AGREEING ON AND CONTROLLING SERVICE LEVELS
IN SERVICE-ORIENTED ARCHITECTURES
Benjamin Heckmann
1
, Andrew D. Phippen
2
, Ronald C. Moore
1
and Christoph Wentzel
1
1
University of Applied Sciences Darmstadt, Haardtring 100, 64295, Darmstadt, Germany
2
University of Plymouth, Drake Circus, PL4 8AA, Plymouth, U.K.
Keywords:
Availability, Business Processes, Monitoring, Reliability, SOA.
Abstract:
Business Service Level Agreements (BSLAs) are introduced as a generalised concept to agree on feasibility
and workload of business processes hosted in service-oriented architectures as an alternative to technical SLA.
Based on BSLAs an according approach to control feasibility at runtime is presented.
1 INTRODUCTION
In Service-oriented Architectures (SOA) Service
Level Agreements (SLAs) (Group et al., 2011) are
used to specify technical thresholds, corresponding
actions to keep them and penalties when failing. From
a business point of view, only the feasibility of busi-
ness processes is of interest for economical success.
Business processes are feasible if a given workload is
processed in a certain maximum time frame. Work-
loads of processes are the sum of all current actively
processed process instances. Determining whether a
business process based on a SOA is feasible can be
complex for highly meshed service cascades includ-
ing redundant alternate service offers (e.g., for load
balancing).
This research is conducted in cooperation with
secco
1
as business partner, introducing the problem
statement. The applied research aims at providing a
concept to monitor the feasibility and workload of
business processes, hosted on multi-tier IT service
provision infrastructures for SOA services, without
the need for active technical monitoring. To elabo-
rate an solution approach, a typical multi-tier IT ser-
vice provision infrastructure operated in the context
of our business partner is analysed in section 3. A
generalised concept to track feasibility and workload
of business processes hosted based on multi-tier in-
frastructures is elaborated in section 4. In section 5,
a BSLA monitoring framework is implemented as a
proof-of-concept.
1
secco advanced GmbH, Grossostheim, Germany,
http://www.seccoadvanced.de.
2 RELATED WORKS
This paper offers an approach to technically con-
verge the quality-related ontologies of service, expe-
rience, and business as introduced in (Van Moorsel,
2001; Dobson and Sanchez-Macian, 2006). Most
other authors address technical perspectives on SLA
in SOA. From the IT architecture point of view,
authors deal with SLA descriptions of performance
modelling (Brebner, 2008), SLA-driven development
(Muthusamy et al., 2009) or dependability through-
out the life cycle (Stantchev and Malek, 2010). In
operations management SLA are mostly enforced
through an distribute-and-enforce tactic. By (Hsu
et al., 2008; Raibulet and Massarelli, 2008; Chen
et al., 2009; Muthusamy and Jacobsen, 2008) highly
detailed SLAs are defined, distributed and then en-
forced on each member of a service cascade. The
complexity to manage such approaches increases with
the complexity of the given cascade. (Stantchev and
Schroepfer, 2008) decouples SLA operations man-
agement from the complexity of a service cascade.
This paper presents a similar approach and advances
it by embedding BSLA in a whole life cycle concept
(Heckmann and Phippen, 2010). Technical operations
is focused on the technical monitoring of technical re-
source thresholds. Three types can be distinguished:
active, passive and agent-based (Utlik and Alexeyev,
2010). Other authors propose the workflow monitor-
ing of business process workloads. It is focused on the
workflow state rather than underlying technical mea-
surements (Ou et al., 2008). In contrast, (Moser et al.,
2008) aims to provide a non-intrusive workflow mon-
267
Heckmann B., D. Phippen A., C. Moore R. and Wentzel C..
AGREEING ON AND CONTROLLING SERVICE LEVELS IN SERVICE-ORIENTED ARCHITECTURES.
DOI: 10.5220/0003871702670270
In Proceedings of the 2nd International Conference on Cloud Computing and Services Science (CLOSER-2012), pages 267-270
ISBN: 978-989-8565-05-1
Copyright
c
2012 SCITEPRESS (Science and Technology Publications, Lda.)
itoring approach combined with active SLA manage-
ment. This paper broadens this approach to incorpo-
rate technical monitoring data and address general IT
services based on IP networks.
3 MULTI-TIER
INFRASTRUCTURE ANALYSIS
In the context of secco, SOA infrastructures consist of
7 horizontal layers, shown in Figure 1. Business pro-
cesses are represented by technical workflows acting
as service consumers on the top layer. Business func-
tionality is provided by the orchestration (Andrews
et al., 2003) of application layer services, for example
web services (Haas and Brown, 2004). These service
instances are hosted on the application infrastructure
layer (e.g., within database systems or application
servers). All software components from upper lay-
ers are deployed on the operating system layer, each
instance running in a virtual machine on the virtual in-
frastructure layer. The virtual hardware is mapped to
resources on the physical systems layer. As the final
layer, the network services connect these systems re-
lying on resources such as routers, switches or domain
name services. Complementary to the previously de-
scribed horizontal multi-tier SOA infrastructure, there
is the vertical technical monitoring layer. It evaluates
technical measuring points of the horizontal layers,
such as network availability, CPU load, memory con-
sumption or storage usage.
Analysed service cascades include redundant ser-
vice offers and the support for dynamic coupling
2
be-
tween service consumer and provider should be con-
sidered. The technical monitoring solutions Amber-
point, Progress Actional, SOA Manager Service Man-
ager, Oracle Enterprise Manager SOA Management
Pack and OpTier CoreFirst do not offer sufficient in-
formation to gather quantifications of failure impacts
and reliable conclusions on the feasibility of the im-
plemented business processes in the given SOA in-
frastructures. Specifically analyses of the current fea-
sibility of business processes in scenarios with redun-
dant service offers fail due to the evaluation of secco
3
.
4 SOLUTION DESIGN
To agree on feasibility and workload of business pro-
cesses Business Service Level Agreements (BSLA)
2
Intermediate logic that changes the invocation target of
a service request at runtime.
3
Based on an internal technical report in June 2009.
are proposed as abstraction layer for the contracting
of service quality between service consumer and ser-
vice provider. BSLAs are focused on the description
of the estimated usage behaviour, extended by the
declaration of the maximum allowed response time
for service requests and can optionally be enriched by
the declaration of maintenance windows, maximum
unplanned downtimes, fines, pricing or other non-
functional properties. BSLAs are aimed at replacing
SLAs. In BSLAs the consumer’s usage behaviour is
described by Usage Patterns (Heckmann and Phip-
pen, 2010), which offer an approach for the descrip-
tion of the quantitative consumer-provider-relation in
terms of request frequency and processing complex-
ity. BSLAs enable analyses on business process feasi-
bility and workload by specifying the contracted us-
age. In opposite, the monitored usage reflects the cur-
rent request amount and resource utilisation within
IT infrastructures. The business process’s workload
is determined by comparing its contracted and mon-
itored usage, assuming all infrastructure components
are technically available. To enable monitoring of the
request amount this paper proposes the use of a cen-
tralised request routing component, named Service
Broker
4
(see Figure 1). The Service Broker provides a
measuring point for request amounts per business pro-
cess, which represent the process’s workload, taking
the contracted usage as reference. The business pro-
cess’s feasibility is lead back from its workload com-
bined with information about the technical availabil-
ity of all infrastructure components hosting the pro-
cess.
The aggregation of technical monitoring informa-
tion in service cascades hosting business processes is
addressed by a topology graph. The term topology
graph is introduced to reflect the functional depen-
dencies between the components in an IT infrastruc-
ture. To build the topology graph infrastructure com-
ponents can be retrieved from a configuration man-
agement database (CMDB) (Group et al., 2011). To
represent redundant service offers within a topology
graph service lines are introduced. A service line is
a logical group of infrastructure components that are
necessary to provide an application layer service. To
aggregate resource utilisation of service line spanning
resources the term component category is introduced.
Component categories logically group infrastructure
components that provide similar functionalities (e.g.,
application servers, which provide hosting of applica-
tion layer services), see Figure 1 as example. To cal-
culate the resource utilisation, each topology graph
node is enriched with interpreted technical monitor-
4
The Service Broker acts as a economical load-balancer
for cloud infrastructures (Heckmann, 2007).
CLOSER2012-2ndInternationalConferenceonCloudComputingandServicesScience
268
Figure 1: Multi-tier architecture including service broker, service line and component category as example.
ing information. The enriched graph is introduced as
availability graph. Considered are two levels of mon-
itoring data interpretation.
In case of state-based analyses the availability of
a graph node is lead back by interpreting state-related
technical monitoring data of the represented infras-
tructure component, such as ping states
5
retrieved
from a technical monitoring system. Interpretation of
this data limits results to two simple states: available
and non-available. This variant is simpler to impose,
but is less significant when deducting the feasibility of
constitutive business processes. For load-based anal-
yses the availability of a graph node is estimated com-
paring current resource utilisation with its maximum
capacity. The utilisation is calculated based on load-
related technical monitoring data like CPU load. This
enables the provision of proportional metrics reflect-
ing the current availability of the represented resource
(e.g., 20 % utilisation of a DNS server).
The Service Broker estimates a business process
as feasible if in the availability graph all state-based
nodes of at least one service line are available and
the resource utilisation of all load-based component
categories offer sufficient reserves to process the es-
timated usage. The estimated usage for a given time
frame is calculated by subtracting the monitored us-
age for a given business process from its contracted
usage. In strictly state-based availability graphs only
the process workload is taken into account when cal-
culating the estimated usage, otherwise also the re-
source utilisation is incorporated.
5 PROOF-OF-CONCEPT
IMPLEMENTATION
As proof-of-concept an application was implemented
representing the state-based availability graph of an
5
Ping enables the monitoring of the technical network
interface of a remote system and is specified in the Internet
Control Message Protocol (ICMP) (Postel, 1981).
exemplary business process. For technical monitor-
ing the implementation bears on Zabbix
6
. The appli-
cation uses a given XML configuration file represent-
ing the topology graph. Its topology graph consists of
nine nodes representing two service lines. Each node
description is enriched with a reference to its Zabbix
database identifier. The application extracts the state
values for all nodes from the monitoring system. It
autonomously determines the service lines and aggre-
gates their availability states. As first outcomes, the
ability to realise state-based availability graphs based
on technical monitoring data is presented. The valida-
tion of the determination reliability is subject of future
research.
6 CONCLUSIONS
This paper introduces a alternate abstraction layer in
order to agree on the feasibility and workload of a
business process instead of technical thresholds of the
underlying technology as known from common SLA.
This layer is called Business Service Level Agree-
ments (BSLA) and establishes a black box around
service capacity and technical implementation, thus
loosening the coupling between technical service pro-
vision and business service consumption on the level
of service agreements. Based on the identified quali-
fied technical indicators, the paper evolves that BSLA
approach. Corresponding, an approach for the techni-
cal monitor and enforcement of BSLAs during oper-
ations is presented. Concluding, a proof-of-concept
implementation demonstrates the capabilities of these
approaches.
REFERENCES
Andrews, T., Curbera, F., Dholakia, H., Goland, Y., Klein,
J., Leymann, F., Liu, K., Roller, D., Smith, D.,
6
http://www.zabbix.com
AGREEINGONANDCONTROLLINGSERVICELEVELSINSERVICE-ORIENTEDARCHITECTURES
269
Thatte, S., Trickovic, I., and Weerawarana, S. (2003).
BPEL4WS, Business Process Execution Language for
Web Services Version 1.1. IBM, BEA Systems, Mi-
crosoft, SAP AG, Siebel Systems.
Brebner, P. C. (2008). Performance modeling for service
oriented architectures. In Companion of the 30th inter-
national conference on Software engineering, pages
953–954, Leipzig, Germany. ACM.
Chen, Y., Iyer, S., Milojicic, D., and Sahai, A. (2009). A
systematic and practical approach to generating poli-
cies from service level objectives. In Integrated Net-
work Management, 2009. IM ’09. IFIP/IEEE Interna-
tional Symposium on, pages 89 –96.
Dobson, G. and Sanchez-Macian, A. (2006). Towards uni-
fied QoS/SLA ontologies. In IEEE Services Com-
puting Workshops, 2006. SCW ’06, pages 169–174.
IEEE.
Group, A., TSO, and Office, C. (2011). ITIL.
http://www.itil-officialsite.com.
Haas, H. and Brown, A. (2004). Web services glossary.
http://www.w3.org/TR/ws-gloss/.
Heckmann, B. (2007). Service provision in a utility comput-
ing environment. In Proceedings of the Third Collab-
orative Research Symposium on Security, E-Learning,
Internet and Networking, pages 185–198, Plymouth,
UK. Lulu.com.
Heckmann, B. and Phippen, A. (2010). Quantitative and
qualitative description of the consumer to provider re-
lation in the context of utility computing. In Proceed-
ings of the Eighth International Network Conference
(INC 2010), pages 335–344, Heidelberg, Germany.
Hsu, C., Liao, Y., and Kuo, C. (2008). Disassembling
SLAs for follow-up processes in an SOA system.
In 2008 11th International Conference on Computer
and Information Technology, pages 37–42, Khulna,
Bangladesh.
Moser, O., Rosenberg, F., and Dustdar, S. (2008). Non-
intrusive monitoring and service adaptation for WS-
BPEL. In Proceeding of the 17th international con-
ference on World Wide Web, pages 815–824, Beijing,
China. ACM.
Muthusamy, V. and Jacobsen, H. (2008). SLA-driven dis-
tributed application development. In Proceedings of
the 3rd workshop on Middleware for service oriented
computing, pages 31–36, Leuven, Belgium. ACM.
Muthusamy, V., Jacobsen, H., Chau, T., Chan, A., and
Coulthard, P. (2009). SLA-driven business process
management in SOA. In Proceedings of the 2009 Con-
ference of the Center for Advanced Studies on Col-
laborative Research, pages 86–100, Ontario, Canada.
ACM.
Ou, T., Sun, W., Guo, C., and Li, J. (2008). Visual-
ized monitoring of virtual business process for SOA.
In Proceedings of the 2008 IEEE International Con-
ference on e-Business Engineering, pages 767–770.
IEEE Computer Society.
Postel, J. (1981). Internet control message protocol - RFC
792. http://tools.ietf.org/html/rfc792.
Raibulet, C. and Massarelli, M. (2008). Managing non-
functional aspects in SOA through SLA. In 2008 19th
International Conference on Database and Expert
Systems Applications, pages 701–705, Turin, Italy.
Stantchev, V. and Malek, M. (2010). Addressing depend-
ability throughout the SOA life cycle. IEEE Transac-
tions on Services Computing, 99(PrePrints).
Stantchev, V. and Schroepfer, C. (2008). Techniques for ser-
vice level enforcement in web-services based systems.
In Proceedings of the 10th International Conference
on Information Integration and Web-based Applica-
tions & Services, pages 7–14, Linz, Austria. ACM.
Utlik, A. and Alexeyev, N. (2010). Comparative analysis of
service level agreement monitoring methods. In Mod-
ern Problems of Radio Engineering, Telecommunica-
tions and Computer Science (TCSET), 2010 Interna-
tional Conference on, pages 346 –346.
Van Moorsel, A. (2001). Metrics for the internet age: Qual-
ity of experience and quality of business. 5th Per-
formability Workshop.
CLOSER2012-2ndInternationalConferenceonCloudComputingandServicesScience
270