Student ID CS 736 Midterm Fear Factor Fall 2002 Please Read All Questions Carefully There are eight 8 total numbered pages Please put your NAME on this page and your STUDENT ID on this and all other pages Name 1 Student ID This is the grading page Points Total Possible Part I Short Answers 45 Part II Long Answers 55 2 Student ID Part I Short Questions For each of the following questions in this section CIRCLE ALL THAT APPLY 1 Which of the following could be said about the message based communication abstraction that is used in Nucleus a Performance Makes it easier to build a high performance system b Robustness Makes it easier to build memory protected higher level operating systems c Modularity Makes it easier to replace an existing service with a better one d Distribution Makes it easier to move a service to a distributed multi computer setting 2 When all of the data blocks in an AutoRAID storage system are constantly being written to and the system is exporting its maximum amount of capacity which of the following are true a Most of the data is in RAID 5 storage b Each write to the system could trigger both a block promotion and a block demotion c The system is performing worse than a standard mirrored RAID 1 storage system d The system is performing worse than a standard RAID 5 storage system 3 In Pilot a page has three flags associated with it referenced written and write protected Given this which of the following are true a Pilot can track to some extent which pages are more frequently accessed than others b Pilot can prevent one process from reading another process s memory c Pilot can ensure that code does not get accidentally over written d Pilot can track which pages are modified by a process 4 Which of the following components of a modern operating system are not found in some form in the THE system a Locking and synchronization b Virtual memory paging c Threads and concurrency d Protected address spaces 5 To be able to virtualize the MIPS and run IRIX on top of Disco a few changes had to be made to IRIX Which of the following changes were made to enable correct virtualization a The OS code and data had to be relocated into mapped memory b The OS idle loop had to be changed to inform Disco the OS was idling c The OS page zeroing code had to be changed to instead request a zero d page from Disco d The OS bcopy had to be changed to a special Disco remapping function during network communication between virtual machines 6 Which of the following are goals in the design of exokernel a Avoid resource management b Avoid multiplexing resources among competing processes c Avoid using physical names i e physical page numbers d Avoid visible revocation of resources 3 Student ID 7 When the LFS cleaner is determining whether a block of a file is live which of the following data structures might it consult a The inode map b The segment summary block c The inode of the file which the block belongs to d The inode of the directory that contains the file 8 Which of the following were done by Chen and Bershad in their study of the impact of the structure of operating systems on memory system performance in order to make sure they could solely focus on structurally induced costs and overheads a Slowed down the clock interrupt b Removed the effects of policy decisions made by Ultrix and Mach c Picked a set of workloads that stressed all components of the operating system d Compared the run times of the workloads across the Ultrix and Mach platforms 9 The gray box file layout detector and controller FLDC exploits FFS behavior to improve performance by reading files in sorted inode number order Under which of the following scenarios will the FLDC likely give you little or no benefit a When the file system has aged significantly b When the files that are accessed are all small i e 4KB in size c When the files that are accessed are spread out across a number of directories d When only one large file is repeatedly accessed sequentially 10 Which of the following statements can be used to describe a Hydra capability a Can t be forged b Has access rights associated with it c Can be passed from one process to another d Is managed by the kernel 11 Cylinder groups are used in FFS to a Group related data blocks together b Group related inodes together c Group related directories together d Increase disk efficiency 12 Which of the following is true about the original UNIX system a Made it easy to write test and share programs b Was built to run in a memory rich and highly CPU powered environment c Designers of the system used the system a lot d Designers were shooting for the most original and innovative system they could think of 4 Student ID 13 In comparing a standard mirrored RAID storage system RAID level 1 of D total disks to a standard block level striping storage system RAID level 0 also with D disks which of the following are always true a The mirrored system appears to be half as big as the striping system to the file system b The mirrored system tolerates more disk failures than the striping system c The mirrored system performs worse than the striping system d The mirrored system tolerates power outages more gracefully than the striping system 14 Vino prevents downloaded code or grafts from doing certain things within the kernel Which of the following potential graft actions are prevented by Vino a Changing a global scheduling policy b Changing a thread scheduling policy for the user s threads c Running an infinite loop d Grabbing locks 5 Student ID Part II Long Questions 1 RAIDing the File System 20 points In this question you will explore issues related to running file systems efficiently on top of RAID devices 1 Assume we are running a file system on top of RAID 5 Briefly describe RAID 5 and how it works A picture is worth a thousand words or at least a few points on this exam 2 Let s say we want to run LFS on top of RAID 5 First describe how LFS writes data to disk What are the basic concepts and structures involved 3 Is LFS a good file system for RAID 5 i e does it potentially solve some of the problems associated with RAID 5 performance Justify 4 To work efficiently on top of RAID 5 LFS may need to discover some parameters about the RAID system Using gray box techniques design a simple algorithm that discovers the stripe size of the RAID A loose sketch is fine no need for a lot of details State your assumptions One thing you can assume is that the block size of the RAID is known 6