Consider a multithreaded computer system that uses the following model to execute programs. Each thread consists of three phases: preload, execute, and post-store. A memory processor (MP) executes...

Consider a multithreaded computer system that uses the following model to execute programs. Each thread consists of three phases: preload, execute, and post-store. A memory processor (MP) executes preload and post-store phases, providing access to memory resident data. An execute processor (EP) provides service during execute phase. New threads are enabled as some threads complete their execution and supply data to waiting threads (modeled as Synch service). Consider the following queueing network as a model of the system. The service time of MP is based on the average number of load and store instructions, while the service time of EP is based on the average number of non-memory instructions. The service time at Synch depends on the average number of inputs needed by a thread (and provided by other threads). Explore the response time of such a system for different number of threads (jobs in the system), by varying the service times at each server. You should examine any available benchmarks to estimate the various service times. Assume that each thread visits EP and Synch once, and visits MP twice. For example, you can try with these service times: MP = 3; EP = 10, Sync = 6 (the time unit is one instruction cycle, and using a 1 GHz processor, the time unit is a nanosecond). Using N = 10, 20, 50 threads, perform MVA for this exercise.