COP Operating Systems Design Principles (Spring 2012)

Transcription

1 COP Operating Systems Design Principles (Spring 2012) Homework 1 Problem 1: 1.1 Difference in physical and the logical storage units affecting (i) the readwrite coherence and (ii) the atomicity. This example of storage abstraction consists of a primary and secondary hard drive working together in a computer. The physical drive units are managed by a single logical unit called Logical volume manager (LVM). The LVM is run by the operating system. Read write coherence: Consider that, there are multiple processes that continuously write the data alternatively between the primary and secondary drive. During the read operation, if the primary and secondary drive is being read in a different sequence without coherence, then the read operation would not read the recently updated data. Similarly, the logical unit (LVM) will also be affected when it is not allowing the process to read coherently with most recent write. This causes writing of wrong state values in the partitions. Atomicity: Consider the primary and secondary physical drives are used in a server that handles credit card transactions. If there is a power failure or hard disk failure and during these situations if only the primary drive has written the data, then after reboot the hard drives would reflect wrong states. Thus the physical units should make sure that the data write during failures should be completed or cleared. Similarly, if there are two processes, each writing data in its corresponding hard drive number fails to complete writing in one logical partition before writing in the next, it affects before-or-after atomicity. Thus it is very important for both physical and logical units to meticulously maintain read write coherence and atomicity. 1.2 Illustrating the phenomena of (i) incommensurate scaling; (ii) the propagation of effects for each of the three types of systems the microprocessors, the communication hardware, and the optical storage. Incommensurate scaling is a property of most systems that as the system grows (or shrinks) in size, not all parts grow (or shrink) at the same rate, thus stressing the system design.

2 Propagation of effects is a property of most systems: a change in one part of the system causes effects in areas of the system that are far removed from the changed part. A good system design tends to minimize propagation of effects. Microprocessors: Incommensurate scaling: Comparing Intel s various processors 4004(First Processor) has 2300 transistors and clock rate of 740KHz, 8080 has 6000 transistors and clock rate of 2MHz whereas the latest processor Core i7 has 995 million transistors and clock rate of 3.8 GHz. We see that the ratio of these two quantities in the 3 processors is not proportional. This is incommensurate scaling where different parts of the system do not have the same growth. The propagation of effects: With the above processors as the number of transistors increases the heat dissipated also increases which in turn affects other parts of the processors. This illustrates propagation of effects. Communication hardware: Incommensurate scaling: In multi radio multi interface networks there can be a more than one channels through which the communication can take place. Increasing the number of channels can increase the capacity and thereby result in better throughput. However increasing the number of channels beyond a certain value will result in interference therefore there has to be a check on this. The propagation of effects: In the above example increase in number of channels cause increased interference which might lead to bad quality of the received data. The effects are propagated. Optical storage: Incommensurate scaling: With the increasing capacity or size of the optical storage the time for retrieval of data keeps increasing. The propagation of effects: More data storage more chances of it getting corrupted. Or with the increased size and features the cost of it keeps increasing.

3 1.3 Example illustrating pure and composite physical realization for each of the three basic abstractions. (1) Interpreter, (2) storage, and (3) communication channels. To illustrate the difference in pure and composite system, the example of a cloud storage system is chosen. Pure system: Interpreters: The message passing between user device and cloud storage virtualization software. Communication channel: Ethernet wired cables, even wireless. Storage system: Fault tolerant cloud virtual storage software as the logical unit and high performance physical solid state drives. Composite system: A Border Gateway Protocol (BGP) can be considered as a composite system with, Interpreters: The BGP daemon itself. Communication channel: Ethernet wired cables, underwater fiber optic cables. Storage: Autonomous system lookup table (stores multiple autonomous system to route data from one edge to another in the internet. Problem 2: Answer: Yes the Harvard architecture reduces the possibility of errors due to soft modularity. Discussion: The specific errors that can be prevented by the Harvard architecture. 1. In Von Neumann architecture, the shared bus between the program memory and data memory leads to the Von Neumann bottleneck, which is the limited throughput between the CPU and memory compared to the amount of memory. This seriously limits the effective processing speed when the CPU is required to perform minimal processing on large amounts of data. Whereas Harvard architecture machine has distinct code and data address spaces. Thus it has increased throughput as compared to Von Neumann architecture. 2. In computers with von Neumann architecture the CPU can either read an instruction or read/write data from/to the memory. Both cannot occur at the same time since the instructions and data use the same bus system. Whereas with Harvard architecture, the CPU can both read an instruction and perform a data memory access at the same time. A Harvard architecture

4 computer can thus be faster for a given circuit complexity because instruction fetches and data access do not contend for a single memory pathway. Problem 3: 3.1 Advantages of the HTTP servers being stateless. 1. The server design is simplified as it does not have to dynamically allocate memory for each and every client to store their information. 2. If a client s connection is abruptly cut in the middle of a transaction, there is no necessity of changing the current state of the server. 3. The number of active clients can be considerably high as a client consumes resources on the web server only during the processing of a request-response cycle and during the time delay between the response and the next request the server can accommodate other clients. 3.2 Dealing with errors like lost data packets, failure of the server, during the transfer of an object consisting of a large number of blocks. 1. Errors within a packet i.e. bit level errors are taken care by checksum. 2. Most of the http connections run over TCP as it is a reliable protocol. Lost data packets are taken care by TCP by retransmissions. The sender either waits for timeout or uses duplicate acknowledgements (in the case of fast retransmit) or retransmits the lost packet. 3. Failure of the servers is handling by sending the corresponding error codes to the sender. Example: Connection timed out, server unavailable, etc. 4. Objects consisting of a large number of blocks are usually identified and reassembled at the receiver. So the receiver keeps track of the last received sequence number of the data block. Any missing data block is identified and retransmitted. 3.3 Servers maintaining information about clients. Servers can maintain client information if they are statefull. These are done by the servers to connect back to the clients. While browsing a webpage the user is able to navigate to the forward and back page and this is possible because the server still has a connection running with the client. This happens in TCP a there is a proper connection establishment and connection close. If the client abruptly disconnects from the server the server still has the connection for the client open until it times out.

5 Problem 4: As all the processes are allowed to read simultaneously there is no necessity for priority this portion of the system. Priority comes into picture only during write. The processes write with the priority p1>p2>p3>p4. Problem 5: Name mapping algorithms. Three main types of name mapping algorithms are table lookup, recursive lookup and multiple lookup. Those used in the internet are as follows. 1. DNS: Domain Name System. Here the domain name is mapped to a unique IP address. It follows hierarchy. DNS operation is at the application layer. 2. NAT: Network Address Translation. Private address is translated into public address to be identified by the outside world. It is a many to one mapping.

6 3. ARP: Address Resolution Protocol. IP address is broadcasted and the corresponding Mac address is got. RARP is the vice versa. 4. Other mappings possible are the MAC addresses to the port numbers in a switch. Processing of IP packets on the receiving host Problem 6: Stack Algorithm Slot 1

7 Slot 2 Slot 3 Slot 4

8 Slot 5 Slot 6

9 Slot 7 Slot 8 Slot 9

10 Slot 10 Slot 11 Slot 12

11 Problem 7: FCFS Algorithm Slot 1 Slot 2 Slot 3

12 Slot 4 Slot 5 Slot 6 Slot 7

13 Slot 8 Slot 9 Slot 10 Slot 11

14 The algorithm required 11 time slots to solve all the collisions.