Chapter 17: Distributed Systems (DS)

Transcription

1 Chapter 17: Distributed Systems (DS) Silberschatz, Galvin and Gagne 2013

2 Chapter 17: Distributed Systems Advantages of Distributed Systems Types of Network-Based Operating Systems Network Structure Communication Structure Communication Protocols An Example: TCP/IP Robustness Design Issues Distributed File System 17.2 Silberschatz, Galvin and Gagne 2013

3 Chapter Objectives To provide a high-level overview of distributed systems and the networks that interconnect them To discuss the general structure of distributed operating systems To explain general communication structure and communication protocols To describe issues concerning the design of distributed systems 17.3 Silberschatz, Galvin and Gagne 2013

4 Overview Distributed system (DS) is collection of loosely coupled processors interconnected by a communications network These processors only share the network & don t share memory or a clock; each processor has its own memory and run an independent OS Processors variously called nodes, computers, machines, hosts, sites Site indicates the location of a machine Node refers to a specific system at a site In general, there is one node at one site a server has a set of resources which a client node at a different site wants to use 17.4 Silberschatz, Galvin and Gagne 2013

5 Why building Distributed Systems? The reasons for building DS 1- Resource sharing Sharing and printing files at remote sites Processing information in a distributed database Using remote specialized hardware devices (e.g., supercomputers) 2- Computation speedup A distributed system allows us to distribute the sub-computations among the various sites and run them concurrently load sharing or job migration Moving some jobs from a heavy loaded site to other, lightly loaded sites 17.5 Silberschatz, Galvin and Gagne 2013

6 3- Reliability Why building Distributed Systems? (cont) If one site fails in a distributed system, the remaining sites can continue operating to provide better reliability Satisfied with enough redundancy of hardware and data detect and recover from site failure, function transfer, reintegrate failed site 4- Communication Low level communication: message passing High level communication: e.g., , chat, login, RPC, file transfer 5- Computers can be downsized moving from large system to multiple smaller systems performing distributed computing Adv.: more flexibility, better user interfaces and easier maintenance 17.6 Silberschatz, Galvin and Gagne 2013

7 17.7 Silberschatz, Galvin and Gagne 2013

8 Types of OS for Distributed Systems Network Operating Systems (NOS) Distributed Operating Systems (DOS) 17.9 Silberschatz, Galvin and Gagne 2013

9 Network-Operating Systems (NOS) Simpler to implement but generally more difficult for users to access and utilize than DOS Users must be aware of multiplicity of machines e.g., users need to know the IP address (or server name) of each remote machine to connect to it Access to resources of various machines is done explicitly by: (2 major functions of NOS) Remote Logging into the appropriate remote machine (telnet, ssh) Remote File Transfer Disadv.: transferring data from remote machines to local machines (FTP, SFTP) Users must change paradigms establish a session, give network-based commands More difficult for users Didn t support real file sharing Silberschatz, Galvin and Gagne 2013

10 تكريم أوائل امتحان الميدترم Silberschatz, Galvin and Gagne 2013

11 Distributed-Operating Systems Users not aware of multiplicity of machines Access to remote resources similar to access to local resources Data Migration Transfer data by transferring entire file Called Automated FTP is used with AFS (stands for?) or Transferring only those portions of the file necessary for the immediate task (similar to?) e.g., NFS, CIFS Computation Migration transfer the computation, rather than the data, across the system. Via Remote Procedure Calls (RPCs): uses network protocols to execute a routine on a remote system messaging system RPC and message passing can be used in same computation Silberschatz, Galvin and Gagne 2013

12 Distributed-Operating Systems (Cont.) Process Migration execute an entire process, or parts of it, at different sites Is a logical extension of computation migration Reasons for process migration: Load balancing distribute processes across network to even the workload Computation speedup sub-processes can run concurrently on different sites Hardware preference process execution may require specialized processor Software preference required software may be available at only a particular site Data access run process remotely, rather than transfer all data locally Consider the World Wide Web (WWW) as the biggest example for a distributed operating system provides data migration (between a web server and a web client). provides computation migration (e.g., a web client could trigger a database operation on a web server) Provides process migration: Java applets and Javascript scripts are sent from the server to the client Note: NOS provides most of these features, but DOS makes them seamless and easily accessible Silberschatz, Galvin and Gagne 2013

13 Network Structure Two types of networks: Local-area networks (LAN) Wide-area networks (WAN) Main difference: The network types differ in how they are geographically distributed? Silberschatz, Galvin and Gagne 2013

14 Local Area Network (LAN) Local-Area Network (LAN) designed to cover small geographical area Adv.: higher speed and lower error rate than WAN Multiple topologies like star (most commonly used) or ring Speeds from 1Mbps(Appletalk, bluetooth) to 40 Gbps for fastest Ethernet Links are made of Consists of twisted pair copper or optical fiber (most commonly used: because it provides higher communication over longer distance than the copper) multiple computers (mainframes through mobile devices) Peripheral devices(printers, storage arrays), routers (specialized network communication processors) providing access to other networks LAN can be constructed using Ethernet most common way to construct LANs Ethernet provides a set of standards (e.g., cables, the structure of data sent among cables, H/W connect cables) Has no central controller ( because it is a Multi-access bus) so, it provides (star bus topology) Defined by IEEE standard Wireless spectrum (WiFi) Adv.: no cabling to connect communicating hosts Disadv.: slower than Ethernet. (Data transfer rate depends on what?) follow IEEE standards» IEEE g standard implemented at 54 Mbps (in practical, less than half)» IEEE Operating System Concepts 9 th n standard provide higher data rate (in practical, 75 Mbps) Edition Silberschatz, Galvin and Gagne 2013

15 Local-area Network Star Network Multi-access bus Silberschatz, Galvin and Gagne 2013

16 Network Types (Cont.) Wide-Area Network (WAN) links distributed over a large geographical area The first WAN was Arpanet (composed of four-site network) grows to internet (network of networks) Communication links: Disadav.: slow and unreliable Are based on telephone lines Are controlled by special network connection processors, called routers Communicating sites over network Transferring data (i.e., routing msgs) among various sites e.g., Internet WAN enables different hosts (i.e., not homogeneous) to communicate Hosts differ in all dimensions (e.g., OS, speed, CPU type) but WAN allows communications Speeds T1 (telephone-system service) link is Mbps T3 is 28 x T1s = 45 Mbps OC-12 is 622 Mbps WANs and LANs interconnect However, it is difficult to tell where one ends and the other starts Silberschatz, Galvin and Gagne 2013

17 Communication Processors in a Wide-Area Network Silberschatz, Galvin and Gagne 2013

18 Communication Structure The design of a communication network must address four basic issues: Naming and name resolution - How do two processes locate each other to communicate? Routing strategies - How are messages sent through the network? Packet strategies - Are packets sent individually or as a sequence? Connection strategies - How do two processes send a sequence of messages? Silberschatz, Galvin and Gagne 2013

19 Naming and Name Resolution Naming of the systems in the network Address messages with the process-id Identify processes on remote systems by <host-name, identifier> pair Host-name: unique name within the network Identifier: unique number for a process within a host Domain name system (DNS) specifies the naming structure of the hosts, as well as name to address resolution (Internet) Map host-name to a host-id Silberschatz, Galvin and Gagne 2013

20 Routing Strategies When a process at site A wants to communicate with a process at site B, how is the message sent? If there is only one physical path between A and B route the message on it If there multiple paths use one of the routing schemes fixed routing, virtual routing, and dynamic routing Fixed routing - A path from A to B is specified in advance; path changes only if a hardware failure disables it Usually the shortest path is chosen so, communication costs are minimized Adv.: Ensures that messages will be delivered in the order in which they were sent Disadv.: cannot adapt to load changes or link failure Virtual routing- A path from A to B is fixed for the duration of one session. Different sessions involving messages from A to B may have different paths e.g., A short session may as a file transfer. A long session may be as a remote-login period Adv.: Ensures that messages will be delivered in the order in which they were sent Partial remedy for adapting to load changes Silberschatz, Galvin and Gagne 2013

21 Routing Strategies (Cont.) Dynamic routing - The path used to send a message from site A to site B is chosen only when a message is sent Therefore, different message between A and B has different paths Usually a site sends a message to another site on the path least used at that particular time Adv.: Adapts to load changes by avoiding routing messages on heavily used path Disadv.: Messages may arrive out of order Solution: appending a sequence number to each message Most complex to set up and run UNIX provides both fixed routing (used with simple network) dynamic routing (used with complex network) Hosts may have fixed routes and gateways (e.g., router) connecting networks together may have dynamic routes Silberschatz, Galvin and Gagne 2013

22 Routing Strategies (Cont.) Router is a network communications processor responsible for routing messages Maybe Special-purpose device Host-computer with routing software Checks its routing tables to determine only the destination host (i.e., where to send messages) Static routing table only changed manually Dynamic routing table changed via routing protocol Silberschatz, Galvin and Gagne 2013

23 Packet Strategies Messages vary in length simplified design breaks them into fixed-length messages, called packets (or frames, or datagrams) The messages may be: Unreliable Called, Connectionless message is just one packet Disavd.: the sender has no guarantee that the packet reached its destination Reliable Can send multiple packets between source and destination an acknowledgement packet is returned from the destination indicating that the original packet arrived Silberschatz, Galvin and Gagne 2013

24 Connection Strategies 1. Circuit switching - A permanent physical link is established for the duration of the communication (e.g., telephone system) Tradeoff: requires setup time, but incurs less overhead for shipping each message, and may waste network bandwidth 2. Message switching - A temporary link is established for the duration of one message transfer (e.g., post-office mailing system) 3. Packet switching - Messages of variable length are divided into fixed-length packets which are sent to the destination Each packet may take a different path through the network The packets must be reassembled into messages as they arrive Tradeoff: Both of message and packet switching require less setup time, but incur more overhead per message Note: It is not harmful for data to be broken into packets, possibly routed separately, and reassembled at the destination Silberschatz, Galvin and Gagne 2013

25 Robustness A distributed system may suffer from various types of hardware failure e.g., link failure, site failure, message loss (generally, it is not easy to differentiate among these failures) Therefore, we must ensure that the DS is robust The robustness of a DS can be satisfied through: 1. Failure detection 2. Reconfiguration 3. Recovery Silberschatz, Galvin and Gagne 2013

26 Robustness: Failure Detection Detecting hardware failure is difficult To detect a link and site failure, a heartbeat protocol (called, handshaking procedure) can be used Assume Site A and Site B have established a link At fixed intervals, each site will exchange an I-am-up message indicating that they are up and running If Site A does not receive a message within the fixed interval, it assumes either Site B is not up or The message was lost Site A can now send an Are-you-up? message to Site B If Site A does not receive a reply, it can repeat the message or try an alternate route to Site B If Site A does not ultimately receive a reply from Site B, it concludes some type of failure has occurred Types of failures: - Site B is down - The direct link between A and B is down - The alternate link from A to B is down - The message has been lost However, Site A cannot determine exactly which of these failures has occurred Silberschatz, Galvin and Gagne 2013

27 Robustness: Reconfiguration When Site A determines a failure has occurred, it must reconfigure the system: 1. If the link from A to B has failed, this must be broadcast to every site in the system 2. If a site has failed, every other site must also be notified indicating that the services offered by the failed site are no longer available Robustness: Recovery When the link or the site becomes available again, this information must again be broadcast to all other sites Silberschatz, Galvin and Gagne 2013

28 Design Issues 1. Transparency the distributed system should appear as a conventional and centralized system to the user Transparency in DS includes two aspects: the users can use the remote resources same as the local ones Allow the users to log into any machine in the system rather than forcing them to use a specific machine (i.e., user mobility) 2. Fault tolerance the distributed system should continue to function in the face of failure 3. Scalability the capability of a system to adapt to increased service load as demands increase, the system should easily accept the addition of new resources to accommodate the increased demand e.g., Hadoop open source programming framework for processing large datasets (i.e., big data) in distributed environments (based on Google s MapReduce) Silberschatz, Galvin and Gagne 2013

29 Distributed File System Distributed file system (DFS) a distributed implementation of the classical time-sharing model of a file system, where multiple users share files and storage resources DFS manages set of dispersed storage devices (composed of different, remotely located, smaller storage spaces) Examples of DFS: NFS (Unix-based DFS), OpenAFS (open-source distributed file system), Hadoop DFS Challenges include: Naming and Transparency Remote File Access Silberschatz, Galvin and Gagne 2013

30 DFS Structure 1. Service software entity running on one or more machines and providing a particular type of function to a priori unknown clients 2. Server service software running on a single machine 3. Client process that can invoke a service using a set of operations that forms its client interface A client interface for a file service is formed by a set of primitive file operations (create, delete, read, write) Client interface of a DFS should be transparent, i.e., not distinguish between local and remote files Therefore, a DFS is a file system whose clients, servers, and storage devices are dispersed among the machines of a distributed system Silberschatz, Galvin and Gagne 2013

31 Challenge: Naming and Transparency Naming mapping between logical and physical objects Multilevel mapping abstraction of a file that hides the details of how and where on the disk the file is actually stored A transparent DFS hides the location where in the network the file is stored For a file being replicated in several sites, the mapping returns a set of the locations of this file s replicas Such that, both the existence of multiple copies and their location are hidden For name mapping, differentiate between Location transparency file name does not reveal the file s physical storage location Location independence file name is independent of the file s physical storage location file name does not need to be changed when the file s physical storage location changes e.g., OpenAFS supports location independency e.g., NFS supports both of location transparency and location independence Silberschatz, Galvin and Gagne 2013

32 Challenge: Remote File Access Remote-service mechanism is a transfer approach to transfer files from server to clients Can be implemented using RPC Improve performance by using cache In traditional FS, we cache to reduce disk I/O, while in DFS, we cache to reduce both of network traffic and disk I/O Reduce network traffic by retaining recently accessed disk blocks in a cache, so that repeated accesses to the same information can be handled locally If needed data not already cached, a copy of data is brought from the server to the user Accesses are performed on the cached copy Files identified with one master copy residing at the server machine, but copies of (parts of) the file are scattered in different caches Cache-consistency problem the problem of keeping the cached copies consistent with the master file In DFS, Cache is called network virtual memory Silberschatz, Galvin and Gagne 2013

33 Cache Location on Disk vs. in Main Memory Advantages of disk caches More reliable Cached data kept on disk are still there during recovery and don t need to be fetched again Advantages of main-memory caches Permit workstations to be diskless Data can be accessed more quickly Performance speedup in bigger memories Server caches (used to speed up disk I/O) are in main memory regardless of where user caches are located; using main-memory caches on the user machine permits a single caching mechanism for servers and users e.g., main-memory caches used in NFS Silberschatz, Galvin and Gagne 2013

34 Cache Update Policy 1. Write-through policy write data through to disk (i.e., server s disk) as soon as they are placed on any cache Reliable, but poor performance 2. Delayed-write policy(called, write-back caching) modifications are written to the cache and then are written through to the server at a later time Adv.: DisAdv.: Write accesses complete quickly; Some data may be overwritten before they are written back, and so only that last updates will be written back Poor reliability; unwritten data will be lost whenever a user machine crashes Variations Write on ejection from cache Scan cache at regular intervals and flush blocks that have been modified since the last scan (e.g., NFS uses this approach) write-on-close: writes data back to the server when the file is closed Best for files that are open for long periods and frequently modified e.g., OpenAFS applies write-on-close policy Silberschatz, Galvin and Gagne 2013

35 Cache Consistency Is locally cached copy of the data consistent with the master copy? 1. Client-initiated approach Client initiates a validity check to server Server checks whether the local data (i.e., cached copy) are consistent with the master copy When the client sends the validity check? (before every access, on file open, at fixed time interval) 2. Server-initiated approach Server records, for each client, the (parts of) files it caches When server detects a potential inconsistency, it must react A potential inconsistency occurs when two different clients in conflicting modes cache the same file Silberschatz, Galvin and Gagne 2013

36 End of Chapter 17 Silberschatz, Galvin and Gagne 2013