One of the biggest impediments to widespread live of virtual machines is the per

Question

One of the biggest impediments to widespread live of virtual machines is the performance overhead incurred by running a virtual machine. The table below lists various performance parameters and application behavior. Calculate the CPI for the system listed above assuming that there are no accesses to I/O. What is the CPI if the VMM performance impact doubles? If it is cut in half? If a virtual machine software company wishes to obtain a 10% performance degradation, what is the longest possible penalty to trap to the VMM? I/O accesses often have a large impact on overall system performance. Calculate the CPI of a machine using the performance characteristics above, assuming a nonvirtualized system. Calculate the CPI again, this time using a virtualized system. How do these CPIs change if the system has half the I/O accesses? Explain why I/O bound applications haw a smaller impact from virtualization. Compare and contrast the ideas of virtual memory and virtual machines. How do the goals of each compare? What are the pros and cons of cach? List a few eases where virtual memory is desired, and a few cases where virtual machines are desired.

Explanation / Answer

please rate - thanks 5.14.1 a. CPI: 2.0 + (100/10,000 × (20 + 150)) = 3.7 CPI: 2.0 + (100/10,000 × (20 + 300)) = 5.2 CPI: 2.0 + (100/10,000 × (20 + 75)) = 2.95 To obtain a 10% performance degradation, we must solve: 1.1 × (2.0 + (100/10,000 × 20)) = 2.0 + (100/10,000× (20 + n)) We fi nd that n = 22 cycles b. CPI for the system with no accesses to I/O CPI: BaseCPI + ((priv OS access/10000) × (perf impact trapguestOS + perf impact trap VMM)) CPI: BaseCPI + ((priv OS access/10000) × (perf impact trapguestOS + 2 × perf impact trap VMM)) CPI: BaseCPI + ((priv OS access/10000) × (perf impact trapguestOS + 0.5 × perf impact trap VMM)) CPI: 1.5 + (110/10,000 × (25 + 160)) = 3.535 CPI: 1.5 + (110/10,000 × (25 + 320)) = 5.295 CPI: 1.5 + (110/10,000 × (25 + 80)) = 2.655 To obtain a 10% performance degradation, we must solve: 1.1 × (BaseCPI + (priv OS access/10000 × perf impacttrap guestOS)) = BaseCPI + ((priv OS access/10000) × (perf impact trapguestOS + n)) 1.1 × (1.5 + (110/10000 × 25)) = (1.5 + (110/10000× (25 + n))) 1.1 × 1.775 = 1.5 + (0.011 × (25 + n)) 1.9525 – 1.5 = 0.011 × (25 + n) 0.4525/0.011 = 25 + n we fi nd that n = 20 cycles is longest possible penalty to trap tothe VMM 5.14.2 a. CPI, non virtualized = 2.0 + 80/10,000 × 20 + 20/10,000× 1000 = 2.0 + 0.16 + 2.0 = 4.16 CPI, virtualized = 2.0 + 80/10,000 × (20 + 150) + 20/10,000× (1000 + 150) = 2.0 + 1.36 + 2.3 = 5.66 I/O bound applications have a smaller impact from virtualizationbecause, comparatively, a much longer time is spent on waiting for the I/O accesses tocomplete. b. CPI (non virtualised): BaseCPI + (priv OS access-I/Oaccesses)/10000 × perf impact trap guestOS) + I/O accesses/10000 × I/O access time CPI (virtualised): BaseCPI + (priv OS access-I/O accesses)/10000× (perf impact trap guestOS + perf impact trap VMM)) + (I/O accesses/10000 × (I/Oaccess time + perf impact trap VMM)) CPI (non virtualised): 1.5 + (110 – 10)/10000 × 25 +10/10000 × 1000 = 1.5 + 0.225 + 1 = 2.725 CPI (virtualised): 1.5 + (110 – 10)/10000 × (25 + 160)+ 10/10000 × (1000 + 160) = 1.5 + 1.665 + 1.16 = 4.325 I/O bound applications have a smaller impact from virtualizationbecause, comparatively, a much longer time is spent on waiting for the I/O accesses tocomplete. 5.14.3 Virtual memory aims to provide each application with theillusion of the entire address space of the machine. Virtual machines aims toprovide each operating system with the illusion of having the entire machine to itsdisposal. Thus they both serve very similar goals, and offer benefi ts such asincreased security. Virtual memory can allow for many applications running in the same memoryspace to not have to manage keeping their memory separate.

Related Questions

Navigate

• Browse (All)
• Browse O
• Subjects

---

---

Integrity-first tutoring: explanations and feedback only — we do not complete graded work. Learn more.