SALMonADA: A Pla o m o Moni o ing and Explaining Viola ions o
WS-Ag eemen -Complian Documen s
C. Mülle (1), M. O iol(2), M. Rod íguez(2), X. F anch(2), J. Ma co(2), M. Resinas(1), A. Ruiz–Co és(1)
(1)Uni e si y o Se ille, LSI, Se ille (Spain), ISA esea ch g oup
{cmulle , esinas,a uiz}@us.es
(2)Uni e si a Poli ècnica de Ca alunya, Ba celona (Spain), GESSI esea ch g oup
{mo iol,jma co}@lsi.upc.edu, {ma c , anch}@essi.upc.edu
Abs ac —Quali y assu ance echniques ha e been de eloped
o supe ise he se ice quali y (QoS) ag eed be ween se ice-
based sys ems (SBSs) consume s and p o ide s. Such QoS is
usually included in se ice le el ag eemen s (SLAs) and hus,
SLA moni o ing pla o ms ha e been de eloped suppo ing
iola ion de ec ion. Howe e , jus a ew o hem p o ide expla-
na ion o he iola ions caused by obse ed QoS a moni o ing
ime, bu no in an use - iendly o ma . The e o e, we p opose
a gene al moni o ing and analysis concep ual e e ence model
and we ins an ia ed i wi h SALMonADA, a SBS ha no i ies
he clien s wi h iola ions and hei causes in hei own easy-
o-unde s and speci ica ion e ms. In addi ion, ou pla o m
pe o ms an ea ly analysis no i ica ion ha a oids delays in he
clien no i ica ion ime when a iola ion akes place. Mo eo e ,
we ha e implemen ed a web applica ion as a SALMonADA
clien , o p o e how i moni o s, analyses and epo s o hei
clien s he se ice le el ul illmen o eal se ices subjec o a
SLA speci ied wi h WS–Ag eemen .
Keywo ds-moni o ing; analysis; iola ion de ec ion; iola ion
explana ion;
I. INTRODUCTION AND MOTIVATION
Se ice le el ag eemen s (SLAs) es ablish he se ice
quali y (QoS) ag eed be ween se ice-based sys ems (SBSs)
consume s and p o ide s, and hus, quali y assu ance ech-
niques a e needed o supe ise he SLAs ul illmen . These
echniques equi e moni o ing pla o ms enhanced wi h anal-
ysis capabili ies o eason abou he moni o ed in o ma ion
in o de o ex ac use ul in o ma ion o he pa ies.
Many esea ch e o s ha e been made ying o ob-
ain use ul moni o ing in o ma ion, s a ing om gene al
moni o ing amewo k [1], [2]. Thus, se e al p oposals can
be ound p o iding a di e en kind o in o ma ion om
moni o ed SLAs, such as: iola ion de ec ion in he SOA
es ing con ex [3], [4]; asynch onous iola ion de ec ion
epo s o subsc ibed clien s ha wai o he moni o ing
in o ma ion ins ead o eques ing i [5], [6]; e en -based
iola ion explana ion in SBSs [7], [8]; and dynamic SBS
adap a ion when a SLA iola ion is de ec ed.
In his pape we p opose a gene al moni o ing and anal-
ysis concep ual e e ence model in which se e al agen s
ex ac use ul in o ma ion om SLAs a moni o ing. Fo
ha pu pose, h ee kinds o documen s a e handled by
agen s, namely: SLAs, moni o ing managemen documen s
(MMDs) o con igu e and manage he moni o s, and se ice
le el ul illmen (SLF) o epo he iola ions and hei
causes. In addi ion, we ins an ia e he concep ual model wi h
SALMonADA, a se ice-based sys em (SBS) ha in eg a es
upg aded e sions o p e iously de eloped SLA moni o -
ing (SALMon [9]) and analysis (ADA [10]) p oposals1.
SALMonADA p o ides he ollowing con ibu ions o such
echniques wi h he aim o ex ac ing use ul in o ma ion a
SLAs moni o ing: i s , i no i ies he clien wi h iola ions
and hei causes in hei own easy- o-unde s and speci i-
ca ion e ms; second, i suppo s exp essi e and easy- o-
unde s and SLAs speci ied wi h WS–Ag eemen [11]; and
inally i pe o ms an ea ly analysis no i ica ion ha suppo s
he SLA ul illmen analysis when a iola ion has jus been
obse ed, educing he clien no i ica ion ime.
P oposals like [12] and [13] assuming he a ailabili y
o a moni o ing and analysis engine, bene i om using
SALMonADA since hey a e p o ided wi h such a se ice
le el ul illmen in o ma ion needed o adap he SBS, ene-
go ia e he SLA, achie e epu a ion s a is ics, e c.
The pape is o ganised as ollows. Rela ed wo k is e ised
in Sec ion II. WS–Ag eemen speci ica ion is in oduced in
Sec ion III including an example o ou suppo ed SLAs.
The concep ual e e ence model is de ailed in IV, while i s
SALMonADA ins an ia ion is included in V. Sec ion VI
and VII de ail he moni o ing and analysis SALMonADA
componen s, espec i ely. Sec ion VIII epo s an e alua ion
o ou p oposal. And Sec ion IX concludes he pape wi h
a discussion o con ibu ions.
II. RELATED WORK
On he one hand, many moni o ing amewo ks ha e
been p oposed, [1], [2], [7] conside ing moni o ing and e en
analysis agen s, bu o he bes o ou knowledge, none o
hem p opose a sepa a ion o conce ns be ween SLAs, he
moni o ing in o ma ion, and he moni o ing analysis esul ;
as we do o SLAs, MMDs, and SLFs.
1Bo h p oposals ha e been widely e ised o suppo he no el ies o he
p oposed concep ual e e ence model (e.g. he managemen o SLAs and
SLF ha e been included in ADA; and he MMDs managemen in SALMon).
Table I
RESPONSETIME<100 IN GETRATE OPERATION USING EC [7]
EC Fo mula lines T u h Value
o all 1 : ime
exis s 2 : ime -
Happens(ic:ge Ra e(ID,coun y2,coun y1),
1,R( 1, 1))^ T ue
Happens(i :ge Ra e(ID), 2,R( 1, 2)) T ue
oc:sel :sub( 2, 1)<100 False
On he o he hand, se e al echniques o ex ac use ul
in o ma ion a SLA moni o ing ha e been de eloped. Thus,
we can ind p oposals p o iding iola ion de ec ion o WS–
Ag eemen documen s in he SOA es ing con ex [3], [4].
Howe e , such es ing p oposals moni o he se ice o de ec
iola ions a es ing and no while he se ice is consumed.
O he p oposals such as [5], [6], dealing he la e wi h non–
WS–Ag eemen documen s, p o ide asynch onous iola ion
de ec ion epo s o subsc ibed clien s ha wai o he
SLA moni o ing in o ma ion ins ead o eques ing i , as
commonly pe o med by o he p oposals. Mo eo e , he e
a e p oposals such as [14], [15], [16], [17] ha go u he and
when hey de ec a SLA iola ion hey dynamically adap
he SBSs ollowing di e en s a egies, bu his dynamic
eac ion is ou o he scope o he pape .
As a as we know, he e is only a se o p oposals om
Mahbub and Spanoudakis ha p o ides iola ion explana-
ion in SBSs [7], [8]. Such p oposals use e en calculus
(EC) and hey epo an e en -based explana ion o he
SBSs iola ion as ollows: "The ope a ion e en has
iola ed he EC o mula Fo he e m T". Fo ins ance,
Table I depic s a ou -lines EC o mula ha is epo ed as
explana ion o a clien in o ming abou he esponse ime
iola ion o he ge Ra e ope a ion. Such a o mula is
included inside WS–Ag eemen documen e ms2.
III. WS–AGREEMENT IN A NUTSHELL
The WS–Ag eemen ecommenda ion [11] desc ibes bo h
an XML–based language and a p o ocol ha acili a es he
publica ion, disco e y, and moni o ing o SLAs be ween
wo pa ies, usually a se ice p o ide and a se ice con-
sume . The SLAs a e c ea ed a e a nego ia ion p ocess
and hey comp ise an ag eemen iden i ie , an ag eemen
con ex con aining in o ma ion abou he in ol ed pa ies,
and ag eemen e ms ha desc ibe bo h he cha ac e is ics
o he se ices o be p o ided in se ice e ms and he
gua an ees on such se ices in gua an ee e ms. No e ha
WS–Ag eemen only de ines he gene al s uc u e o a SLA
and he kind o e ms i may include. Howe e , i does no
speci y any ocabula y o exp ess he ea u es o he se ice.
Se ice e ms a e di ided in o wo elemen s: i s , se ice
desc ip ion e ms ha de ine he ea u es o he se ice
ha will be deli e ed unde an ag eemen ; and second,
2This sample is included in [7] a page 26.
<Ag eemen Ag eemen Id="1.0"...>
<Name>SALMonADA-complian ADA SLA</Name>
<Con ex >
<Ini ia o >IneedSLAAnalysisCo p.</Ag eemen Ini ia o >
<Responde >ADA Tool (ISA G oup)</Ag eemen Responde >
<Se iceP o ide >Ag eemen Responde </Se iceP o ide >
<Expi a ionTime>2013-01-01T00:00:00</Expi a ionTime>
</Con ex >
<Te ms Name="ADASe ice">...
<Se iceP ope ies Name="SALMon-complian me ics"...>
<Va iable Name="A e ageResponseTime"
Me ic="me ics/Floa ">...</Va iable>
<Va iable Name="Gene alResponseTime"
Me ic="me ics/Floa ">...</Va iable>
<Va iable Name="A e ageA ailabili y"
Me ic="me ics/Pe cen age">...</Va iable>
</Se iceP ope ies>
<Se iceDesc ip ionTe m Name="ADA-SDT"
Se iceName="ADASe ice">
<WebSe iceIn o ma ion Name="ADASe ice-WSDL">
<desc ip ion>ADA is a SLA analyse </desc ip ion>
<wsdl>h p://www.isa.us.es:8081/ADASe ice?wsdl</..>
<endp>h p://www.isa.us.es:8081/ADASe ice?wsdl</..>
<ope a ion opName="checkDocumen Consis ency">...
... mo e ope a ions a e included ...
</WebSe iceIn o ma ion>
</Se iceDesc ip ionTe m>
<Gua an eeTe m Name="Gene alA ailabili y"...>
<SLO> A e ageA ailabili y >= 95 </SLO>
</Gua an eeTe m>
<Gua an eeTe m Name="gene alResponseTimeRela ions"...>
<Se iceScope Se iceName="ADASe ice">
checkDocumen Consis ency, xmlToWSAg4People,
wsag4PeopleToXML, ge Me icFile
</Se iceScope>
<SLO> A e ageResponseTime<=Gene alResponseTime </SLO>
</Gua an eeTe m> ... mo e gua an ees a e included ...
</Te ms>
</Ag eemen >
Figu e 1. Main elemen s o SALMonADA-complian ADA SLA
se ice p ope ies ha de ine named, se ice– ela ed se s o
a iables ha can be used o he speci ica ion o gua an-
ee e ms and mus be he e o e conside ed o ag eemen
moni o ing. All a iables mus include a domain–speci ic
me ic de ini ion o speci y he seman ics and ype o a
a iable. How he se ice desc ip ion e ms a e o ganized
and exp essed is le open by he ecommenda ion. Thus,
as depic ed in he documen o Fig. 1, we use ou own
domain speci ic language (DSL) o de ining he ADA
analysis se ice, including h ee p ope ies: he a e age be-
ween se e al esponse ime measu es o he same ope a ion
(A e ageResponseTime); he a e age esponse ime o
any ope a ion se ice (Gene alResponseTime); and he
a e age se ice a ailabili y (A e ageA ailabili y).
Gua an ee e ms desc ibe he se ice le el objec i es
(SLOs) ha an obliga ed pa y, usually he se ice p o ide ,
mus ul ill as pa o he SLA. The SLO is an asse ion
de ined o e moni o able a iables de ined in he se ice
p ope ies sec ion o he ag eemen documen , and o e
ex e nal ac o s such as da e, ime, e c. The SLOs can
be exp essed using any sui able asse ion language. In ou
example, he p o ide assu es a minimum a ailabili y o he
se ice, and a ela ion assu ing ha he a e age esponse
ime o 4 pa icula se ice ope a ions is less o equal o
SLA Moni o -Analyse
Se ice
SLA Manage Clien
SLA
SLA
MMD
Moni o ing
di ec i es
Analyse
Moni o ed
QoS
Se ice Le el
Ful illmen (SLF)
Upda ed
MMD
SLA is he ag eed Se ice Le el
Ag eemen be ween he se ice
p o ide and he se ice consume
o he usage o Se ice
Se ice
ou pu
SLA Reposi o y
MMD
Reposi o y
Analysis
esul s
[ca los:] Compac ando pa a mejo a la isibilidad
QoSReposi o y
SLA2MMD
Se ice-o ien ed Sys em
MMD Manage
Moni o
Figu e 2. Concep ual Re e ence Model
he a e age esponse ime o any se ice ope a ion (c .
Sec ion VII o mo e de ails abou he asse ion language).
IV. CONCEPTUAL REFERENCE MODEL
In his sec ion we p opose a concep ual a chi ec u e as a
concep ual e e ence model. This a chi ec u e can be ins an-
ia ed by se e al design a chi ec u es o moni o ing SLAs,
ensu ing a decoupled s uc u e be ween he moni o ing o
he se ices and he analysis o he SLA compliance. We
illus a e he a chi ec u e using he SAP-TAM No a ion [18].
We b ie ly desc ibe he e he di e en agen s o he sys em,
as depic ed in Fig. 2, wi h hei equi ed esponsibili ies.
Clien : is he use o he pla o m. The esponsibili ies
o he clien is o p o ide he SLA o moni o , and i s goal
is o e ie e he esul s o he moni o ed SLA, s uc u ed
in a documen named Se ice Le el Ful illmen (SLF). I
is impo an no o assimila e he SALMonADA clien wi h
he consume o he se ice ( he SALMonADA clien could
be ei he he consume o he se ice, he p o ide o a hi d
pa y in e es ed in moni o ing he assessmen o he SLA).
SLA Manage : is an agen esponsible o e ie ing
and managing he moni o ed SLAs (s o age, p o isioning,
dele ion...). The SLAs a e s o ed in a SLA Reposi o y.
SLA2MMD: is an agen ha decouples he SLA Man-
age om he MMD Manage . I wo ks as a b idge be ween
he SLA (a con ac ual speci ica ion unde s ood by SLA-
dependen agen s) wi h he MMD (a speci ica ion o he
moni o ing di ec i es o con igu e a moni o ).
MMD Manage : Simila ly o he SLA Manage , he
MMD Manage is esponsible o managing he MMD
documen s. I p o ides he Moni o ing di ec i es o he
Moni o , and i is used o p o ide an upda ed MMD wi h
he moni o ed alues. The MMDs a e s o ed in a eposi o y.
Moni o : is he agen esponsible o moni o ing he
se ices QoS. The obse ed QoS a e s o ed in a eposi o y.
Analyse : is he agen used o check i he moni o ed
QoS o a se ice (ob ained h ough he MMD wi h alues)
is complian wi h he ag eed QoS included in he SLA.
The p oposed a chi ec u e o he concep ual model p o-
ides a e e ence o ins an ia e he di e en agen s wi h con-
SALMonADA
s a Moni o ing(MMD, nEndpoin ):
idSALMonClien
s opMoni o ing(idSALMon-
Clien ): bool
ge MMD(idSALMon-
Clien ): bool
No i y(MMD, id-
SALMonClien )
«con olle »
SALMonADA compose
ge Ag eemen (idSLA):
WSAg eemen
s o eAg eemen (WS-
Ag eemen ): idSLA
ge SLF(idSLA, MMD): SLF
dele eAg eemen
(idSLA):bool
gene a eMMD(WSAg eemen ):MMD
e ie eMeasu e(MMD)
s a Moni o ing
(WSAg eemen ,
nEndpoin ): idClien
s opMoni o ing(idClien )
:bool
ge SLF(idClien )
: SLF
ge MMD(idClien )
: MMD
No i y(SLF, idClien ) upda eMeasu e(measu es, MMD)
MMD Pa se
«se ice»
ADA
[ca los:] Clien a la izquie da
Clien
Consume
P o ide
«se ice»
SALMon
Figu e 3. A chi ec onic Model o SALMonADA
c e e solu ions, es ablishing a clea sepa a ion o conce ns
on he managemen o he SLAs, he MMDs and he SLFs.
V. THE SALMONADA PLATFORM
In his sec ion we p o ide a design a chi ec u e as an
ins an ia ion o he p e ious concep ual model. We p opose
SALMonADA, a pla o m able o moni o SLAs speci ied
in WS–Ag eemen . The p oposed solu ion combines wo
exis ing amewo ks ha ha e been ex ended o ealize his
p ojec : SALMon[9] and ADA[10]. We also desc ibe in his
sec ion he wo app oaches ha SALMonADA has o ob ain
he esul s o he analysis.
A. Design Le el A chi ec u e
As shown in Fig. 3, we ha e de eloped SALMonADA as
a SBS wi h he ollowing elemen s:
Clien : p o ides he SLA o moni o exp essed in WS–
Ag eemen . I is able o e ie e ei he he MMD o he SLF,
and i desi ed, i can also ecei e no i ica ions when he SLA
has been iola ed (see subsec ion V-B).
SALMonADA compose : is he se ice ha composes
he in e nal se ices o he pla o m. I p o ides he in e ace
o he clien and manages he execu ion p ocess o he
sys em. I also adds an independence laye on he in e ac ion
equi ed be ween he analysis o he SLAs (pe o med by
ADA) and he moni o ing o he se ices QoS (pe o med
by SALMon). Such a decoupled s uc u e allows o add o
modi y he in e nal componen s in a e y lexible manne .
(i.e. allows o eplace he moni o o he analyze wi hou
a ec ing he o he elemen s o he pla o m).
SALMon: is he se ice esponsible o moni o ing he
se ices QoS. I ac s as bo h he MMD Manage and Moni o
agen s o he concep ual model. A de ailed desc ip ion o he
SALMon beha io is included in sec ion VI.
ADA: is he se ice esponsible o managing and
analyze he di e en WS–Ag eemen documen s. I suppo s
he analysis o WS–Ag eemen s wi h exp essi e asse ions
inside gua an ee e ms. I ac s as bo h he SLA Manage and
Analyze o he concep ual model. A de ailed desc ip ion o
he ADA beha io is included in sec ion VII.
«moni o »
SALMon
Consume
Se ice
P o ide
Se ice
« esou ce»
MMD Pa se
«subsc ibe »
SALMonADA
Compose
«analyse »
ADA
sl :Se iceLe el
Ful illmen
«
use /applica...
SALMonADA
Clien
esponse
measu es
MMD
MMD
subsc ibe s<moni o ingSessionID, subsc ibe No i ica ionEndpoin >
loop (subsc ibe s no i ica ion)
[ o each s:subsc ibe ] SALMonADA Compose
is one o he SALMon
Subsc ibe s
esponse
clien ID
clien :Clien
al (no i ying he igh clien )
[clien .clien No i ica ionEndpoin != null]
measu e
analysisResul <Ful illmen ,
Viola ionExplana ion>
sl
se iceReques (pa ams)
s a Measu e()
se iceReques (pa ams)
s opMeasu e()
ge Se iceMMD(use C eden ials)
upda eMeasu e(measu es, MMD)
upda eMMD(MMD)
ge MMDSubsc ibe s(MMD)
new(s.subsc ibe No i ica ionEndpoin )
no i y(MMD, s.moni o ingSessionID)
ge Clien ID(s.moni o ingSessionID)
ge Clien (clien ID)
ge Se icelLe elFul illmen (clien .slaID, MMD)
e ie eMeasu e(MMD)
analyse(clien .slaID, measu e)
new(analysisResul , MMD)
new(clien .clien No i ica ionEndpoin )
no i y(sl , clien ID)
Figu e 4. Asynch onous SALMonADA app oach.
MMD Pa se : is he se ice ha implemen s he
SLA2MMD agen ex ac ing he moni o ing in o ma ion,
and i also implemen s he unc ionali y o in e ac wi h he
MMDs ( e ie e o upda e alues). Thus, he MMD s uc-
u e, whose in o ma ion is used by all pla o m componen s,
is decoupled om bo h ADA and SALMon. The e o e,
di e en MMD s uc u es can be de eloped, i needed.
B. The Asynch onous and Synch onous App oaches
SALMonADA pla o m is designed and de eloped o sup-
po asynch onous and synch onous in e ac ion s yles wi h
hei clien s. Thus, a clien , based on i s own bene i , may
choose i s p e e ed app oach. Independen ly o he selec ed
app oach a clien mus s a and s op he SALMonADA
moni o ing o be subsc ibed/unsubsc ibed as clien . The s a
p ocess equi es a WS–Ag eemen documen o moni o i s
ul illmen and such a p ocess sligh ly a ies o clien s using
asynch onous app oach because hey mus also p o ide
whe e he no i ica ion is awai ed (no i ica ion endpoin ).
Asynch onous App oach: i is he mos con enien way
o in e ac wi h he pla o m due o he asynch onous na u e
o SALMonADA se ice moni o ing and analysing. In his
sense, he pla o m inco po a es an ea ly analysis no i ica ion
ha suppo he SLA ul illmen analysis as soon as a
iola ion is obse ed. Thus, he SLF no i ica ion is sen
o he clien wi hou mo e delay han he analysis ime.
The e o e, SALMonADA no i ies hei clien s only when he
moni o ed se ice has jus been used by a se ice consume
and a SLA iola ion is incu ed. As depic ed in sequence
diag am o Fig. 4, once he clien has s a ed o moni o , he
p o ide se ice included in he epo ed WS–Ag eemen
documen is moni o ed by he SALMon componen . Nex ,
he MMD c ea ed om he moni o ed WS–Ag eemen doc-
umen is sen o he MMD Pa se wi h moni o ed measu es
o be upda ed. Finally, he new MMD is no i ied o he
SALMonADA compose ha sends i o he ADA componen
o analyse he se ice le el ul illmen o he co esponding
WS–Ag eemen documen (c . Sec ion VII o mo e de ails).
Then, he clien is no i ied abou such SLF in o ming abou :
he WS–Ag eemen documen ul illmen o no ; and in
he la e case, bo h: he speci ic iola ed WS–Ag eemen
e ms and he iola ing me ics, a e included as iola ion
explana ion. Sec ion VIII includes an example o how his
SLF is epo ed o use s. No e ha SALMonADA suppo s
he same endpoin ac ing as di e en clien s, o ins ance,
one o hem o ge he SLF, o he o s o e epu a ion analy ics
o he se ice consume and p o ide , o e en o pe o m
sel -adap a ion s a egies.
Synch onous App oach: i allows he clien o con ol
when SALMonADA ope a ions a e eques ed o ge : he
cu en MMD wi h he mos ecen moni o ing in o ma ion
ob ained by SALMon; o he cu en SLF o he WS–
Ag eemen documen analysed by ADA. Howe e , he a ail-
abili y o new moni o ing in o ma ion is no assu ed by
he pla o m due o he a o emen ioned asynch onous na u e
o i s moni o ing and analysis. Thus, i is possible o he
clien o ge he same moni o ing in o ma ion in consecu i e
MMD eques s. Only i he SALMonADA clien is ac ing
as consume o p o ide se ice, he a ailabili y o new
moni o ing in o ma ion is known.
VI. SALMON COMPONENT IN A NUTSHELL
SALMon is a amewo k aimed a moni o ing he QoS o
se ices [9]. I has been de eloped as a SBS i sel , p o iding
hence an easy in eg a ion on amewo ks de eloped as a
se ice-o ien ed sys em, such as Sel -Adap i e SBS [19] o
Cloud moni o ing [20] amewo ks. In his wo k, SALMon
has been enhanced wi h he MMD Manage Se ice, which
can be in oked h ough s anda d SOAP-based web se ice
p o ocols. The MMD Manage Se ice, in u n, in okes he
al eady exis ing Moni o Se ice o con igu e he moni o .
Fig. 5 depic s he ex ended SALMon componen s.
MMD Manage : is he se ice ha s o es he MMDs in
he eposi o y and con igu es he Moni o acco dingly. To do
so, i pa ses he MMD using he MMD Pa se componen .
Moni o Se ice: The Moni o Se ice is esponsible
o e ie ing he QoS o he se ices. To do so, i c ea es
he equi ed Measu e Ins umen s o ob ain he QoS Da a.
SALMon
MMD Reposi o y
QoS
Reposi .
QoS
Da a
«se ice»
Publishe
Ge moni o ing da a
Calcula e QoS
S o e moni o ing da a
[ca los:] Compac ando pa a mejo a la isibilidad
«se ice»
Moni o
Moni o ing Engine
En e p ise Se ice Bus (ESB)
Measu e
Ins umen
[Ma c:] Con el MMD manage
MMD Manage
SOAP
messages
«se ice»
MMD
Manage
MMD da a
s a Moni o ing
(MMD,nEndpoin )
s opMoni o ing
(idSALMonClien )
ge MMD
(idSALMonClien )
MMD Pa se
Figu e 5. Technical A chi ec u e o he ex ended SALMon
Moni o ing is pe o med by means o an En e p ise Se ice
Bus (ESB) (i.e., ins ead o in oking he se ices di ec ly, all
eques s and esponses a e sen h ough he ESB) which in
u n eeds he Measu e Ins umen s.
Measu e Ins umen : is he componen ha ob ains he
alues o a basic quali y me ic, whe eas de i ed me ics a e
calcula ed by compu ing he equi ed o mula (e.g. a e age).
The Measu e Ins umen s a e ac i a ed depending on he
quali y me ics o measu e. Once ac i a ed, hey ecei e he
SOAP messages h ough he ESB ha in e cep s hem.
Publishe Se ice: implemen s he Obse e pa e n o
se ices in a SBS. I no i ies any ele an s a e change o any
subsc ibed se ice. This pa e n equi es ha he subsc ibed
se ices ( he obse e s) implemen he equi ed in e ace o
ecei e such a no i ica ion. This is achie ed by de ining a
common WSDL-in e ace wi h he no i y me hod.
VII. ADA COMPONENT IN A NUTSHELL
ADA is an Ag eemen Documen Analysis amewo k
aimed a ex ac ing use ul in o ma ion om ag eemen docu-
men s a any SLA li e-cycle s age [10]. I has been de eloped
based on ou p e ious heo e ical wo ks on applying he
cons ain sa is ac ion p oblem (CSP) [21] pa adigm o he
au oma ed p ocu emen o web se ices [22] and explana-
ion o WS–Ag eemen documen inconsis ency and non-
compliance si ua ions [23], [24]. The main ADA ea u es
a e: (1) in e ope abili y h ough a iple dis ibu ion model,
namely: as a Ja a lib a y, as an OSGi3se ice and as a web
se ice; and (2) sol e independen h ough he use o a
seman ic mapping be ween WS–Ag eemen documen s and
CSP pa adigm, ha p o ec s ou design om he possible
a ia ions de i ed om using di e en sol e s. In his wo k,
ADA has been enhanced wi h he ADA Manage and se e al
analysis acili ies depic ed in Fig. 6 and de ailed as ollows.
ADA Manage : is esponsible o SLA s o age and
e ie al om he eposi o y; as well as he ansla ion
be ween se e al SLA models o a WS–Ag eemen -based
no malised one ha ADA is able o analyse.
3www.osgi.o g
ADA – moni o ing complian e sion
ge Ag eemen (slaID)
WS-Ag eemen
Reposi o y
Ge SLA da a
S o e SLA da a
dele e-
Ag eemen (slaID)
s o e-
Ag eemen (WSAg)
ADA Manage
SLA Manage
CSP Sol e
CSP Mapping
Ful illmen
Analyse
Viola ion
Explaine
ADA Analyse
SLA T ansla o
SLA da a
Analysis
Resul s
Ge Analysis da a
S o e Analysis da a
SLF da a
ge Se iceLe elFul-
illmen (slaID, MMD)
MMD Pa se
Analysis
Con olle
e ie e-
Measu e(MMD)
Ve sión de ADA muy SALMonADA
La e sión de ADA más gené ica debe ía ene elemen os
Como:
1. la achada de ADA con las ope aciones de análisis
2. La achada WSDL
3. OSGi complian
4. Desglosa el ansla o en XML o4People, y DSL pa se s
Figu e 6. Technical A chi ec u e o he moni o ing-complian ADA.
ADA Analyse : is esponsible o he ul illmen anal-
ysis be ween he WS–Ag eemen documen and he MMD
wi h moni o ed measu es, as well as he c ea ion o iola ion
explana ion when an un ul illmen is de ec ed. This com-
ponen is also in cha ge o he SLF s o age and e ie al.
No e ha he use o he CSP pa adigm allows ADA o
suppo s he ollowing easy- o-unde s and o non- echnical
use s exp essi e asse ion language inside WS–Ag eemen
documen s (c . SLOs o Fig. 1):
P::= P opLP|T– p edica e, opL∈ {∧ |∨ | ¬ | ⇒ |⇔}
T::= E opCE– e m, opC∈ { =| 6=|>|≥|<|≤}
E::= ID opAID |ID |li – exp ession, opAis an algeb aic ope a o
– de ined on he domain o he se ice
– p ope ies, a iables, and li e als
VIII. EXPLAINING VIOLATIONS WITH SALMONADA
Fo demons a ion pu poses, we ha e implemen ed a web
applica ion4as a SALMonADA clien in o de o speci y o
upload he WS–Ag eemen documen s o moni o , execu e
SALMonADA and ecei e he esul s. In his web appli-
ca ion, we ha e in oduced he WS–Ag eemen s o ADA
and SALMon hemsel es. By moni o ing he SLAs o hese
se ices, we assess on he one hand, he unc ionali y o
SALMonADA, and on he o he , he non- unc ional aspec s
o i s main componen s. Mo eo e , as pa o he demon-
s a ion, we ha e simula ed he consume s ha execu e
ADA and SALMon se ices. We desc ibe he e he asyn-
ch onous SALMonADA app oach h ough moni o ing he
ADA se ice and analysing he se ice le el ul illmen wi h
he WS–Ag eemen documen in o de o epo iola ion
explana ions.
To moni o he WS–Ag eemen , he SALMonADA clien
in okes he s a Moni o ing me hod speci ying he ADA
WS–Ag eemen and he endpoin o he no i ica ion. In
he demons a ed scena io, i is he same web applica ion,
bu any o he clien can be subsc ibed o he no i ica ion
4SALMonADA web applica ion can be ied a www.isa.us.es/ada.sou ce/
SLAnalyze / and a sc eencas is a ailable a gessi.lsi.upc.edu/salmon/ada/

Figu e 7. Repo ing a iola ion wi h SALMonADA
and explaining o he iola ions o he same ADA WS–
Ag eemen , such as a se ice epu a ion agen , se ice
adap a ion amewo ks, e c. Using his asynch onous ap-
p oach, as soon as a iola ion is de ec ed, i is au oma -
ically analysed, and hen epo ed o all he subsc ibed
agen s. As Fig. 7 depic s, he web applica ion highligh s
as iola ion explana ion ha he A e ageResponseTime
o explainNonCompliance ope a ion is he iola ing
me ic because i was measu ed as 3.421 seconds, while
he gua an ee e m obliga es he p o ide o espond in less
han 2seconds.
The WS–Ag eemen o ADA, al eady shown in
Fig. 1, includes mo e gua an ee e ms o moni o ,
some o hem including SLOs in ol ing mo e han jus
one quali y me ic. Hence, an app op ia e explana ion
iden i ying no only he gua an ee e m ha a e in ol ed
in he iola ion o he SLA, bu also he conc e e
iola ing me ics, is equi ed. Fo ins ance, some
ope a ions ha e a highe p io i y and a e equi ed o
be as e han he a e age esponse ime o he di e en
me hods o he se ice (A e ageResponseTime
<= Gene alResponseTime). In his case, ou
explana ion would iden i y i he iola ing me ic is ei he
A e ageResponseTime o Gene alResponseTime
because a simple iden i ica ion o he iola ed
e m is no enough o g asp he iola ion cause.
Simila ly, SALMonADA suppo s he explana ion
o iola ions o mo e exp essi e SLOs as ollows.
The p o ide may gua an ee a di e en a e age
esponse ime limi o he slowe se ice ope a ions,
depending on he gene al esponse ime o he
se ice: ((Gene alResponseTime >= 0 AND
Gene alResponseTime < 2)IMPLIES (A e age-
ResponseTime < 3)) AND ((Gene alResponse-
Time >= 2 AND Gene alResponseTime <= 4)
IMPLIES (A e ageResponseTime < 5)).
IX. CONCLUSIONS AND DISCUSSION
In his pape we p esen a concep ual e e ence model o
a moni o ing and analysis amewo k, and SALMonADA as
one o i possible ins an ia ions.
The p oposed concep ual e e ence model p o ides a
e e ence a chi ec u e o de elop a pla o m o moni o ing
SLAs. The p oposed a chi ec u e ensu es and p o ides he
ollowing se o ea u es o implemen a conc e e pla o m:
•A lexible and highly decoupled a chi ec u e is p e-
sen ed. The di e en agen s o he sys em deal exac ly
wi h he in o ma ion equi ed in sepa a e documen s:
SLA, MMD and SLF.
•We p opose he MMD as a unique documen o manage
he moni o s, in o de o (1) speci y he moni o ing
di ec i es o e ie e he di e en me ics and (2) epo
he measu ed esul s o e he speci ied me ics.
•We in oduce he SLF as he documen ha explains
clea ly he iola ions o he SLA. O he app oaches
[7], [8] suppo an e en -based iola ions explana ions
based on E en Calculus. Howe e , i has in ou con-
side a ion, he ollowing d awbacks: (1) he clien (end-
use ) mus be an expe in EC o speci y he gua an ee
e ms wi h EC o mulas, bu also o unde s and he
iola ion explana ions; and (2) i is di icul o he
clien (applica ion o end-use ) o g asp he iola ion
o igin specially when mo e han one me ic a e ela ed
in he iola ing e en and/o he iola ed EC o mula.
We p esen as an ins an ia ion o concep ual e e ence
model, he SALMonADA pla o m. As Sec ion V-B de-
sc ibes, he pla o m ex ac s he moni o ing in o ma ion
om he clien WS–Ag eemen documen as o he au ho s
p opose [5], [1]. Ou p oposal di e s om hese wo ks since
we s o e i in an independen MMD ha will be upda ed
wi h he moni o ed in o ma ion when he se ice is used
a se ice p o isioning ime. Such upda ed MMD is used
o analyse he se ice le el ul illmen in o de o epo o
he clien s an easy- o-unde s and explana ion including he
iola ed SLA e ms and i s iola ing me ics. The ea u es
ha p o ides ou pla o m can be summa ized as ollows:
•SBS: I is a SBS by i sel , and because o i s decoupled
s uc u e, he inhe en se ices can be eplaced by
o he s i hey jus implemen he equi ed in e ace.
•WS–Ag eemen complian : i is able o analyze exp es-
si e SLOs in WS–Ag eemen documen s. Mo eo e ,
he clien s do no need o be expe s in any easoning
pa adigm: nei he EC no CSP, because we suppo an
asse ion language (de ailed in Sec ion VII) inside WS–
Ag eemen documen s ha is easie o unde s and.
•Ea ly no i ica ion: We p o ide an ea ly no i ica ion
mechanism based on he obse e pa e n use ul o
sel -adap i e SBS, and o he in e es ed pa ies such as
se ice epu a ion agen s.
Howe e , he pla o m p esen ed in [7], [8] has wo key
poin s ha di e s om ou p oposal and makes i e y
appealing: (1) hey conside iola ions o an expec ed beha -
io al o he SBS o moni o by adding assump ions inside he
WS–Ag eemen documen ; and (2) as hey moni o e en s
wi h EC, i s p oposal is able o deal wi h me ics depending
on a ime in e al. Bo h a e pa o he possible imp o emen
o ou wo k, he i s in e ms o expec ed ope a ions execu-
ion low (i is possible by de ining a p ecedence o de o
se ice ope a ions inside he SLA); and he la e using he
empo al analysis o ADA ha is cu en ly no conside ed
in he p oposal.
ACKNOWLEDGMENT
This wo k has been pa ially suppo ed by: S–Cube, he
Eu opean Ne wo k o Excellence in So wa e Se ices and
Sys ems; he Eu opean Commission (FEDER); he Spanish
Go e nmen unde he CICYT p ojec s SETI (TIN2009–
07366) and P oS–Req (TIN2010–19130–C02–01); and by
he Andalusian Go e nmen unde he p ojec s THEOS
(TIC–5906) and ISABEL (P07–TIC–2533).
REFERENCES
[1] A. D. e al., “Web se ices on demand: WSLA-d i en au-
oma ed managemen ,” IBM Sys ems Jou nal, ol. 43, no. 1,
pp. 136–158, 2004.
[2] A. Kelle and H. Ludwig, “The WSLA F amewo k : Spec-
i ying and Moni o ing Se ice Le el Ag eemen s o Web
Se ices,” Ne wo k, ol. 11, no. 1, pp. 57–81, 2003.
[3] M. Palacios, J. Ga cia-Fanjul, J. Tuya, and C. De La Ri a, “A
P oac i e App oach o Tes Se ice Le el Ag eemen s,” Fi h
In e na ional Con e ence on So wa e Enginee ing Ad ances,
pp. 453–458, 2010.
[4] M. Di Pen a, G. Can o a, G. Esposi o, V. Mazza, and
M. B uno, “Sea ch-based es ing o se ice le el ag eemen s,”
in P oc. o he 9 h annual con e ence on Gene ic and e olu-
iona y compu a ion, pp. 1090–1097, ACM, 2007.
[5] M. Comuzzi and C. Ko sokalis, “Es ablishing and moni o ing
SLAs in complex se ice based sys ems,” Web Se ices, 2009.
[6] A. Michlmay , F. Rosenbe g, P. Lei ne , and S. Dus da ,
“Comp ehensi e qos moni o ing o web se ices and e en -
based sla iola ion de ec ion,” in P oc. o he 4 h In e na ional
Wo kshop on Middlewa e o Se ice O ien ed Compu ing,
MWSOC ’09, pp. 1–6, ACM, 2009.
[7] K. Mahbub and G. Spanoudakis, “Moni o ing ws-ag eemen
s: An e en calculus-based app oach,” in Tes and Analysis o
Web Se ices, pp. 265–306, Sp inge , 2007.
[8] G. Spanoudakis and K. Mahbub, “Non-in usi e moni o ing
o se ice-based sys ems,” In e na ional Jou nal o Coope a-
i e In o ma ion Sys ems, ol. 15, no. 3, pp. 325–358, 2006.
[9] M. O iol, X. F anch, J. Ma co, and D. Amelle , “Moni o ing
adap able soa-sys ems using salmon,” in Wo kshop on Se ice
Moni o ing, Adap a ion and Beyond (Mona+), pp. 19–28,
2008.
[10] C. Mülle , M. Resinas, and A. Ruiz-Co és, “A F amewo k
o Analyse WS–Ag eemen Documen s,” in P oc. o The 4 h
Wo kshop on Non-Func ional P ope ies and SLA Manage-
men in Se ice-O ien ed Compu ing (NFPSLAM-SOC’10),
2010.
[11] A. And ieux e al., “Web Se ices Ag eemen Speci ica ion
(WS-Ag eemen ) ( . g d- .192),” 2011. OGF - G id Resou ce
Alloca ion Ag eemen P o ocol WG.
[12] M. Papazoglou, V. And ikopoulos, and S. Benbe nou, “Man-
aging e ol ing se ices,” So wa e, IEEE, ol. 28, no. 3,
pp. 49 –55, 2011.
[13] O. F. Rana, M. Wa nie , T. B. Quillinan, F. B azie , and
D. Cojoca asu, “Managing Viola ions in Se ice Le el Ag ee-
men s,” in G id Middlewa e and Se ices Chap e Ti le -
Managing Viola ions in Se ice Le el Ag eemen s, pp. 349–
358, 2008.
[14] M. Comuzzi and G. Spanoudakis, “Dynamic se -up o mon-
i o ing in as uc u es o se ice based sys ems,” in SAC,
pp. 2414–2421, ACM, 2010.
[15] B. Pe nici, “Adap a ion o web se ices based on QoS sa is-
ac ion,” Se ice-O ien ed Compu ing, pp. 65–75, 2011.
[16] R. Kazhamiakin, B. We zs ein, D. Ka as oyano a, M. Pis o e,
and F. Leymann, “Adap a ion o se ice-based applica ions
based on p ocess quali y ac o analysis,” in ICSOC/Se ice-
Wa e Wo kshops, ol. 6275 o LNCS, pp. 395–404, 2009.
[17] O. Mose and F. Rosenbe g, “Non-in usi e moni o ing and
se ice adap a ion o WS-BPEL,” in P oc. o he 17 h
in e na ional con e ence in WWW, pp. 815–824, 2008.
[18] S. AG, S anda dized Technical A chi ec u e Modeling: Con-
cep ual and Design Le el. Ve sion 1.0. SAP, 2007.
[19] O. Sammodi, A. Me zge , X. F anch, M. O iol, J. Ma co, and
K. Pohl, “Usage-based online es ing o p oac i e adap a ion
o se ice-based applica ions,” in COMPSAC, pp. 582–587,
IEEE Compu e Socie y, 2011.
[20] A. Ke esz, G. Kecskeme i, M. O iol, A. C. Ma osi,
X. F anch, and J. Ma co, “In eg a ed moni o ing app oach o
seamless se ice p o isioning in ede a ed clouds,” in PDP,
2012.
[21] E. Tsang, Founda ions o Cons ain Sa is ac ion. Academic
P ess, 1995.
[22] A. Ruiz-Co és, O. Ma ín-Díaz, A. Du án, and M. To o,
“Imp o ing he Au oma ic P ocu emen o Web Se ices
using Cons ain P og amming,” In . Jou nal on Coope a i e
In o ma ion Sys ems, ol. 14, no. 4, 2005.
[23] C. Mülle , A. Ruiz-Co és, and M. Resinas, “An Ini ial
App oach o Explaining SLA Inconsis encies,” in P oc. o
he 6 h In . Con . on Se ice-O ien ed Compu ing (ICSOC),
pp. 394–406, 2008.
[24] C. Mülle , M. Resinas, and A. Ruiz-Co és, “Explaining he
Non-Compliance be ween Templa es and Ag eemen O e s
in WS-Ag eemen *,” in P oc. o he 7 h In e na ional Con e -
ence on Se ice-O ien ed Compu ing (ICSOC), pp. 237–252,
2009.