AGREEING ON AND CONTROLLING SERVICE LEVELS

IN SERVICE-ORIENTED ARCHITECTURES

Benjamin Heckmann

, Andrew D. Phippen

, Ronald C. Moore

and Christoph Wentzel

University of Applied Sciences Darmstadt, Haardtring 100, 64295, Darmstadt, Germany

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

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.

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

 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

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

4 SOLUTION DESIGN

To agree on feasibility and workload of business pro-

cesses Business Service Level Agreements (BSLA)

Intermediate logic that changes the invocation target of

a service request at runtime.

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

(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-

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

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

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

. 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.,

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