ROTTENSTREICHetal.:COMPRESSINGFORWARDINGTABLESFORDATACENTERSCALABILITY145
ByDe nition4oftheFIBencodingwidth,wecanseethattheencodingwidthofanencodingschemeCA=(¯σ1,σ2)canbepresentedas
(C A)=imax
∈[1,h]
(¯σ1(ai1))
+
(σ2(ai2))
=ymax
∈Sφ
(13)
2
σ¯1(y)+ (σ2(y)).
Weobservethatinordertocalculatetheoptimalencoding
σ2giventhatσ1:=¯σy∈S1,itisenoughtoknowφ (¯σσ¯1(y)forany
2whiletheexactvaluesof1(s1,1)),..., (¯σ1(srequired.Thisisbecause
1,n))arenotnecessarilyOPT(A|σ1:=σ¯1)=Cmin
(CA)
A=( σ1:=¯σ1,σ2)
=Cmin
A=(σ1:=¯σ1,σ2)
imax∈[1,h]
(¯
σ1(ai1))+ (σ2(ai))
2=
Cmin
max φ(y)+ (σ σ¯A=(σ1:=¯σ1,σ2)y∈S2
12(y)).
(14)Beforepresentingthealgorithm,wesuggestthefollowing
lemma.
Lemma5.Thereexistsa(pre x)encodingσ¯(σσ2ofS2suchthatthe1,σFIB2)=(¯encoding1,σ¯2)achievestheminimalencodingwidthofall
schemes{(σ1,σ2)|σ1:=¯σ
1}andsatis es( y∈S2),φσ¯1(y)+ (¯σ2(y))=OPT(A|σ1:=σ¯1).(15)Proof.LetC A=(σ1schemethatsatis es (C
:=σ¯1,σ 2)beaFIBencoding
A)=OPT(A|σ( y∈S1:=σ¯1).If2),φσ¯1(y)+ ( σ
2(y))=OPT(A|σσ¯ 1:=σ¯1)thende nition2=σ
2canbetherequestedencoding.Otherwise,bytheof (C A),( y∈Sσ¯2),φσ¯1(y)+ ( σ
2(y))≤ (C OPT(A|σA)=1:=1).ThisisbecausebyEquationCσ (13),for A=(¯1,σ 2)wehave (C A)=maxy∈S2φσ¯1(y)+
(σ
2(y)).Fory∈S2weusethenotationq(y)=
OPT(A|σ¯encodingσ¯1:=σ1) (φσ¯1(y)+ ( σ
2(y))).Wede nean2ofS2basedon σ
2asσ¯2(y)=σ 2(y)·0q(y)wherethe·signisusedtorepresenttheconcatenationoperation.In
particular,iffory∈S σ¯2,φσ¯1(y)+ (σ
2(y))=OPT(A|σ1)thenσ¯y∈S2(y)=σ 2(y)·0q(y)
=σ 1:=2(y).Wecanseethat( 2),φσ¯1(y)+ (¯σOPT(A|σ2(y))=φσ¯1(y)+ ( σ
2(y))+q(y)=1:=¯σ
1).Thusσ¯2satis estherequiredequality.Tocompletetheproofwehavetoshowalsothatσ¯apre xencoding.Wecan rstverifythatthen2isindeeddifferent.Ifforanyx,y∈S2=|Scodewordsofσ2|σ¯2(x)=σ
(x)2are·0q(x)=σ 0q(y)=σ¯2(s.t.x=y),
σ 22(y)·2(y)thennecessarily2(x)isapre xofσ (x)2(y)(ifq(x)≥q(y))orσ ofσ ≤q(y)),acontradictionsince2(y)isapre x2(x)(ifqσ pre xencoding.σ¯2isa)=σ 2alsopreserversthepre xpropertysinceifσ¯2(x2(x)·0q(x)isapre xofσ¯2(y)=σ 2(y)·0q(y)thenagaineitherσ 2(x)isapre xofσ 2(y)orσ 2(y)isapre xofσ2(x).
BasedonthelastlemmawecandeducefromKraft’s
inequality,theexactvalueofOPT(A|σ1:=¯σ
1).Theorem6.TheoptimalconditionalFIBencodingwidthof
Asatis es
OPT(A|σ1:=¯σ)= 1 ¯(y)
1 log2
2φσ (16)y∈S 2 .Proof.Weσ¯ rst
1)≥log wouldliketo showthatOPT(A|σ1:=2y∈Sφσ¯(y221
).ConsideraFIBencod-ing(σ1,σ2)=(¯σσ¯1,σ¯2)thathasanencodingwidthofOPT(A|σ (¯σ1:=1)and( y∈S2)satis esφσ¯1(y)+
2(y))=OPT(A|σ¯1).Thelastequalitycanbepresentedalsoas (¯σ1:=σ
2(y))=OPT(A|σ1:=σ¯1) φσ¯1(y).SuchanencodingexistsbyLemma5.ByKraft’sinequalityσ¯2satis es
1≥
2 (¯σ2(y))
= 2 OPT(A|σ1:=¯σ1) φσ¯1(y)y∈S2
y∈S2
2
OPT(A|σ1:=¯σ1)
≥
2φ¯σ
1(y)y
∈S2
OPT(A|σ1:=σ¯1)≥
φσ¯1(y)
log2
2
(17)
y∈S 2
,
wherethelastinequalityisobtainedsinceOPT(A|σσ¯)isofcourseaninteger.Tocompletetheproof1:=1wewould
log also liketoshowthatOPT(A|σ1:=σ¯1)≤2φ(y) σ¯1by ndingaFIB(σy∈S22
encodingscheme1,σ2)=(¯σ1,σ¯2)whoseencodingwidthsatis esthisupper Todolog bound. 2
so,wedenotebyvσ¯1theterm
2y∈Sφσ¯21(y).Then,2 (vσ¯1 φσ¯1(y))= 2φσ¯1(y) vσ¯1= 2φ¯σ1(y)·2 v¯σ
1y∈S2
y∈S2
≤
2
φσ¯1(y)
·
y∈S2
2
φσ¯ 1
1(y)
=1.y∈S2y∈S2
(18)
Thusthereexistsapre xencodingσ¯y))=(v(y))fory∈2withcodewordlengths (¯σ2(σ¯1 φσ¯1SC2.TheFIBencodingschemeA=(¯σ1,σ¯2)satis es
(C=max y∈Sφ
A)2
σ¯1(y)+ (¯σ2(y))(19)=max y∈Sφy)+(v(y)))
2
σ¯1(σ¯1 φσ¯1=max y∈Sv
2
σ¯1=vσ¯1.
Then,OPT(A|σ1:=σ¯1)≤v
σ¯1=log
2
y∈Sφσ¯2
2
1(y)
.
Now,whenthecodewordlengthsoftheelementsofSfromLemma5andTheorem6,itiseasyto nd2are
knownsuchanencodingasexplainedearlierinSectionIII.
Basedonthesuggestedoptimalconditionalencodingfromabove,wenowsuggesttwodirectionsto ndanef cientencodingschemeforagiventwo-columnFIBtable.Ofcourse,thesedirectionsarealternativesfortheencodingofgeneralFIBtablespresentedinSectionIII-Bthatcanbeappliedalsofortwo-columntables.The rstdirectionmaybeconsidered
Compressing Video for TaskStream Currently, TaskStream allows users to upload videos up to 100 megabytes in size. Therefore, in order to upload a video...
PPMexe: PPM for Compressing Software Milenko Drini?? and Darko Kirovski? c ? Computer Science Department, UCLA, Los Angeles, USA ? Microsoft Research, ...
Face Fixer Compressing Polygon MeshesFace Fixer Compressing Polygon Meshes隐藏>> UNC Technical Report TR-00-04 Face Fixer: Compressing Polygon Meshes with ...
Compressing Hexahedral Volume Meshes Martin Isenburg? University of North Carolina at Chapel Hill Pierre Alliez? INRIA Sophia-Antipolis Abstract Unstructured ...
Compressing Mixed-State Sources by Sending Classical Information_专业资料。We consider visible compression for discrete memoryless sources of mixed quantum ...
Managing Gigabytes Compressing and Indexing Documents and Images_专业资料。As today’s information explosion generates greater and greater volumes of raw data...
Compressing File Collections with a TSP-Based Approach Dimitre Trendafilov Nasir Memon Torsten Suel Department of Computer and Information Science Technical ...
Compressing cube-connected cycles and butterfly networks We consider the simulation of large cube-connected cycles (CCC) and large butterfly networks (BFN)...
The objectives for compressing DEMs are obtain...更多>>Abstract. Wavelet decomposition is a well-known technique to compress image data. Here, we have ...
Re-Store: A System for Compressing, Browsing, and Searching Large Documents (Extended Abstract of Invited Presentation) Alistair Moffat Raymond Wan ...
热门文档