Consider a two-level memory hierarchy, M1 and M2 with access time t1 and t2 , costs per byte c1 and c2 , and capacities s1 and s2 , respectively. The cache hit ratio h1 ratio h1 = 0.95 at the first level. (Note that t2 is the access time between the CPU and M2 , not between M1 and M2 ).

a) Calculate the effective access time teff & total cost of this memory system if

t1 = 7ns and t2 = 48ns , costs per byte c1 = 0.7 $/Kbyte and c2 = 0.4 $/Kbyte , and capacities s1

= 256 KB and s2 = 512 KB

b) Suppose t1 = 20ns and t2 = unknown , s1 = 512 KB and s2 = unknown, c1 = 0.01 $/byte and c2 =

0.0005 $/byte.

How large a capacity of M2 (s2 = ?) can you acquire without exceeding budget limit (15000 $)

How fast a main memory (t2 = ?) do you need to achieve an effective access time of teff =

40ns in the entire memory system under the above hot ratio assumptions?

Consider a two-level memory hierarchy, M and M2 with access time t, and t2, costs per byte ci and c2, and capacities si and 2, respectively. The cache hit ratio hi ratio hi=0.95 at the first level. (Note that tz is the access time between the CPU and M2, not between M and M). a) Calculate the effective access time ter & total cost of this memory system if t1 = 7ns and t2 = 48ns, costs per byte ci = 0.7 S/Kbyte and c2 = 0.4 S/Kbyte, and capacities si = 256 KB and s2 = 512 KB b) Suppose t; = 20ns and t2 = unknown, si = 512 KB and s2 = unknown, c = 0.01 S/byte and C2 = 0.0005 $/byte. How large a capacity of M2 (82-?) can you acquire without exceeding budget limit (15000 How fast a main memory (t2 = ?) do you need to achieve an effective access time of te Ons in the entire memory system under the above hot ratio assumptions