Lecture 6.

Online Algorithms

Introduction to Algorithms
Da Nang University of Science and Technology

Dang Thien Binh

[email protected]

Intelligent Networking Laboratory

Copyright 2000-2022 Intelligent
DANG
Networking
THIEN BINH
Laboratory
1/23
 Contents

 Introduction
 Online algorithms analysis
 Competitive analysis
 Probabilistic analysis
 Motivating examples:
 Ski rental problem
 2-lane line search problem
 Paging problem
 Bélády's anomaly

Intelligent Networking Laboratory DANG THIEN BINH 2/23

 Introduction

 An online algorithm is one that can process its input

piece-by-piece, without having the entire input
available from the beginning
 For example, insertion sort is an online algorithm because:
 It considers one input element per iteration
 It produces a partial solution without considering future
elements

 In contrast, an offline algorithm is given the whole

problem data from the beginning
 For example, selection sort is an offline algorithm because:
 It requires access to the entire input to sort an array by
repeatedly finding the minimum element

Intelligent Networking Laboratory DANG THIEN BINH 3/23

 Online algorithms analysis

 Competitive analysis
 An online algorithm is c-competitive if there is a constant
b for all sequences s of operations
CON(s) ≤ c COPT(s) + b
Where CON(s) is the cost of online algorithm on the sequence s
and COPT(s) is the optimal offline cost for the same sequence
 Competitive ratio: The ratio between the output
generated by an online algorithm and the output produced
by an optimal offline algorithm
 Probabilistic analysis
 Assume a distribution generating the input then find an
algorithm which minimizes the expected cost of the online
algorithm
Intelligent Networking Laboratory DANG THIEN BINH 4/23
 Ski rental problem (1/4)

 Problem description: We will go skiing in

Pyeongchang for the first time. We have two
options for the equipment: (1) Buying it costs $500;
(2) Renting it for a weekend costs $50. Should we
buy or rent it?
 Our goal is to minimize the cost for equipment
 Observations: The cost depends on how many times
we will go skiing. Specifically,
 If we will go skiing 11 times or more, then it is better to
buy the equipment
 If we will go skiing 9 times or fewer, then it is better to
rent the equipment

Intelligent Networking Laboratory DANG THIEN BINH 5/23

 Ski rental problem (2/4)

 An online algorithm called better-late-than-never:

 rent until you realize you should have bought, then buy. (In our case: rent
9 times, then buy).
 will be a number k such that after renting k-1 times, we will buy the
equipment (just before your kth time)
 Formally, if the rental cost is r and the purchase cost is p then the
algorithm is to rent ⌈p/r⌉ − 1 times and then buy
 Competitive analysis:
 For k < 10, we should rent -> the cost for optimal offline algorithm is 50k,
so the competitive ratio is
50𝑘
=1
50𝑘
 If k ≥ 10, we should buy -> the cost for optimal offline algorithm is 500, so
the competitive ratio is
𝑝
𝑟 −1 +𝑝 𝑟
𝑟 = 2− ≤2
𝑝 𝑝

Intelligent Networking Laboratory DANG THIEN BINH 6/23

 Ski rental problem (4/4)

 Conclusion: The algorithm better-late-than-never

has competitive ratio ≤ 2. If the purchase cost p is a
multiple of rental cost r, then the competitive ratio
is 2 - r/p
 So better-late-than-never algorithm is (2-r/p)-competitive

Intelligent Networking Laboratory DANG THIEN BINH 7/23

 2-lane line search problem (1/5)

 Problem description: A robot starts at the origin of

the x-axis. An object has been placed somewhere
on the x-axis. The robot can switch direction of
travel immediately, but in order for the robot to
determine that there is an object at location x’,
the robot has to be physically present at x’. How
should the robot explore the x-axis in order to find
the object as soon as possible?
Left or Right?
How should I go?

Left Right
x’ O x

Intelligent Networking Laboratory DANG THIEN BINH 8/23

 2-lane line search problem (2/5)

 We investigate an online algorithm called doubling

strategy to find the object
Input:
1) Current direction:= Left
2) Current position:= 0
3) d:= 1
Repeat:
1) Travel d unit on the current direction
2) If found the object then finish
3) Else return to starting point
4) d:= 2d
5) Flip the current direction
Intelligent Networking Laboratory DANG THIEN BINH 9/23
 2-lane line search problem (3/5)

 The robot follows an online algorithm named

doubling strategy to find the object

x’ Flip direction O Start Flip direction x

Found object 1
1 2
4 2
4 8
𝜀 4 8

Figure 3. The worst case when the robot finds the object

Intelligent Networking Laboratory DANG THIEN BINH 10/23

 2-lane line search problem (4/5)

 We derive the competitive ratio of the doubling

strategy
 If the object locates at position x′ = 2𝑗 + 𝜀.
 We have the total distance of this algorithm:
𝐶𝑜𝑠𝑡𝑂𝑁 = 2 1 + 2 + 4 + … + 2𝑗+1 + 2𝑗 + 𝜀

1 ∗ 1 − 2𝑗+2
= 2∗ + 2𝑗 + 𝜀
1−2
= 2 ∗ (2𝑗+2 − 1) + 2𝑗 + 𝜀
= (23 +1) ∗ 2𝑗 + 𝜀 − 2
≈ 9 ∗ 2𝑗 + 𝜀
where j is the number of iterations and 𝜀 is a small distance
Intelligent Networking Laboratory DANG THIEN BINH 11/23
 2-lane line search problem (5/5)

 The cost of optimal offline algorithm when the

robot knows where the object locates at x′ = 2𝑗 + 𝜀
𝐶𝑜𝑠𝑡𝑂𝑃𝑇 = 𝑥′ = 2𝑗 + 𝜀
 Finally, combine 𝐶𝑜𝑠𝑡𝑂𝑁 and 𝐶𝑜𝑠𝑡𝑂𝑃𝑇, we obtain the
competitive ratio as following
𝐶𝑜𝑠𝑡𝑂𝑁 9∗2𝑗 +𝜀
Competitive ratio= = ≈9
𝐶𝑜𝑠𝑡𝑂𝑃𝑇 2𝑗 +𝜀

 Conclusion: The doubling strategy is 9-competitive

for the 2-lane line search problem

Intelligent Networking Laboratory DANG THIEN BINH 12/23

 Paging problem (1/5)

 Problem description: We have a disk with N pages, and

cache with space k < N. When a memory request is
made, if the page isn’t in the cache -> page fault. We
then need to bring the page into the cache and throw
something else out if our cache is full. Our goal is to
minimize the number of page faults. The algorithmic
question is: which page should we throw out?
 There are some online algorithms for this problem:
 Least-Recently-Used (LRU) (We will investigate this
algorithm)
 First-In-First-Out (FIFO)
 Flush-When-Full (FWF)

Intelligent Networking Laboratory DANG THIEN BINH 13/23

 Paging problem (2/5)

 Least-Recently-Used (LRU) algorithm evicts the

page that has been unused for the longest time

 Competitive ratio is used to analyze LRU algorithm

# 𝑜𝑓 𝑒𝑣𝑖𝑐𝑡𝑖𝑜𝑛𝑠 𝑜𝑓 𝑜𝑛𝑙𝑖𝑛𝑒 𝑎𝑙𝑔𝑜𝑟𝑖𝑡ℎ𝑚(σ)
Competitive ratio=
# 𝑜𝑓 𝑒𝑣𝑖𝑐𝑡𝑖𝑜𝑛𝑠 𝑜𝑓 𝑜𝑝𝑡𝑖𝑚𝑎𝑙 𝑜𝑓𝑓𝑙𝑖𝑛𝑒 𝑎𝑙𝑔𝑜𝑟𝑖𝑡ℎ𝑚(σ)

where σ is the input sequence of requests

Intelligent Networking Laboratory DANG THIEN BINH 14/23

 Paging problem (3/5)

 We will derive the competitive ratio of LRU algorithm

 Consider the cache has k slots, which are initially
empty and the input sequence
σ = 1, 2, 3, …, k, k+1, 1, 2, 3, …, k
 After arriving of first k inputs, cache looks like

 When k+1 arrives, 1 will be evicted because 1 is unused for

the longest time, cache then looks like

 Continue…

Intelligent Networking Laboratory DANG THIEN BINH 15/23

 Paging problem (4/5)

 When 1 arrives again, 2 will be evicted, cache looks like

 Similarly, all the k pages will be evicted one by one

 And, when the last input k arrives, k+1 will be evicted

Intelligent Networking Laboratory DANG THIEN BINH 16/23

 Paging problem (5/5)

 So, the total number of evictions using LRU is k+1

 Now, if we use the optimal algorithm for the same
input sequence σ, the cache will look like

 Hence, the number of evictions for optimal algorithm is 2

𝑘+1
Competitive ratio =
2

Intelligent Networking Laboratory DANG THIEN BINH 17/23

 Bélády's anomaly (1/5)
 Bélády's anomaly is a phenomenon where increasing the
size of cache memory results in an increase of page
faults
 Bélády's anomaly is commonly experienced with FIFO
algorithm but never occurs in LRU algorithm
 Reason behind Bélády's anomaly:
 LRU is a stack-based algorithm -> Do not suffer Bélády's
anomaly
 Stack property: the set of pages that were presented in cache
memory when the size of cache memory is k will be
compulsorily presented in cache memory if the size of cache
memory increases to k’>k

Intelligent Networking Laboratory DANG THIEN BINH 18/23

 Bélády's anomaly (2/5)
 FIFO algorithm example:
 Given the request: 1, 2, 3, 4, 1, 2, 5, 1, 2, 3, 4, 5
 Case 1: If k = 3, the given request using FIFO
algorithm yields a total of 9 page faults (PF)
Stage 1 Stage 2 Stage 3 Stage 4 Stage 5 Stage 6 Stage 7 Stage 8 Stage 9 Stage 10 Stage 11 Stage 12
1 1 1 2 3 4 1 1 1 2 5 5
2 2 3 4 1 2 2 2 5 3 3
3 4 1 2 5 5 5 3 4 4
PF PF PF PF PF PF PF O O PF PF O

 Case 2: If k = 4, the given request using FIFO

algorithm yields a total of 10 page faults (PF)
Stage 1 Stage 2 Stage 3 Stage 4 Stage 5 Stage 6 Stage 7 Stage 8 Stage 9 Stage 10 Stage 11 Stage 12
1 1 1 1 1 1 2 3 4 5 1 2
2 2 2 2 2 3 4 5 1 2 3
3 3 3 3 4 5 1 2 3 4
4 4 4 5 1 2 3 4 5
PF PF PF PF O O PF PF PF PF PF PF

Intelligent Networking Laboratory DANG THIEN BINH 19/23

 Bélády's anomaly (3/5)
 At stage 7 and stage 8 in Case 2, cache memory
does not contain the set of pages that are
presented in the corresponding stages in Case 1
 Thus, FIFO algorithm does not follow the stack
property
 FIFO algorithm suffers Bélády's anomaly
 The number of page faults increases from 9 to 10 when
increasing the size of cache memory from 3 to 4
 Note: It would be wrong to say that “FIFO algorithm
always suffers from Bélády's anomaly”

Intelligent Networking Laboratory DANG THIEN BINH 20/23

 Bélády's anomaly (4/5)
 LRU algorithm example:
 Given the request: 1, 2, 3, 4, 1, 2, 5, 1, 2, 3, 4, 5
 If k = 3, the given request using LRU algorithm
yields a total of 10 page faults (PF)
Stage 1 Stage 2 Stage 3 Stage 4 Stage 5 Stage 6 Stage 7 Stage 8 Stage 9 Stage 10 Stage 11 Stage 12
1 2 3 4 1 2 5 1 2 3 4 5
1 2 3 4 1 2 5 1 2 3 4
1 2 3 4 1 2 5 1 2 3
PF PF PF PF PF PF PF O O PF PF PF

 If k = 4, the given request using LRU algorithm

yields a total of 8 page faults (PF)
Stage 1 Stage 2 Stage 3 Stage 4 Stage 5 Stage 6 Stage 7 Stage 8 Stage 9 Stage 10 Stage 11 Stage 12
1 2 3 4 1 2 5 1 2 3 4 5
1 2 3 4 1 2 5 1 2 3 4
1 2 3 4 1 2 5 1 2 3
1 2 3 4 4 4 5 1 2
PF PF PF PF O O PF O O PF PF PF

Intelligent Networking Laboratory DANG THIEN BINH 21/23

 Bélády's anomaly (5/5)
 At stages in Case 2, cache memory contains the set
of pages that are presented in the corresponding
stages in Case 1
 LRU algorithm follows the stack property
 LRU does not suffer Bélády's anomaly:
 The number of page faults decreases from 10 to 8 when
increasing the size of cache memory from 3 to 4

Intelligent Networking Laboratory DANG THIEN BINH 22/23

 Thanks to contributors
Mr. Phuoc-Nguyen Bui (2022)
Dr. Thien-Binh Dang (2017 – 2022)
Prof. Hyunseung Choo (2001 – 2022)

Intelligent Networking Laboratory

Copyright 2000-2022 Intelligent
DANGNetworking
THIEN BINH
Laboratory
23/23