Global Memory Systems [5 points] Consider a GMS system with 10 Nodes ‘A’ to ‘J’. A process ‘x’ on Node ‘A’ intends to read-share a page that is  currently actively used only by a process ‘y’ on Node ‘B’.When Node A’s GMS mechanism initiates the access, indicate on which Node(s) each of the following data structures  are accessed.Assume no co-location constraints for any of the data structures a) [1 point] POD

RPC Latency Limits The context for this question is the same as the previous question. [6 points] A legacy low-latency RPC system for a distributed operating system operates over 100 Gbps Ethernet hardware that occasionally exhibits packet corruption, even when Ethernet CRC checks pass. To guard against such errors, the system currently uses software-based UDP checksums, which add about 180 µs per RPC, contributing roughly 7% of the total latency for a minimal RPC call. An improved integrity verification mechanism is being considered to reduce latency while maintaining correctness. The following 2 approaches are being considered – Hardware-only -> Use NIC checksum offload, no software checksums Hardware Software Combined -> Use NIC checksum offload together with an additional lightweight application-level checksum in the RPC protocol. b) [2 points] If you have machines that have cross-datacenter communication across the Wide Area Internet, what check-sum strategy would you use? Justify your answer.

Distributed File Systems [2 points] In xFS Paper, it’s mentioned that metadata Manager Recovery is not Implemented. Propose a high-level algorithm (conceptual level, no need for data structures) for how the xFS system could safely recover from a manager’s crash?

Potpourri [2 points] Answer True/False with justification. No credit without justification. In Treadmarks, upon a page fault for a page X in Node N1, the DSM software on N1 broadcasts the virtual page number (VPN) of the faulting page to all the peer nodes.

Distributed File Systems Background: The Google File System was designed a few years after xFS and is heavily influenced by xFS. GFS also separates its metadata management logic from its data storage. It uses a single “Master” server to store all file system metadata. It uses many “Chunkservers” to store actual data blocks. A client first contacts the single Master to get metadata and then contacts the Chunkservers directly to get data. Give one pro and one con of the GFS approach to metadata management.

EJB [2 points] An EJB application built for an internal office network works well. The company now wants to make this service securely available to remote clients over the public Internet (a slow, high-latency WAN). The service uses fine-grained data objects to get parallelism and performance (like Entity Beans). Explain why a developer would strongly recommend implementing a Session Facade pattern before exposing this service to the wide area network

Potpourri The context for this question is the same as the previous question. [6 points]  Your team is experimenting with different RPC communication methods for designing a distributed system. Two design choices are being considered- Design A involves LRPC-style mechanism where the client and server communicate through shared memory in an SMP. Design B involves traditional RPC where the client and server communicate across the LAN as discussed in the Thekkath & Levy paper. d) [1 point] For design B, what is the minimum number of context switches required for a single, blocking RPC call-reply cycle?

Potpourri [2 points] Answer True/False with justification. No credit without justification. Multi-writer coherence protocol works with SC, RC and LRC memory models.

Which condition can occur without any motor weakness?

Reduced loudness and lack of intonation (monotone) is a feature of