Unit Two Objectives CSCI 524 Distributed Operating System Design Design Considerations for the Transformation of MINIX into a Distributed Operating System. 1)Assuming a performance design goal, describe three ways in which the DOS design will be constrained. 2)Describe the role of a "PID" and how it will be extended in the DOS and the new or expanded services that will use the PIDs. 3)Explain what is meant by the expression "building DOS components subservient to the BOS" and explain how this differs from other DOS designs (five parts). 4)Describe the role of the network manager. 5)Explain in your own words how the network manager "acts as the interface between local synchronous operating system services and network-wide asynchronous services. 6)Describe the role of the resource manager and explain how it addresses the global state problem. 7)Describe the buffering problem observed in the V-kernel. 8)Describe the role of the communication manager and explain how it addresses the buffering problem. 9)Explain the following "srcv(mqual, pid, &buff, sizeof(buff));" 10)Explain how a process becomes blocked on the CDQ. 11)Explain how a remote user unblocks the local user. 12)Explain how message fragmentation is accommodated. Omit)Explain how the communication manager processes incoming requests for which there is no processes. 14)Explain why or why not "strict inter-host synchronization is not required. Omit)Describe the sequence of events that allows a local host to detect and access files on another host. Omit)Describe the sequence of events that allows a network-based request access to local resources. 17)Describe Sun's NFS design and explain how this design differs. Omit)Describe the actions required to move a process image between two hosts. 19)Support or refute the concept of mobile process images is a "bad" idea (be specific). 20)Explain how the proposed DOS design works towards the following five conclusions: a)Elimination of crosspoint switches. b)Elimination of implicit network buffering schemes. c)Elimination of expensive context switching. d)Handling of multiple network requests. e)Elimination of duplicate BOS and DOS services. PODOS - The Design and Implementation of a Performance Oriented Linux Cluster 1)Support or refute that PODOS offers a testbed for "a clustering system, but at the same time provides a good resource sharing environment." (Unified file system? Traditional network stack?) 2)Describe the PODOS network topology (three parts). 3)Describe the PODOS packet protocol and relate its design to traditional network protocol stacks. 4)Describe six fields of a CDT record, and describe the two types of communication supported by the communication manager. 5)Define "transmission group" and explain how it is implemented in PODOS as it relates to interfaces, scheduling, and protocol. 6)Describe the goal and results of Experiment 1, PODOS vs Sock. 7)Describe the goal and results of Experiment 2, Sock&PODOS vs Tran grp. 8)Describe the goal and results of Experiment 3, PODOS vs Tran grp. 9)Describe the goal and results of Experiment 4, CDT vs Sock. 10)Contrast NFS, Coda, and PDFS design goals. 11)Describe the control flow pass-off mechanism between the local file manager and the PODOS network manager. 12)Explain how the Linux kernel clone() service call differs from the traditional user level fork() system call. Explain how PODOS exploits this feature to implement its PDFS. 13)Given the design goals of PODOS, support or refute the use of read-ahead caching. 14)Describe the goal and results of the file fetch duration experiment with NFS versus PDFS for idle and lightly loaded Ethernets. 15)Explain the extension to PODOS PIDs and the file read/write semantics and explain the resulting design goal.