1. Distributed Systems

2. Introduction to Distributed Systems Why do we develop distributed systems? –availability of powerful yet cheap microprocessors (PCs, workstations), continuing advances in communication technology. What is a distributed system? A distributed system is a collection of independent computers connected via a high speed computers and appear to the users of the system as a single system.

3. Examples of DS –Network of workstations :Personal workstations + a pool of processors + single file system –Distributed manufacturing system (e.g. automated assembly line) :Robots on the assembly line + Robots in the parts department –Network of branch office computers : A large bank with hundreds of branch offices all over the world

4. Advantages of Distributed Systems over Centralized Systems Economics: a collection of microprocessors offer a better price/performance ratio than mainframes. Low price/performance ratio: cost effective way to increase computing power. microprocessors offer a better price/performance than mainframes Speed: a distributed system may have more total computing power than a mainframe. Ex. 10,000 CPU chips, each running at 50 MIPS. Not possible to build 500,000 MIPS single processor since it would require 0.002 nsec instruction cycle. Enhanced performance through load distributing. a distributed system may have more total computing power than a mainframe

5. Advantages of Distributed Systems over Centralized Systems Inherent distribution: Some applications are inherently distributed. Ex. a supermarket chain, co- operative work ( a group of people located far away stations, can work together and create join report). Reliability: If one machine crashes, the system as a whole can still survive. Higher availability and improved reliability. Ex: Nuclear reactors, aircrafts. Incremental growth: Computing power can be added in small increments. Modular expandability i. e Gradual expansion is possible. Another deriving force: the existence of large number of personal computers, the need for people to collaborate and share information

6. Advantages of Distributed Systems over Independent PCs –Data sharing: allow many users to access to a common data base.Ex: Airline/ Railway Reservation system –Resource Sharing: expensive peripherals like color printers, Phototypesetters, storage devices can be shared in DS. –Communication: make human-to-human communication easier, for example, by E-mail, chat. –Flexibility: spread the workload over the available machines in the most cost effective way

7. Disadvantages of Distributed Systems –Software: difficult to develop software, PL and applications for distributed systems. –Network: saturation, lossy transmissions. the network can saturate or cause other problems i.e may loose messages, & can be overloaded –Security: easy access also applies to secrete data.

8. Hardware Concepts of DS MIMD (Multiple-Instruction Multiple-Data) Tightly Coupled versus Loosely Coupled MIMD  Tightly coupled systems (multiprocessors)( pure DS ) oshared memory ointermachine delay short, data rate high  Loosely coupled systems (multicomputers)( EX: LAN) oprivate memory ointermachine delay long, data rate low

10. Bus versus Switched MIMD Bus: a single network, backplane, bus, cable or other medium that connects all machines. E.g., cable TV Switched: individual wires from machine to machine, with many different wiring patterns in use. Masg can be passed from m/c to m/c. The roots can be changed by using switch Ex: world wide telephonic system. Categories of MIMD( DS ) : –Multiprocessors (shared memory) Bus multiprocessor Switched multiprocessor –Multicomputers (private memory) Bus Multicomputers Switched Multicomputers

11. Bus-based Multiprocessors Bus-based multiprocessors:  hit rate  coherence: If A writes a word in Memory & if B reads the word just written by A then the memory is called as Coherent. Imp Aspect of DS.  Bus will be always Busy i.e Overloaded so performance of Sys may degraded to avoid this a high sped cache can be added in between cpu and bus.  Disadvantage of cache is memory can become incoherent. The solution is to use write-through cache.

13.  write-through cache:if a word is written to cache it should be written to memory as well i.e propagate write immediately  snoopy cache: whenever cache sees a write occuring to mem address present in cache it either removes that entry from cache or update it with new value this is called as Snoopy Cache.

14. Switched Multiprocessor

15. For Crossbar switch we need n CPUs & n Memories so n 2 crossbar switches are needed.for large n it will be difficult. Solution is use Omega Network : –contains four 2*2 switches. –Each having 2 inputs and 2 outputs –Each switch can route either input to either output.

16. Multicomputers Bus-Based Multicomputers –Each CPU has direct connection to its own local memory. –easy to build  communication volume much smaller.  relatively slow speed LAN (10-100 Mbps, compared to 300 Mbps and up for a backplane bus)

17. Bus Multicomputer

18. Switched Multicomputers Switched Multicomputers: -Each CPU has direct and exclusive access to its own, private memory. -interconnection networks: E.g., grid, hypercube - Grids : easy to understand and lay out on PCB. Best suited for 2 D nature problems Ex: Graph Theory. -hypercube: n-dimensional cube: each CPU has n connections to other CPUs. - msg passing is easy becoz of nearest neighbors are connected.

19. Switched Multicomputer

20. Software Concepts Software more important for users Three types: 1.Network Operating Systems 2.(True) Distributed Systems 3.Multiprocessor Time Sharing

21. Network Operating Systems  loosely-coupled software on loosely-coupled hardware  A network of workstations connected by LAN  Here each user has a workstation for its exclusive use along with Exclusive OS.  All commands normally run locally.  Sometimes user can log into another machine remotely orlogin machine orcp machine1:file1 machine2:file2 oTo avoid communication problems one approach is to provide a shared global file system accessible from all workstation.  Files servers: the file system supported by one or more machines is called as file servers.client and server model. Refer fig …  Clients mount directories on file servers  Best known network OS: oSun’s NFS (network file servers) for shared file systems oA situation where each m/c has high degree of Autonomy and there are a few system-wide requirements that os is called as NOS.

24. NFS NFS Architecture –Server exports directories –Clients mount exported directories NFS Protocols –For handling mounting –For read/write: no open/close, stateless NFS Implementation

25. (True) Distributed Systems  tightly-coupled software on loosely-coupled hardware  provide a single-system image or a virtual uniprocessor  a single, global interprocess communication mechanism, process management, file system; the same system call interface everywhere  Ideal definition: “ A distributed system runs on a collection of computers that do not have shared memory, yet looks like a single computer to its users.”

26. Multiprocessor Operating Systems Tightly-coupled software on tightly-coupled hardware  Examples: high-performance servers  shared memory  single run queue  traditional file system as on a single-processor system: central block cache

29. Design Issues of Distributed Systems Transparency Flexibility Reliability Performance Scalability

30. 1. Transparency How to achieve the single-system image, i.e., how to make a collection of computers appear as a single computer. A system that realizes this goal is said to be transparent. Hiding all the distribution from the users as well as the application programs can be achieved at two levels: 1)hide the distribution from users. 2)at a lower level, make the system look transparent to programs. 1) and 2) requires uniform interfaces such as access to files, communication.

31. Types of transparency –Location Transparency: users cannot tell where hardware and software resources such as CPUs, printers, files, data bases are located. The name of resource must not encode the location of resource, so machine1: prog.c or /machine1/proc.c are not acceptable. –Migration Transparency: resources must be free to move from one location to another without their names changed. E.g., /usr/lee, /central/usr/lee –Replication Transparency: OS can make additional copies of files and resources on its own without the users noticing. –Concurrency Transparency: if the system is CT then The user will not notice the existence of another user. to achieve CT, one mechanism is to lock the resource automatically when one had started to use it, unlocking it when the access was finished. Lock and unlock for mutual exclusion. –Parallelism Transparency: Automatic use of parallelism without having to program explicitly. There are times when Users do not want complete transparency.

33. 2. Flexibility Make it easier to change Monolithic Kernel: each machine should run a traditional kernel that provides most services itself. systems calls are trapped and executed by the kernel. All system calls are served by the kernel, e.g., UNIX. Adv: performance Ex: today’s centralized operating system with n/w facilities and remote services, Sprite os Microkernel: provides minimal services. 1) IPC 2) some memory management 3) some low-level process management and scheduling 4) low-level i/o E.g.,Amoeba, Mach can support multiple file systems, multiple system interfaces. Most of DS has this kind of system.

35. 3. Reliability The goal behind designing DS is to make them more reliable than single processor system. Aspects of reliability : –Availability: fraction of time the system is usable. Another tool to improve availability is redundancy i.e key pieces of h/w and s/w can be replicated. –Security : files and other resources must be protected from unauthorized users. –Fault tolerance: need to mask failures, recover from errors.

36. 4. Performance Performance problem is compounded by the fact that is communication which is essential in DS. Performance loss due to communication delays: –fine-grained parallelism: high degree of interaction, not good for DS. –coarse-grained parallelism : best fit for DS ( jobs that involve large computations, less interactions and little data) Performance loss due to making the system fault tolerant.

37. 5. Scalability Systems grow with time or become obsolete. Techniques that require resources linearly in terms of the size of the system are not scalable. e.g., broadcast based query won't work for large distributed systems. Examples of bottlenecks oCentralized components: a single mail server oCentralized tables: a single URL address book oCentralized algorithms: routing based on complete information, only decentralized algorithm should be used in Ds.

39. Characteristics of Decentralized Algorithm No m/c has complete information about the system state. m/c makes decision based on local information only Failure of one m/c does not ruin the algorithm There is no implicit assumption that a global clock exists.

40. THANX

41. Experiments & Assignments 1.Simulating NOS commands: 1.Implementation of Stack 2.Implementation of Queue. 2.RMI supporting Distributing Computing in JAVA: 1.Design a GUI based Calculator 2.Retrieve Time & Date function from server to Client. 3.Develop a Program for Client Chat Server. 4.Remote Objects for Database Access. 5.Amoeba, V-System, Mach, Chorus – Case study 6.Centralized & Distributed Algorithms for Clock synchronization. 7.Deadlock prevention & Deadlock detection algorithms. 8.Election algorithms for co-ordinator. 9.Load balancing Algorithms.