ѧ
JOURNAL OF SOFTWARE
2000Vol.11No.2P.251-255



һֱܷģͺʵ
ľ
ժҪΪĿɿԼͨĹˮƽ,һܼŶȷģ.ģܹʵʱؼԤĸܲ,ʱȣĿɿԣperformance-reliabilityԣperformance-availability͸ƽ޹ʱ䣨mean time to failure,MTTF.ڴģʮָ,ڸ߿ɿĹ,˸ģ͵Ľģͼʵ.
ؼʣ,ܿɿ,ܿ,ƽ޹ʱ,ܹ.
ţTP393
A Performance Analysis Model and Its Implementation for Network Management
LI Mu-jinWANG Guang-xing
(Research Center for Network and Communications Northeastern University,Shenyang,110006)
AbstractTo improve network reliability and management level in today's high-speed telecommunication networks,a performance and dependability analysis model used for network management is proposed in this paper,which computes in real-time and forecasts performance-reliability,performance-availability for the overall network performance parameter (e.g.bandwidth,throughput and latency) and the MTTF (mean time to failure).Due to the arising complexity of exact model solution,for highly reliable network's management an approximation of the model and its implementation are also presented.
Key wordsActive network management,performance-reliability,performance-availability,MTTF (mean time to failure),performance management.
ڵ,ͨԽԽҪ,ĿԺͿɿ岻ݴǵ,,ĹߺͷܹӦǰͽͨŻҪ.Ŀǰ,еģͣ1,2һ.Ͻܵ½,,ѰһµķģԤ.ĲӦģͣstochastic reward model,SRM3,ݹĲܼͬ䶯̬ʵȶܲĿɿԣperformance-reliabilityԣperformance-availability͸ƽ޹ʱ䣨mean time to failure,MTTFȽзԤ,ʹԱܹϻʱȡҪĴʩֹϵķ,֤ṩķQoS,һֹΪ.ڴ˷ʹøӻ,,ڸ߿ɿĹ,Ǹ˸ģ͵Ľģͼʵ.
1ܼŶȷģ
1.1 
һɲͬԪأ,·ͽڵ㣩ɵĿ޸.ÿԪؾ״̬״̬͹״̬.һϷ֮,޸滻Ϳʼ.ÿԪ޸֮ͻָ״̬.ͬĹ϶ΪͬӰ.Щϵܵ½,ȻԼ,ṩһԽܵķ;Щ֮ΪϡĹϽṩһɽܵķ,,һϡΪṩһɽܵķ״̬תƵṩһɽܵķ״̬.һϡ·ڽڵ϶,ӰΪ,սṩһɽܵķʱʵӣ.
Щ,SRMģͨ4.Ǿ״̬ռ䦸ʱ{Z(t),t0},ǳƸùΪṹ״̬.Ӧṹfr:,ʵriǵi״̬ĻӦ,ڵi״̬.״̬ռ䦸ɷֳཻӼU,Uʾ״̬,ʾ״̬.|U|ʾU״̬.P(t)ʾгʼֲP(0)ת״̬µĸ,Pi(t)ʾtʱ̵i״̬ʱĸ.ͬ,бʾƽ״̬ĸ,iʾڸƽ״̬µi״̬ĸ.ʧһ,t=0ʱΪ״̬,򦸵ĵ1Ԫǹ״̬.P(u)(t)Ϊ繤״̬|U|ά.ԪP(u)i(t)(iU)ʾڣ0,tʱûϷʱ,tʱ̵i״̬ʱĸ.
1.2һ㹫ʽ
ܿɿPR(t)Ϊڣ0,tʱûϷʱ,tʱ̵˲ʱӦѧ,޹ڼ,
(1)
޹ƽMPTFΪϷ֮ǰۻӦѧ֮ͣ
(2)
tʱ̵˲ʱӦѧ,ܿPA(t),
(3)
1.3ģ 
ǿǽṹ״̬{Z(t),t0}ǰϣsem-Markov̵4,
Qij(x)=Pr{Z(T1)=j,T1x|Y0=i}
תƸ,ҾQ(t)=Qij(t)ݱΪϹ̵ĺ.pij=Pr{Z(T1)=j|Y0=i}pij=Qij(),ڵi״̬ʱƽֵΪ,Qi(x)=Pr{T1x|.
ף4ݽһõڵiiUʼ״̬ܿɿԺƽ޹ܵĹʽ
(4)
(5)
ԼܿԵļ㹫ʽ
(6)
,ǽṹ״̬{Z(t),t0}һСQ=qijݵһʱCTMCcontinuous time homogeneous Markov chain,qijijʾӵi״̬תƵj״̬תƸʺǵiԽԪ.Qi(x)=1-e-qix,ʽ46Ϳɵõǵֵ.
Ӧֻ,Ϊ״̬ʱ,ӦΪ1Ϊ״̬ʱ,ӦΪ0,ʽͿԵõƽ޹ʱMTTFļ㹫ʽ.
nڵ·ʱ,2n״̬.̵Ľ,ҪתΪ2nһ΢ֺͻַĽ.nϴʱ,Ϊ.
2߿ɿܼŶȷģ
Ϊ˼,Ǽڵ·޸/滻ƽʱԶСMTTF,ڱ,ǳΪ߿ɿ,ǺϵԪعϵ״̬,Ҹģһ۵ľȷģ֮С5.
һ޸,䲻ͬĽڵ·ǵϿܷ.i¡ϡĹ,ܼ,п½Լj¡ϡĹ.ǵĹΪk(k=1,2,...,i+j),ck(k=1,2,...,i)Ϊk¡ϡĹϵĿɻָ,ָʱԲ.һɻָĹϷ֮,Ϳʼù.kɻָĹϵʱκһʷֲ,ʷֲΪGck(t),Gck(0)=0,ĵkɻָĹƽʱMTTCkmean time to reconfigurationΪMTTCk=ҡ0(1-Gck(x))dx.ͬ,һɻָĹϷ֮,Ϳʼ޸/滻.k(k=1,2,...,i+j)ϵ޸/滻ʱκһʷֲ,ֲΪGRk(t),GRk(0)=0,ĵkɻָĹϵƽ޸ʱMTTRkmean time to repairΪMTTRk=ҡ0(1-GRk(x))dx.
Щ,SRMĽṹ״̷̬ӰϻӦ̣0״̬ʾеĽڵ·޹,ĻӦΪr0.xk(k=1,2,...,i)״̬ʾΪrkĿɻָĹϵķ.kɻָYk¡ϡ,ӦΪ0.ڼ,ڵxk(k=1,2,...,i)״̬.ýʱ,ڵ0״̬.xk(k=i+1,i+2,...,i+j)״̬ʾ.
,Q(t)=Qij(t)ݵķʽйƽȹ̵4ƪ޶ʡԣ
(7)
ʽ4,5ͣ6,õ㹫ʽ8,9ͣ10
8

9
(10)

PASxk(t)=ri(1-GRk(t))+t0PAS0(t-x)dQxk0(x),k=1,2,...,i,
PASxk(t)=t0PAS0(t-x)dQxk0(x),k=i+1,2,...,i+j,
PASyk(t)=t0PASxk(t-x)dQxkyk(x),k=1,2,...,i.
3ֵ
һֻۼ㹫ʽ8ͣ10ֵⷨ.
һڵļ㹫ʽ8ͣ10ɱʾһʽķ飺
A(t)-t0W(x)A(t-x)dx=g(t),11
A(t),W(t)g(t)ֱΪ.
ǵA(t)ɢΪA(nh),Էʽ11дɣ
12
hΪ,enΪ,h2Ľ.,h㹻Сʱ,enɺԲ,ǵõ
(13)
,A(nh)ʾnhɢĽƽ,淽ұߵļԴ˵,ʵʱҪ,ҪѡQn,i.Trapezoidal6,Qn,0=Qn,n=1/2,Qn,i=1,i=1,2,...,n-1,ʽ13ʽ
14
Iʾm+1׵λ.I-(h/2)W(0)ݡ0ʱ,ܹؼ{A(h),A(2h),...}.EkiʾʹTrapezoidalki̵Ľƽ߽,h2Ľ.ͨף6ݵĽ,õ

ͼ1ʾԴST֮ͨ.ST֮ͨҪ2,ָṩST֮ҪĴ.ST֮ڷƽ״̬¸ʱPRS0(t)PAS0(t)ֵ1ͱ2.PAS0(t)ֵʵϾǴST֮tʱõĴ.

Fig.1
ͼ1
Table 1Performance reliability
1ܿɿ

n123456789
PRS09.99829.99659.99489.99319.99149.98979.98819.98659.9850

(h=0.2,i=810-5,ci=1,i=1,...,6,=5)
Table 2Performance availability
2ܿ

n123456789
PAS09.99839.99679.99529.99389.99269.99149.99059.98969.9889

(h=0.2,i=810-5,ck=1,k=1,...,14,i=1,...,14,=5)
ڵNii=0,1,...,5·Lii=1,2,7,8ĴΪ2,·Lii=3,4,5,6ĴΪ.ڵNi·LiĹΪii=1,...,14,.ϵͳܳɹؼ⵽,ҳֹϺ޸,·ɹ,໥.ݸ߿ɿܼŶȷģͺ֪
(1) L1,L2,L3,L6,L7L86Ӱ,,ǵri=2,ҿԻָ,ci=1(i=1,2,3,4,5,6.
(2) N0,N1,N2,N3,N4,N5,L4L58ϵ,ri=0(i=7,...,14).
GRk(t)ӱ׼ֲ̬,,ʽ8ɵã
15

PRSxk(t)=ri(1-GRk(t))+t0f(x)PRS0(t-x)dx,k=1,2,3,4,5,6,
ͬ
ơ16

PASxk(t)=ri(1-GRk(t))+t0f(x)PAS0(t-x)dx,k=1,2,3,4,5,6,
PASxk(t)=t0f(x)PAS0(t-x)dx,k=7,8,...,14.
4ᡡ
۵ģѾõɹ863߿ƼĿĿ⡰ҵIntranetоĿ,ܹʵʱʾԤ磨Դڵ㵽սڵĴʱĿɿԡԺ͸磨MTTF,ȡЧ.
Ŀоõ863ƼĿ(No.863-511-711-003).
߼飺ľ,1965,,ʿ,ʦ,ҪоΪ,ATM,ýͨ.
,1939,,ʿʦ,ҪоΪ,ATM,ýͨ.
ߵλľ(ѧͨġ110006)
(ѧͨġ110006)
οף
1Rouvellou I.Graph identification techniques applied to network management problems Ph.D.Thesis.Colombia University,1993
2Wolfson O,Sengupta S,Yemini Y.Managing communication networks by monitoring databases.IEEE Transactions on Software Engineering,1991,17(9):944953
3Bobbio A.Stochastic reward models in performance reliability analysis.Journal on Communications,1992,XLIII(1):2735
4Brenner A.Stochastic reward models for performance and dependability analysis of fault tolerant systems.Technical Report,DS3,ETH Zrich:Reliability Laboratory,1995
5Ciardo G,Marie R,Sericola B et al.Performance analysis using semi-Markov reward processes.IEEE Transactions on Computers,1990,C-39(10):12511264
6Jones J G.On the numerical solution of convolution integral equations and systems of equations.Mathematics Computation,1961,15(1):131142
ոڣ1998-10-09
޸ڣ1999-03-10
