$[10\%]$ "Tail latency" is the small percentage of response times from a system, out of all of responses to the input$/$output

$($I$/$O$)$ requests it serves, that take the longest in comparison to the bulk of its response times. They are, quite literally,

the tail cnd of a system's response time spectrum, and are often expressed as the $98$th or $99$th percentile response times.

A long$-$tail latency issuc is observed in the distribution of the write latency because of the management design of

storage devices $($e$.$g$.,$ HDDs and SSDs$).$ For instance, the latency at the $99$th percentile can be $100$x higher than the

average latency. Such a long$-$tail latency causes a significant problem in real$-$time embedded and enterprise$-$server

systems, which need to meet the real$-$time and quality of service $($QoS$)$ requirements. Please answer the following

questions:

$(1)[5\%]$ Please provide four possible causes of tail latency in the storage system design. Brief explanation is needed.

$(2)[5\%]$ Please provide two solutions for resolving the long$-$tail latency issue. A brief explanation is needed.

$[10\%]$ Consider a computer system with a $32-$bit virtual address space where paging is used. Assuming the page size

is $4K$ bytcs and the memory is bytc$-$addressable, please answer the following questions:

$(1)[2\%]$ How many pages can a process have at most? Suppose the maximum physical memory size is $32GB.$ What

is the maximum number of bits for physical addresses? What is the maximum number of frames for the system?

$(2)[2\%]$ Assume one$-$level paging is adopted. Let the memory access time and TLB access time be $100ns$ and $20ns,$

respectively. If we want an effective memory access time of less than $140ns,$ what is the minimal TLB hit ratio

that necds to be achicved?

$(3)[3\%]$ Suppose the virtual memory of the system adopts demand paging. Assume the effective memory access time

of the computer system without any page fault is $100ns,$ and the service time for a page fault is $15ms.$ If the page

fault rate is $0\mathrm{.}0000004,$ what is the effective access time under demand paging?