Note: Answers are given here at the end to check to see if you are correct. You will get zero if you don t show your work or if you copy my answers. Taber and I can t read your mind. J 1) A 2-way set-associative cache consists of four sets. Main memory contains 2K blocks of eight words each and word addressing is used. Show the main memory address format that allows us to map addresses from main memory to cache. Be sure to include the fields as well as their sizes. Tag: 9 bits (14-2-3) Set: 2 bits (2^2 sets) Offset: 3 bits (2^3 words per block) 2) Suppose a computer using fully associative cache has 2^16 bytes of byte-addressable main memory and a cache of 64 blocks, where each cache block contains 32 bytes. a. How many blocks of main memory are there? 2^16/2^5 = 2^11 blocks of main memory b. What is the format of a memory address as seen by the cache, i.e., what are the sizes of the tag and offset fields? Tag: 11 bits (16-5) Offset: 5 bits (2^5 bytes in ea. block) c. To which cache block will the memory address F8C9 map? Since it is a fully associative cache, it can map to any available address. 3) Suppose a process page table contains the entries shown below. Using the format shown in Figure 6.15a, indicate where the process pages are located in memory. Virtual Memory Physical Memory 0 1 2 0 3 3 6 2
4) Calculate the overall speedup of a system that spends 65% of its time on I/O with a disk upgrade that provides for 50% greater throughput. (Use the Amdahl s Law) S = 1/((1-.65)+(.65/1.5)) S = 28% 5) Calculate the overall speedup of a system that spends 40% of its time in calculations with a processor upgrade that provides for 100% greater throughput. (Use the Amdahl s Law) S = 1/((1-.4)+(.4/2)) S = 25% 6) Your friend has just bought a new personal computer. She tells you that her new system runs at 1GHz, which makes it over three times faster than her old 300 MHz system. What would you tell her? (Hint: Consider how Amdahl's Law applies.) According to Amdahl s Law, the increase in system speed is proportional to the utilization of the upgraded component. The new processor that my friend just bought can certainly process data more than 3x as fast, but this does not indicate the entire system will increase by a factor of 3 because there might be other bottlenecks involved. 7) Define the terms seek time, rotational delay, and transfer time. Explain their relationship. Seek time is the time it takes for a disk arm to position itself over the required track. This metric does not include the time that it takes for the head to read the disk directory. Rotational delay is the time that it takes for the required sector to position itself under a read/write head. The sum of the rotational delay and seek time is known as the access time. Similarly, the sum of the access time and the time that it takes to actually read the data from the disk, is known as transfer time. 8) Do you trust disk drive MTTF figures? Explain. No, because the Mean time to Failure statistic is unregulated, which gives the manufacturer free roam on how to portray the figures. The Design Life figure is more accurate. 9) The book said that disks are power hungry as compared to RAM. Why? Disks use more power than RAM because they have to constantly spin platters to seek to the correct location for the requested data. 10) Talk about pros and cons of using disk versus tape for backups. Disk is more expensive than tape backups, but tape is slower to access and write to. 11) You have been hired by an ecommerce company and the company run particular highperformance computer system has been functioning as an e-business server. a) The fact that this system is not using any type of RAID is disturbing to you. Please convince your CIO (Chief Information Officer) to use some type of RAID and support your ideas. Please
explain your CIO of the history of RAID. Your CIO is still not sure about the RAID because he is obsessed with system performance. He feels that more revenue would be lost in the long run if the system slows after RAID is installed. RAID is essential where sensitive data cannot be lost at any cost. There are a few different designs to choose from. RAID-0 is good for system performance but it doesn t have any redundancy. RAID-5 is a good option because it provides performance and data security. b) Congratulations! Your CIO is finally convinced by your ideas and he is ready to buy RAID system. He has narrowed it down to RAID#1 and RAID #5 configurations. Please help him pick the best fit for the current system. Remember, your CIO still care about the performance and cost. Please include pros and cons of each configuration. Out of all the RAID configuration, which RAID level is cheapest? Why? RAID-5 is the best option because it provides fast performance because of data striping, and it also includes redundancy incase of a disk failure. 12) a) Which of the RAID systems described in this chapter cannot tolerate a single disk failure? RAID-0 cannot tolerate a single disk failure because the data is split up amongst all the drives in the array and there is no redundancy. b) Which can tolerate more than one simultaneous disk failure? RAID-6 because it uses a two dimensional parity. 13) Suppose you have just been assigned to a committee that has been tasked with purchasing a new enterprise file server that will support customer account activity as well as many administrative functions, such as producing a weekly payroll. (Yes, a committee frequently makes these decisions!) One of your committee members has just learned that a particular system has blown out the competition in the SPEC CPU2000 benchmarks. He is now insisting that the committee buy one of these systems. What would be your reaction to this? He is basing this purchase on one benchmark. They should look at more benchmark and actual reviews, not just one metric. 14) Suppose you wish to run a program P with 7.5 x 10^9 instructions on a 5 GHz machine with a CPI of 0.8. What is the expected CPU time? (7.5x10^9)*(.8 cycles)*(1/(5x10^9)) = 1.2 seconds 15) Program A runs in 0.34 seconds on a 500 Mhz machine. You know that this program requires 100 million instructions. Calculate CPI..34 = (1x10^8)*(1/(5x10^8))*CPI
CPI = (1/5)*.34 CPI = 1.7 cycles/instruction 16) A program runs in 10 seconds on computer A, at 4GHz. How do you build a computer B to run this program in 6 seconds? The designer has determined that if the clock rate will be increased, it will cause computer B to require 1.2 times more clock cycles than computer A. What clock rate should be used in computer B? 10 = x*(4ghz), x = 4x10^10 6 = 1.2x*(rate) 6/(1.2*(4x10^10)) = 1/(8x10^9) = 8 GHZ 17) Computer A clock cycle time 250 ps, CPI of A = 2 and Computer B clock cycle time 500 ps, CPI of B = 1.2. Assume A and B have same instruction set per program. Which statement is true? -->A) Computer A is 1.2 times faster than B B) Computer A is 4.0 times faster than B C) Computer B is 1.7 times faster than A D) Computer B is 3.4 times faster than A (500*1.2)/(250*2) = 1.2x faster 18) We wish to compare the performance of two different computers: M1 and M2. The following measurements have been made on these computers: Time on M1 Time on M2 Program 1 2.0 seconds 1.5 seconds Program 2 5.0 seconds 10.0 seconds Which computer is faster for each program, and how many times as fast is it? Program 1: M1 is 2/1.5 times faster Program 2: M2 is 10/5 times faster (2x) 19) What is the difference between scalability and elasticity? Scalability is used to describe how well a cloud computing system can dynamically adapt to an increased workload demand. Elasticity is use to describe how well the system can scale outwards with an increased demand. 20) What is the issue with using mean to measure performance?
There are outliers in the data that can skew the results. 21) What is benchmarking? A process that attempts to compare different components on a certain scale to judge the ability of each. 22) From the mainframe era, then the PC era, and some declare that we are now entering a third era. What characterizes this new era (makes it distinct from the previous one), and what is an application / service / system that characterizes it? This new era on computing consists of cloud computing and mobile devices. This provides faster performance for less money. Also, this allows end-users to carry enough hardware to access the cloud that can execute to the necessary processes. 23) Cloud computing seems to be the panacea for data or compute-intensive problems facing companies today, with very small upfront costs and seemingly infinite capacity on demand. With so much going for it, why are some companies staying away? Cloud computing is highly dependent on the Internet, and this presents a huge security risk for large companies with sensitive data. 24) When comparing two systems one of the best way to measure is weighted average. What is the issue with weighted average? Weighted average does not take into account assumptions that do not hold over time. 25) A certain microprocessor requires either 2, 3,4, 8, or 12 machine cycles to perform various operations. Twenty-five percent of its instructions require 2 machine cycles, 20% require 3 machine cycles, 17.5% require 4 machine cycles, 12.5% require 8 machine cycles, and 25% require 12 machine cycles. a. What is the average number of machine cycles per instruction for this microprocessor? 0.25 * 2+ 0.2*3 + 0.175 * 4 + 0.125 * 8 + 0.25 * 12 = 5.8 cycles/instruction b. What is the clock rate (machine cycles per second) required for this microprocessor to be a 1 MIPS processor? 5.8 cycles/instruction * 1,000,000instructions/second = 5.8GHz 26) Read Google paper on Warehouse Scale Computing (WSC). Read Chapter 1, 2.4, 5.1-5.3. I will throw the questions in the quiz and final. Focus on main ideas. Click the PDF link below http://www.morganclaypool.com/doi/abs/10.2200/s00193ed1v01y200905cac006 Answers:
1) 14-bit addresses with 9 bits in the tag field, 2 bits in the set field (since we have four sets), and 3 in the offset field. 2) a. 2^11 b. 16 bit addresses with 11 bits in the tag field and 5 in the offset field c. Since it is associative cache, it can map anywhere 3) 4) 1.28 or 28% 5) 1.25 or 25% 6) You would explain Amdahl s Law. The processor is only one component contributing to the overall performance of a system. Seek time is the time it takes for a disk arm to position itself over a requested track. 7) Rotational delay is the time that it takes for the required sector to position itself under a read/write head. The sum of the rotational delay and seek time is known as the access time. If we add to the access time the time that it takes to actually read the data from the disk, we get transfer time. 8) Answers in the textbook 9) Not given here. 10) Not given here 11) Support your points. There is no single solution. The point of this question is that I wanted everyone to master the RAID system. 12) a. RAID-0. b. RAID-1, RAID-2 and RAID-6. RAID-1 can tolerate multiple disk failures only if the failure does not involve a disk and its mirror image. 13) First, be diplomatic. Suggest that the group investigate whether TPC-C benchmarks are available for this system. Whether or not TPC-C figures are available, you may want to educate this individual as to the meaning of Amdahl s Law and why a fast CPU will not necessarily determine whether the entire system will handle your workload.
14) 1.2 seconds 15) CPI = 1.7 16) Clock Rate B = 8 GhZ (Computer B needs to be 8GhZ) 17) A is true and explain why here to get credit 18) For P1, M2 is 4/3 (2 sec/1.5 sec) times as fast as M1. For P2, M1 is 2 times (10/5.0) times as fast as M2. 19) Answers: Scalability is a characteristic of cloud computing through which increasing workload can be handled by increasing in proportion the amount of resource capacity. It allows the architecture to provide on demand resources if the requirement is being raised by the traffic. Whereas, elasticity is being one of the characteristic provide the concept of commissioning and decommissioning of large amount of resource capacity dynamically. It is measured by the speed by which the resources are coming on demand and the usage of the resources. 20) Answer not given here. Please read your text book in chapter 11. 21) Read Chapter 11 22) It's an era of mobile, social, and ubiquitous computing. Examples? Tweeting/facebook from a smartphone, exploring a city. A lot of processing are also happening on the cloud also. 23) Concern about... 1) service availability (trusting your data to someone else) 2) data lock-in (getting your data out when needed) 3) security / confidentiality / auditability 4) data transfer bottlenecks 5) performance unpredictability 6) scalable storage 7) the ability to scale up quickly. 24) Please read chapter 11 25) a) 0.25 * 2+ 0.2*3 + 0.175 * 4 + 0.125 * 8 + 0.25 * 12 = 5.8 cycles/instruction b) 5.8 cycles/instruction * 1,000,000instructions/second = 5.8 26) Will quiz high level ideas on quiz and final.