Potpourri_1 Giant-Scale Services   1. [2 points] (True/False with justification) Admission control in Giant-scale services is a way of maintaining yield while reducing harvest. 

Failures_and_Recovery_1b LRVM   The context for this question is the same as the previous question. 1. A service that uses the LRVM library is in the middle of an LRVM transaction. There is a power failure before the transaction completes.   (b) [2 points] Upon crash recovery what will be the state of the data segment that was being modified within the transaction at the time of the power failure?

Real_time_and_Multimedia_1b TS-Linux   The context for this question is the same as the previous question. 1. Assume that you are playing a game (G) on a version of TS-Linux in which the scheduler strictly follows the born priority of all the processes in making its scheduling decisions. The game uses ONE-SHOT timer with an overshoot parameter of To. The timer event goes off at time 0. Assume the following: When the timer goes off, the kernel has just started a critical section of duration Tk. For taking the necessary gaming action for the timer event, the video game application (G) has a processing time of Tg and priority P1. In addition to Tg, the application will need to make a blocking call to a Window Manager process (W) has a processing time of Tw and priority P3. When the timer goes off, the scheduling queue is empty. The priorities are ordered as P1 > P2 > P3 > P4 > …. With P1 being the highest priority. There are no periodic timers in the system. Duration of the scheduling time and interrupt handing time are negligible. Note that the given image is simply for your reference, and that the time Th has nothing to do with the solution. Only use the variables given in the text in your solutions.   (b) [2 points] In the middle of W’s execution, another application process (S) whose duration is Ts and priority P2 becomes runnable. Does this change your calculation in the previous part of the question?

Potpourri_5 PTS   5. [2 Points] Imagine multiple producers putting items into the same channel in PTS.  Do the threads need to use mechanisms like pthread’s mutex lock to avoid stepping on one another?  Why or why not?

Security_1 AFS   1. [5 points] Consider the login process in AFS. You are a user of the system and you are assigned a username “user1” and password “pwd123”. You are using the system to make changes to a file right before the assignment deadline and, in a rush, you try to login to the system using the password “pwd1234”. List in detail the steps (the login phase and the potential steps in the bind phase) that will occur when you enter this password till the point you are denied access to the system.

Failures_and_Recovery_2 LRVM   2. [2 points] (True/False with justification) During log truncation, the service continues execution making changes (if need be) to the log. 

Potpourri_4 PTS   4. [2 Points] (True/False with justification) PTS runtime system provides global clock synchronization to enable temporal correlation of timestamped items from different threads into different channels. 

Potpourri_6 QuickSilver   6. [2 points] The number of Transaction Managers (TMs) in Quicksilver (choose one of the following):  One per node in the distributed system  One per transaction tree created in the system  One per application running on Quicksilver  One for the entire distributed system 

Failures_and_Recovery_3c RioVista   The context for this question is the same as the previous question. 3. Consider RioVista, which implements LRVM semantics on top of Rio file cache.   (c) [2 points] Compare the log truncation code of RioVista to that of Satya’s LRVM.

Potpourri_8 QuickSilver   8. [2 points] (True/False with justification) Quicksilver provides a more complete recovery management compared to LRVM.