Presentation on theme: "ARC: A self-tuning, low overhead Replacement Cache"— Presentation transcript:
1ARC: A self-tuning, low overhead Replacement Cache Nimrod Megiddo andDharmendra S. ModhaPresented by Gim Tcaesvk
2Cache Cache is expensive. The replacement policy is the only algorithm of interest.
3Cache management problem Maximize the hit rate.Minimize the overhead.ComputationSpace
4Belady’s MIN Replaces the furthest referenced page. Optimum for every case.Provides upper bound of hit ratio.But,Who knows the future?
5LRU (Least Recently Used) Replaces the least recently used page.Optimum policy for SDD (Stack Depth Distribution).Captures recency but not frequency.232132232231123232113
6LFU (Least Frequently Used) Replaces the least frequently used page.Optimum policy for IRM (Independent Reference Model).Captures frequency but not recency.11123321
7Drawbacks of LFU Logarithmic complexity in cache size High frequently used page is hardly to paged out.Even if it is no longer useful.
8LRU-2 Replaces the least 2nd recently used page. Optimum online policy which knows at most 2 most recent reference for IRM.
9Limitations of LRU-2 Logarithmic complexity Tunable parameter, Due to priority queueTunable parameter,CIP (Correlated Information Period)2(2,4)1(1,3)0(-∞,0)
102Q (modified LRU-2)Uses simple LRU list instead of the priority queue.Two parameter, Kin and KoutKin=Correlated Information PeriodKout=Retained Information Period
11LIRS (Low Inter-reference Recency Set) Maintains 2 LRU stacks of different size.Llirs maintains LRU page at least twice recently seen.Lhirs maintains LRU page only once recently seen.Works well for IRM, but not for SDD.
12FBR (Frequency-Base Replacement) Divide LRU list into 3 sections;NewReference count in new section is not incremented.MiddleOldReplaces page in old section with smallest reference count.
13LRFU (Least Recently/Frequently Used) Exponential smoothing.λ is hard to tune.
14MQ (Multi-Queue) Uses m LRU queues Q0, Q1, … Qm-1 Qi contains page that have been seen [2i,2i+1) times recently.