Operating system support in distributed system

Operating system support in distributed system

• 1.
  Operating System Support •The OS must provide: - Encapsulation - Concurrent Processing - Protection • OS components: - Process Manager - Thread Manager - Communication Manager - Memory Manager - Supervisor Communication manager Threadmanager Memorymanager Supervisor Process manager 1 Isha Padhy, Department of CSE,CBIT, Hyderabad
• 2.
  Operating System Layer Applications,services Computer & Platform Middleware OS: kernel, libraries & servers networkhardware OS1 Computer & networkhardware Node 1 Node 2 Processes, threads, communication, ... OS2 Processes, threads, communication, ... 2 Isha Padhy, Department of CSE,CBIT, Hyderabad
• 3.
  Process/Thread Concepts • Whatare Processor, process, threads? Thread concepts: 1. Address Space - Region Fixed: un-modifiable text region containing program code. Heap: part of which is initialized by values. - There are indefinite no. of regions because a separate stack is needed for each thread. Fixed Heap Stack 3 Isha Padhy, Department of CSE,CBIT, Hyderabad
• 4.
  2. Creation ofnew processes a. Choice of a target host: Node where process will reside, consists of some policy. 1. Transfer Policy: Whether to situate a new process locally or remotely depending on load of local node. 2. Location Policy : which node should host a new process. Location Policy can be a. Static: optimize overall process throughput policy operate w/o regard to current state of the system. - Deterministic: ‘A’ should always transfer process to ‘B’ - Probabilistic: ‘A’ should transfer to any of nodes between ‘B’ to ‘E’ b. Adaptive: Apply methods to make allocation( Ex. Measure load at each node) 4 Isha Padhy, Department of CSE,CBIT, Hyderabad
• 5.
  b. Creation ofnew execution environment: • Two approaches to define address space of newly created process: 1. Statically defined- Address space regions are initialized from an executable file. 2. Defined w.r.t. existing execution envt. Ex. In UNIX the newly created child process can physically share parent’s address space. The region is not getting physically copied but the page frames that make up the inherited region are shared between 2 address spaces.(COW technique) 5 Isha Padhy, Department of CSE,CBIT, Hyderabad
• 6.
  COW(Copy on Write) a)Before write b) After write Shared frame A's page table B's page table Process A’s address space Process B’s address space Kernel RA RB RB copied from RA 6 Isha Padhy, Department of CSE,CBIT, Hyderabad
• 7.
  Threads in DistributionSystem: Multi-threaded Clients • Requirement: conceal the inter-process message propagation time. To hide communication latencies we can initiate communication and proceed with other work. Ex downloading a web page. With Waiting single thread Without Waiting single thread Multi-threading Every element like image, text ,icons etc in a web page requires a TCP/IP connection. So waiting time time will be more if we wait for all the elements to download first and use the web page. Fetch the HTML page and subsequently display it. Display text while downloading image. As soon as main HTML page is fetched every element is fetched with separate threads. 7 Isha Padhy, Department of CSE,CBIT, Hyderabad
• 8.
  Multi-threaded Servers - Serverson uni-processor systems can be developed to exploit parallelism with multithreading. - Ex. File operations with multithreading, without multi-threading, finite- state- machine 8 Isha Padhy, Department of CSE,CBIT, Hyderabad
• 9.
  Architecture of Multi-threaded Servers 1.Worker- pool architecture: Server creates a fixed no. of “worker threads” to serve the requests. Server N threads Input-output Client Thread 2 makes T1 Thread 1 requests to server generates results Requests Receipt & queuing 9 Isha Padhy, Department of CSE,CBIT, Hyderabad
• 10.
  2. Thread- per-request architecture: 3. Thread- per- connection architecture 4. Thread-per-object architecture a. Thread-per-request b. Thread-per-connection c. Thread-per-object remote workers I/O remote remote I/O per-connection threads per-object threads objects objects objects 10 Isha Padhy, Department of CSE,CBIT, Hyderabad
• 11.
  Thread Implementation Kernel LevelImplementation: - Kernels support multi-threaded processes ex. Windows. - These kernel provide thread creation and management system calls, schedule threads. - Threads within a process can take advantage of multiprocessor. - If a thread gets blocked the whole process is not blocked. 11 Isha Padhy, Department of CSE,CBIT, Hyderabad
• 12.
  User- level Implementation •Thread scheduling module is implemented outside the kernel so it can be customized for a particular application. • No system call for switching between threads of same process. • The process notifies the kernel when either: - Virtual Processor(VP) * is idle - Extra virtual processor is required * Kernel is running on a computer with one or more virtual processor( physical processor allotted to virtual machines with different time slot). * Kernel allocates one or more VP to the process threads which are controlled by user level scheduler within each process. 12 Isha Padhy, Department of CSE,CBIT, Hyderabad
• 13.
  The kernel notifiesthe process when any of 4 events occur:- 1. Virtual processor allocated: VP allocated to process, this is the first time slice for it. The scheduler (user level) can load the SA* with the context of a READY thread. 2. If SA is blocked, kernel uses a fresh SA to notify scheduler. The corresponding thread is BLOCKED and a READY thread is allocated to notifying SA. 3. If SA unblocked, kernel informs scheduler to release thread as READY and either allocates a new VP if available or pre-empts another SA in same process. 4. If SA pre-empted , scheduler places the pre-empted thread to READY list and reevaluates the thread allocation. * SA( Scheduler Activation) - Call from kernel to process to notify allocation of a VP. No. of SA= No. of VP assigned to process 13 Isha Padhy, Department of CSE,CBIT, Hyderabad
• 14.
  Operating system Architecture •Monolithic kernels • Micro kernels • Hybrid Kernels 14 Isha Padhy, Department of CSE,CBIT, Hyderabad