadcast
3. Distributed Systems
Spring 2011     92 / 100
Issues Examples      Fault Tolerance in Distributed Systems
Fault Tolerance in DSs - Agreement in Faulty Systems
Fault-tolerant Broadcast - Casual Broadcast
Casual Broadcast:
o the FIFO broadcast is still not sufficient: it is possible to receive a message from a third party, which is a reaction to a particular message before receiving that particular message
o == Casual broadcast • Casual broadcast = Reliable broadcast + casual ordering
o if the broadcast of a message m happens before the broadcast of a
message m', then no correct process delivers m' unless it has previously
delivered m
o broadcastp(m) — broadcastq(m') == deliver,.(m) — deliver,(m') o can be implemented as an extension of the FIFO broadcast
3. Distributed Systems
Spring 2011        93 / 100
Issues Examples      Fault Tolerance in Distributed Systems
Fault Tolerance in DSs - Agreement in Faulty Systems
Fault-tolerant Broadcast - Atomic Broadcast
Atomic Broadcast:
o even the casual broadcast is still not sufficient: sometimes, it is necessary to guarantee the proper in-order delivery of all the replicas o two bank offices: one of them receives the information about adding an interest before adding a particular amount of money to the account, the second one receives these messages contrariwise o == inconsistency o == Atomic broadcast • Atomic broadcast = Reliable broadcast + total ordering
o if correct processes p and q both deliver messages m, m', then p
delivers m before m' if and only if q delivers m before m' o deliverp(m) — deliverp(m') == deliverq(m) — deliverq(m')
o does not exist in asynchronous systems
3. Distributed Systems
Spring 2011        94 / 100
Issues Examples      Fault Tolerance in Distributed Systems
Fault Tolerance in DSs - Agreement in Faulty Systems
Fault-tolerant Broadcast - Timed Reliable Broadcast
Timed Reliable Broadcast:
o a way to practical solution
o introduces an upper limit (time), before which every message has to be delivered
• Timed Reliable broadcast = Reliable broadcast + timeliness
□ there is a known constant A such that if a message is broadcasted at real-time t, then no correct (any) process delivers m after real-time t + A
feasible in synchronous systems
3. Distributed Systems
Spring 2011        95 / 100
Issues Examples      Fault Tolerance in Distributed Systems
Fault Tolerance in DSs - Agreement in Faulty Systems
Failure Detectors I.
impossibility of consensus caused by inability to detect slow process and a failed process
• synchronous systems: let's use timeouts to determine whether a process has crashed
o == Failure Detectors
Failure Detectors (FDs):
a distributed oracle that provides hints about the operational status of processes
o (which processes had failed) o every process maintains its own FD
o and asks just it to determine, whether a process had failed o however:
hints may be incorrect o FD may give different hints to different processes o FD may change its mind (over & over) about the operational status of a process
Ludek Matyska  (FIMU) 3. Distributed Systems Spring 2011        96 / 100
Issues Examples     Fault Tolerance in Distributed Systems
3. Distributed Systems
Spring 2011        97 / 100
Issues Examples      Fault Tolerance in Distributed Systems
Fault Tolerance in DSs - Agreement in Faulty Systems
(Perfect) Failure Detector
Perfect Failure Detector:
o properties:
o Eventual Strong Completeness - eventually every process that has crashed is permanently suspected by all non-crashed processes Eventual Strong Accuracy - no correct process is ever suspected
o hard to implement
o is perfect failure detection necessary for consensus? No.
o == weaker Failure Detector
weaker Failure Detector:
properties:
Strong Completeness - there is a time after which every faulty process is suspected by every correct process
Eventual Strong Accuracy - there is a time after which no correct process is suspected can be used to solve the consensus
o this is the weakest FD that can be used to solve the consensus
Ludek Matyska  (FIMU) 3. Distributed Systems Spring 2011        98 / 100
Conclusion
Lecture overview
1 Distributed Systems ^ Key characteristics
• Challenges and Issues
• Distributed System Architectures
• Inter-process Communication
1 Middleware
• Remote Procedure Calls (RPC)
• Remote Method Invocation (RMI)
9 Common Object Request Broker Architecture (CORBA)
1
2
Web Services
Grid Services
Issues Examples
• Scheduling/Load-balancing in Distributed Systems Fault Tolerance in Distributed Systems
Conclusion
3. Distributed Systems
Spring 2011        99 / 100
Conclusion
Distributed Systems - Further Information
o FI courses:
o PA150: Advanced Operating Sytems Concepts (doc. Staudek) o PA053: Distributed Systems and Middleware (doc. Tuma) o IA039: Supercomputer Architecture and Intensive Computations (prof. Matyska)
o PA177: High Performance Computing (LSU, prof. Sterling) o IV100: Parallel and distributed computations (doc. Kraiovic) o IB109: Design and Implementation of Parallel Systems (dr. Barnat) etc.
o (Used) Literature:
o W. Jia and W. Zhou. Distributed Network Systems: From concepts to
implementations. Springer, 2005. o A. S. Tanenbaum and M. V. Steen. Distributed Systems: Principles and
paradigms. Pearson Prencite Hall, 2007. o G. Coulouris, J. Dollimore, and T. Kindberg. Distributed Systems: Concepts
and design. Addison-Wesley publishers, 2001. o Z. Tari and O. Bukhres. Fundamentals of Distributed Object Systems: The
CORBA perspective. John Wiley & Sons, 2001.
etc.
Ludek Matyska  (FIMU) 3. Distributed Systems Spring 2011        100 / 100