Chapter 1: Distributed Systems: What is a distributed system? Fall 2013

Transcription

1 Chapter 1: Distributed Systems: What is a distributed system? Fall 2013

2 Course Goals and Content n Distributed systems and their: n Basic concepts n Main issues, problems, and solutions n Structured and functionality n Content: n Distributed systems - Architectures, goal, challenges - Where our solutions are applicable n Synchronization: Time, coordination, decision making n Replicas and consistency n Fault tolerance n Large-scale distributed systems in real world 10 September

3 Course Material n Tanenbaum, van Steen: Distributed Systems, Principles and Paradigms; Pearson Prentice Hall 2007 n 2002 edition also ok n Coulouris, Dollimore, Kindberg: Distributed Systems, Concepts and Design; Addison-Wesley 2005 n Selected content from Barroso L. A. and Hölzle U.: The Datacenter as a Computer (online, see course website) n Lecture slides on course website n NOT sufficient by themselves n Help to see what parts in book are most relevant n On some topics, slides cover more material than the book 10 September

4 Course Exams n Normal way (recommended) n Exercises, home exercises, course exam n Grading: n Exam 42 points, date December 11, 2013 n Exercises 12 points (6 exercises, scaled to 0 12) n Home exercises 12 points (4 exercises) n Grading based on 60 point maximum n Need 30 points to pass with minimum 16 points in exam n 50 points will give a 5 n Possible to take as separate exam 10 September

5 Exercises n Bi-weekly exercises: n Smaller assignments n Exception to schedule: has exercise, does not n Complete beforehand, present your work at exercise sessions to gain points n Must make an effort to get points! n Home exercises n 4 larger exercises - 2 writing exercises - 2 programming exercises n Due dates will be announced later, deadlines fixed! 10 September

6 People n Jussi Kangasharju n Lectures: - Period I: Mon and Thu in C222 - Period II: Mon and Thu in D lectures, usually on Mondays - Thursday sessions workshops/discussion/help/ n Exercise group: Fri in B222 (bi-weekly) n Office hour: Mon or ask for appointment by n Ossi Karkulahti n Exercise group: Fri in B222 (bi-weekly) n Home exercises n Office hour: During exercises or ask appointment by 10 September

7 Lectures vs. Workshops n Lectures on Mondays present new material n There can and should be discussion as well J n Workshops on Thursdays are for discussion and help n First three workshops mainly tutorials for new students n Later workshops discussion about exercises n Every week has some scheduled program n Program posted on website in advance 10/09/2013 7

8 Questions? 10 September

9 Chapter Outline n Defining distributed system n Examples of distributed systems n Why distribution? n Goals and challenges of distributed systems n Where is the borderline between a computer and a distributed system? n Examples of distributed architectures 10 September

10 Definition of a Distributed System A distributed system is a collection of independent computers that appears to its users as a single coherent system.... or... as a single system. 10 September

11 Where Does the Definition Leave Us? n Which of the following are distributed systems? Multi-core processor Parallel systems Multi-processor computer One data center Web Internet Computing cluster Corporate intranet Local Area Network Network of data centers 10 September

12 What About the Following? Airplane Car Ticket reservation system Mobile phone My home theater setup Factory Nuclear power plant My office 10/09/

13 Examples of Distributed Systems Distributed application one single system one or several autonomous subsystems a collection of processors => parallel processing => increased performance, reliability, fault tolerance partitioned or replicated data => increased performance, reliability, fault tolerance 10 September

14 Why Distribution? Sharing of information and services Possibility to add components improves availability reliability, fault tolerance performance Leads to scalability And a large set of gotchas 10 September

15 Goals and challenges for distributed systems

16 Goals n Making resources accessible n Distribution transparency n Openness n Scalability n Security n System design requirements 10 September

17 Challenges for Making Resources Accessible n Naming n Access control n Security n Availability n Performance n Mutual exclusion of users, fairness n Consistency in some cases 10 September

18 Challenges for Transparency n The fundamental idea: a collection of independent, autonomous actors n Transparency: Concealment of distribution => user s point of view: a single unified system 10 September

19 Transparencies Transparency Description Access Hide differences in data representation and how a resource is accessed Location Hide where a resource is located (*) Migration Hide that a resource may move to another location (*) (the resource does not notice) Relocation Hide that a resource may be moved to another location (*) while in use (the others don t notice) Replication Hide that a resource is replicated Concurrency Hide that a resource may be shared by several competitive users Failure Hide the failure and recovery of a resource Persistence Hide whether a (software) resource is in memory or on disk (*) Note the various meanings of location : network address (several layers) ; geographical address 10 September

20 Challenges for Transparencies n replications and migration cause need for ensuring consistency and distributed decision-making n failure modes n concurrency n heterogeneity 10 September

21 Omission and arbitrary failures Class of failure Affects Description Fail-stop Process Process halts and remains halted. Other processes may detect this state. Crash Process Process halts and remains halted. Other processes may not be able to detect this state. Omission Channel A message inserted in an outgoing message buffer never arrives at the other end s incoming message buffer. Send-omission Process A process completes send, but the message is not put in its outgoing message buffer. Receive- Process A message is put in a process s incoming message omission buffer, but that process does not receive it. Arbitrary Process or Process/channel exhibits arbitrary behaviour: it may (Byzantine) channel send/transmit arbitrary messages at arbitrary times, commit omissions; a process may stop or take an incorrect step. 10 September

22 Timing failures Class of Failure Affects Description Clock Process Process s local clock exceeds the bounds on its rate of drift from real time. Performance Process Process exceeds the bounds on the interval between two steps. Performance Channel A message s transmission takes longer than the stated bound. 10 September

23 Failure Handling n More components => increased fault rate n Increased possibilities n more redundancy => more possibilities for fault tolerance n no centralized control => no fatal failure n Issues n Detecting failures n Masking failures n Recovery from failures n Tolerating failures n Redundancy n New: partial failures n New: warehouse-scale computing (we return to this) 10 September

24 Concurrency n Concurrency: n Several simultaneous users => integrity of data - mutual exclusion - synchronization - ext: transaction processing in data bases n Replicated data: consistency of information? n Partitioned data: how to determine the state of the system? n Order of messages? n There is no global clock! 10 September

25 Consistency Maintenance n Update... n Replication... n Cache consistency n Failure... n Clock... n User interface September

26 Heterogeneity n Heterogeneity of n networks n computer hardware n operating systems n programming languages n implementations of different developers n Portability, interoperability n Mobile code, adaptability (applets, agents) n Middleware (CORBA etc) n Degree of transparency? Latency? Location-based services? 10 September

27 Challenges for Openness n Openness facilitates n interoperability, portability, extensibility, adaptivity n Activities addresses n extensions: new components n re-implementations (by independent providers) n Supported by n public interfaces n standardized communication protocols 10 September

28 Challenges for Scalability Scalability: n The system will remain effective when there is a significant increase in: n number of resources n number of users n The architecture and the implementation must allow it n The algorithms must be efficient under the circumstances to be expected n Example: the Internet 10 September

29 Challenges: Scalability (cont.) n Controlling the cost of physical resources n Controlling performance loss n Preventing software resources running out n Avoiding performance bottlenecks n Mechanisms (implement functions) & Policies (how to use the mechanisms) n Scaling solutions n asynchronous communication, decreased messaging n caching (all sorts of hierarchical memories: data is closer to the user à no wait / assumes rather stable data!) n distribution i.e. partitioning of tasks or information (domains) (e.g., DNS) 10 September

30 Challenges for Security n Mostly similar to normal challenges in wide-area networks n Sometimes easier (closed, dedicated systems) n Solution techniques n cryptography n authentication n access control techniques n Policies n access control models n information flow models n Leads to: secure channels, secure processes, controlled access, controlled flows 10 September

31 Environment challenges n A distributed system: n HW / SW components in different nodes n components communicate (using messages) n components coordinate actions (using messages) n Distances between nodes vary n in time: from msecs to weeks n in space: from mm s to Mm s n in dependability n Autonomous independent actors (=> even independent failures!) No global clock Global state information not possible 10 September

32 Challenges: Design Requirements n Performance issues n responsiveness n throughput n load sharing, load balancing n issue: algorithm vs. behavior n Quality of service n correctness (in nondeterministic environments) n reliability, availability, fault tolerance n security n performance n adaptability 10 September

33 Where is the borderline between a computer and distributed system?

34 Hardware Concepts n Characteristics which affect the behavior of software systems n The platform... n the individual nodes ( computer, processor ) n communication between two nodes n organization of the system (network of nodes) n... and its characteristics n capacity of nodes n capacity (throughput, delay) of communication links n reliability of communication (and of the nodes) à Which ways to distribute an application are feasible 10 September

35 Basic Organizations of a Node Different basic organizations and memories in distributed computer systems 10 September

36 Multiprocessors (1) A bus-based multiprocessor. Essential characteristics for software design fast and reliable communication (shared memory) => cooperation at instruction level possible bottleneck: memory (especially the hot spots ) 10 September

37 General Multicomputer Systems n Hardware: Possibly very heterogeneous n Loosely connected systems n nodes: autonomous n communication: slow and vulnerable n => cooperation at service level n Application architectures n multiprocessor systems: parallel computation n multicomputer systems: distributed systems n ( how are parallel, concurrent, and distributed systems different?) 10 September

38 Software Concepts System Description Main Goal DOS Tightly-coupled operating system for multiprocessors and homogeneous multicomputers Hide and manage hardware resources NOS Loosely-coupled operating system for heterogeneous multicomputers (LAN and WAN) Offer local services to remote clients Middleware Additional layer atop of NOS implementing general-purpose services Provide distribution transparency DOS: Distributed OS; NOS: Network OS 10 September

39 History of distributed systems n RPC by Birel &Nelson -84 n network operating systems, distributed operating systems, distributed computing environments in mid-1990; middleware referred to relational databases n Distributed operating systems single computer n Distributed process management - process lifecycle, inter-process communication, RPC, messaging n Distributed resource management - resource reservation and locking, deadlock detection n Distributed services - distributed file systems, distributed memory, hierarchical global naming 10 September

40 History of distributed systems n late 1990 s distribution middleware well-known n generic, with distributed services n supports standard transport protocols and provides standard API n available for multiple hardware, protocol stacks, operating systems n e.g., DCE, COM, CORBA n present middlewares for n multimedia, realtime computing, telecom n ecommerce, adaptive / ubiquitous systems 10 September

41 Misconceptions tackled n The network is reliable n The network is secure n The network is homogeneous n The topology does not change n Latency is zero n Bandwith is infinite n Transport cost is zero n There is one administrator n There is inherent, shared knowledge 10 September

42 Multicomputer Operating Systems (1) General structure of a multicomputer operating system 10 September

43 Multicomputer Operating Systems (2) Alternatives for blocking and buffering in message passing. 10 September

44 Distributed Shared Memory Systems (1) a) Pages of address space distributed among four machines b) Situation after CPU 1 references page 10 c) Situation if page 10 is read only and replication is used 10 September

45 Distributed Shared Memory Systems (2) False sharing of a page between two independent processes. 10 September

46 Network Operating System (1) General structure of a network operating system. 10 September

47 Network Operating System (2) Two clients and a server in a network operating system. 10 September

48 Network Operating System (3) Different clients may mount the servers in different places. 10 September

49 Software Layers n Platform: computer & operating system &.. n Middleware: n mask heterogeneity of lower levels n (at least: provide a homogeneous platform ) n mask separation of platform components - implement communication - implement sharing of resources n Applications: , www-browsers, 10 September

50 Positioning Middleware General structure of a distributed system as middleware. 10 September

51 Middleware n Operations offered by middleware n RMI, group communication, notification, replication, (Sun RPC, CORBA, Java RMI, Microsoft DCOM,...) n Services offered by middleware n naming, security, transactions, persistent storage, n Limitations n ignorance of special application-level requirements End-to-end argument: n Communication of application-level peers at both ends is required for reliability 10 September

52 Middleware and Openness In an open middleware-based distributed system, the protocols used by each middleware layer should be the same, as well as the interfaces they offer to applications. 10 September

53 Comparison between Systems Item Distributed OS Multiproc. Multicomp. Network OS Middleware-based OS Degree of transparency Very High High Low High Same OS on all nodes Yes Yes No No Number of copies of OS 1 N N N Basis for communication Shared memory Messages Files Model specific Resource management Global, central Global, distributed Per node Per node Scalability No Moderately Yes Varies Openness Closed Closed Open Open 10 September

54 More examples on distributed software architectures

55 Architectural Models n Architectural models provide a high-level view of the distribution of functionality between system components and the interaction relationships between them n Architectural models define n components (logical components deployed at physical nodes) n communication n Criteria n performance n reliability n scalability,.. 10 September

56 Client-Server Architectures n General interaction between a client and a server.

57 Processing Level The general organization of an Internet search engine into three different layers 10 September

58 Multitiered Architectures (1) Alternative client-server organizations. 10 September

59 Multitiered Architectures (2) Client - server: generalizations request node 1 A node 2 reply B node 3 A client: node 1 server: node 2 B client: node 2 server: node 3 node 4 the concept is related to communication not to nodes 10 September

60 Multitiered Architectures (3) An example of a server acting as a client. 10 September

61 Modern Architectures An example of horizontal distribution of a Web service. 10 September

62 Chapter Summary n Introduction into distributed systems n Challenges and goals of distributing n Examples of distributed systems 10 September