QUERYING WORKFLOWS OVER DISTRIBUTED SYSTEMS

Hanna Kozankiewicz

Institute of Computer Science, Polish Academy of Sciences, Ordona 21, 01-237 Warszawa, Poland

Krzysztof Stencel

Institute of Informatics, Warsaw University, Banacha 2, 02-097 Warszawa, Poland

Kazimierz Subieta

Polish-Japanese Institute of Information Technology, Koszykowa 86, 02-008 Warszawa, Poland

Keywords: Workflow, query language, object-oriented, distributed system, pre-condition, post-condition, SBA, SBQL.

Abstract: We describe a new paradigm of declarative workflow, where an activity is modelled as an object which can

have other activities as subobjects. All activities act in parallel and have pre- and post-conditions

determining when they are to be started and when they are to be terminated. The proposal is based on the

Stack-Based Approach (SBA) to object-oriented query languages, query language SBQL and updatable

SBQL views. The idea is currently implemented within the European project eGov Bus.

1 INTRODUCTION

During our previous experience with workflow

management software (Momotko, 2004), (Momotko,

2005) we have made important observations

concerning the nature of querying a workflow data

environment and the nature of workflows.

Workflows are usually perceived as activities and

transitions (WfMC, 1995) in the spirit of Petri nets.

This view is common and quite successful, but has

disadvantages. There are problems concerning

parallel execution of sub-processes in a distributed

environment. Usually the parallelism is supported by

split/join operations build into the control of the

workflow. However, such synchronization of

parallel processes might be hard in many practical

situations, e.g. when splitting/joining concerns

statically unknown number of sub-processes. Such

situations are frequent in distributed and

heterogeneous workflow environment. There are

other disadvantages of too stiff control flow within

some graph of activities.

In the current projects we investigate an

alternative paradigm, when activities are objects in

an object-oriented model and sub-activities are sub-

objects. By definition, all such objects have some

attributes which can be queried, but they have also

executable parts which by definition run in parallel.

We provide synchronization means known as pre-

conditions and post-conditions that make it is

possible to start and terminate particular activities.

Sequential execution of activities can be forced by a

pre-condition of a next activity that depends on the

state of a previous one. For a complex activity a

post-condition determines when it has to be

terminated. Pre-conditions and post-conditions are

formulated in a query language which may work on

the entire workflow environment, the state of all the

population of workflow processes and the state of

resources (e.g. workflow performers). The

advantages of the approach concern, in particular,

inherent parallelism of activities, easier distribution

of parallel activities among many machines and

dynamic changes of executed processes (e.g. by

runtime inserting a new activity or removing an

activity). Synchronization of workflow sub-

processes by a post-condition corresponds to join

operators in a typical workflow definition language,

but a post-condition expressed as a query is

incomparably more powerful and flexible.

The proposed paradigm makes workflows close

to PERT nets that are used in project planning to

251

Kozankiewicz H., Stencel K. and Subieta K. (2007).

QUERYING WORKFLOWS OVER DISTRIBUTED SYSTEMS.

In Proceedings of the Third International Conference on Web Information Systems and Technologies - Society, e-Business and e-Government /

e-Learning, pages 251-254

DOI: 10.5220/0001285102510254

 SciTePress

analyse critical time paths and other features of

available resources. PERT nets assume inherent

parallelism of processes, constrained however by

their in-out dependencies and availability of

resources necessary for execution. We argue that

this paradigm is worth investigation concerning new

qualities that it may bring for definition and

execution of workflow processes and to make them

more flexible for dynamic changes and parallel

execution.

Our approach is based on the Stack Based

Approach (SBA) to query languages and its

powerful query/programming language SBQL

(Subieta, 2004), (Subieta, 2006). The language can

be used to perform efficient tracking and monitoring

tools for population of workflows, just like in BPQL

(Momotko, 2004). We propose a new metamodel for

workflow processes and then show how SBQL can

be used in workflow-specific applications. An

important element of our architecture is that a

workflow system can work on resources that are

distributed among different organizations.

2 ARCHITECTURE OF A

DISTRIBUTED WORKFLOW

A workflow mechanism operates on top of a virtual

store which integrates data and services from

different distributed sources. The details concerning

the store are described e.g. in (Kozankiewicz, 2005).

It is implemented by means of updatable views.

The resources are supplied by local servers.

Local schemata have to be adapted to the

requirements of the virtual store. To this end, the

administrators of local servers have to implement

wrappers, which provide mappings of local

data/services to the global canonical data/service

model assumed by the virtual store. These wrappers

can be implemented as updatable views. A basic

integration mechanism is on an intermediate server,

which contains integrators that virtually fuse

resources supplied by local servers. Integrators

resolve heterogeneities and redundancies and join

fragmented collections. Integrators are also to be

implemented as updatable views. On a client level

there are customization views that adopt the global

virtual schema to the need of a particular client

application and determines client’s access privileges.

A workflow mechanism is on a special server which

through the virtual store can access all the

distributed resources. The workflow server is

connected to its clients (workflow participants).

Such integration of e-Gov resources supported

by various partly independent institutions allows

building a workflow system over a distributed e-Gov

infrastructure. When resources are virtually

integrated, one can use the query/programming

language SBQL to write applications, with the use of

abstractions such as classes, methods, procedures

and updatable views. SBQL is also used in

definitions of workflow processes to query all the

workflow environment (through the workflow

metamodel) and to formulate pre- and post-

conditions for particular workflow activities.

3 DECLARATIVE WORKFLOWS

Our basic assumptions for workflow systems built

on top of a virtual store are as follows. Each activity

is modelled as an object. It can consist of sub-

activities, i.e. sub-objects. There is no restriction on

the size and the number of levels of the hierarchy of

the activity nesting. The top level activity can be

considered a workflow process.

Activities can be dynamically inserted into a

super-activity or removed from it. In such a way we

provide a flexible mean to adapt processes/activities

to dynamically changing complex application

requirements. There is no explicit graph describing

control flow between activities, because for many

workflow types such graphs are not enough flexible

concerning parallel execution, dependency on

resources necessary for an activity and the

possibility of runtime workflow changes. Instead, all

activities have pre- and post-conditions defining

when they are to be started and when they are to be

finished. Such conditions are expressed in SBQL.

Activities can include inner objects, procedures,

views, etc. Atomic activities have executable parts

that work concurrently to other activities. For

complex activities an executable part can be treated

as sub-activity. A sub-activity can be initialised if

and only if its super-activity has been initialised. All

activities are described by definitions, which can be

used e.g. to create activity instances and to type

checking. Not all sub-activities must be terminated

to terminate the given activity – this depends on its

post-condition only. Similarly, terminating all sub-

activities does not necessarily mean that the given

activity must be terminated too.

One of possible goals of our idea is to support

life events of citizens or enterprises. In Figure 1 we

show how activities related to a life event ‘family

location change’ are represented using our workflow

model. On the first part we show sub-activities such

as changing residence, changing schools, changing

home MD, and so on. On the second part we show

WEBIST 2007 - International Conference on Web Information Systems and Technologies

252

how the process of changing location can be

modelled as a hierarchy of workflow activities. Note

that the activity ‘Change school’ can appear zero,

one, or more times, depending on the number of

children in the family.

Figure 1: Activities as objects.

Activity instances are complex objects with the

following example structure:

ActivityName{ ontology, priority, …

activityState,

precondition, action,

cost, duration, deadline,…

refs to performers,

refs to subactivities,

postcondition,

ref to ActivityDef,

local objects, procedures, …}

Each activity is identified by its name, which bears

its business semantics. An activity can be in several

states (waiting, pending, terminated, etc.). Complex

activities have references to their subactivities and

post-condition. An activity may have a reference to

its definition in some metamodel.

Activity instances can generate activity

instances. In this way, in particular, one can organize

loops of some activities. As objects, activities can be

connected to classes containing some of their

invariant properties. The code of activity actions

may require special workflow-oriented functions,

such as an access to workflow client software.

4 SBQL FOR WORKFLOWS

In SBQL we can easily define workflow-specific

procedures and functions. For simplification we skip

typing information and make some simplifications

concerning right object naming and SBQL syntax.

The following function returns activities which are

in one of the states specified by the parameter and

have the highest priority.

procedure HighestPriorityActiv(states) {

create maxPriority = max(Activity.priority);

return Activity where priority = maxPriority

and activityState ∈ state; }

The function can be called in the following query,

that selects all suspended and pending actives with

deadline during the next 10 days:

HighestPriorityActiv(bag(‘suspended’, ‘pending’))

where deadline – currentDate < 10

By SBQL recursive procedures one can easily

process nested activity structures, e.g. evaluate cost

of an activity and all its sub-activities:

procedure getActivityCost( act ) {

create local c = act.cost;

for each act.subactivity.Activity as a do

c := c + getActivityCost(a));

return c; }

A workflow system can offer a large set of

predefined workflow procedures and functions, for

instance, a function returning the best workflow

participant for the given task, a function that

recursively returns all sub-activities, a procedure that

insert a new activity into the given one, etc.

SBQL can also be used to write pre- and post-

conditions. Below, we present a query which checks

whether we can find an employee with the current

workload less than 10 activities:

exists (Performer where

count(performs.Activity) < 10)

The next query is a post-condition of an activity that

verifies whether all its sub-activities with the name

‘Order Approval’ have been successfully finished:

forall (self.subactivity.OrderApproval) holds

activityState = ‘completed’

Pre- and post-conditions can be the subject of

manipulations during runtime, in particular, new

pre- and post-conditions can substitute the previous

ones by using a proper SBQL statement. Such

property of programming languages is known as

reflection, i.e. dynamic modification of a running

program. Reflexive capabilities require careful

QUERYING WORKFLOWS OVER DISTRIBUTED SYSTEMS

253

treatment in the programming language, in

particular, they are more difficult for strong type

checker. For this reason the function that modifies

pre- and post-conditions is not a regular assignment

function but the function that calls parser, compiler

and optimizer and then converts a condition into an

executable bytecode.

Another application of SBQL queries and

procedures is monitoring and tracing of the entire

population of workflow processes. Below we present

some examples.

• List all pending activities with the deadline

expiring during the next 3 days:

Activity where activityState= ‘pending’ and

deadline – currentDate < 3

• List all pending activities which can be started,

i.e. all their pre-conditions are valid. In this

example we use a function valid that checks if a

set of conditions is fulfilled.

Activity where activityState = ‘pending’

and valid(precondition)

• List all active activities which have already

missed their deadlines.

Activity where activityState = ‘active’ and

deadline < currentDate

• List all activities which are close to their

deadlines and not assigned to anyone yet.

Activity where count(performers) = 0

and deadline – currentDate < 3

• List all activities performed by overloaded

employees, i.e. employees assigned to instances

of activities which must be completed in

following 7 days while their total duration is

greater that 40 working hours:

Activity where sum(performers.Performer.

performs. (Activity where deadline <

currentDate + 7) . duration) > 40

5 CONCLUSIONS

We have shown how a declarative workflow systems

can be implemented on the top of distributed object-

oriented systems. The case covers most of e-

Government applications where complicated

routines should be performed at multiple institutions.

If a citizen or a company faces a life event, a number

of governmental agencies is to be involved in

serving it. The goal of an e-Government system

should be to let this citizen carry out a single visit to

one governmental web page in order to act upon this

life event. He/she should not be forced to visit many

web pages with different rules of the user behaviour.

A carefully designed workflow system on the top

of integrated e-Government resources is able to

achieve this goal. Such a system will facilitate the

concurrency of processes by using more powerful

synchronization mechanisms in comparison to join

and split operators. The critical feature is flexibility

of dynamic changes of processes. Furthermore, such

a system should support effective monitoring and

tracing of the entire population workflow processes

and activities, including resources and documents

necessary to perform them. For such complex

situation it seems that a high level database query

and programming language like SBQL is the best

choice.

ACKNOWLEDGEMENTS

This work is supported by European Commission

under the 6

Framework Programme project e-Gov

Bus, IST-4-026727-ST

REFERENCES

eGov Bus, 2006. Advanced eGovernment Information

Service Bus Project. http://www.egov-

bus.org/web/guest/home

Momotko, M., Subieta, K., 2004. Business Process Query

Language - a Way to Make Workflow Processes More

Flexible. Proc. of ADBIS 2004, Springer LNCS 3255,

pp. 306-321

Momotko, M., 2005. Tools for Monitoring Workflow

Processes to Support Dynamic Workflow Changes.

PhD Thesis, 2005. http://www.ipipan.waw.pl/~subieta/

→ Finished PhD-s

Subieta, K., 2004. Theory and Construction of Object-

Oriented Query Languages. Editors of the Polish-

Japanese Institute of Information Technology, 520

pages (in Polish)

Subieta, K., 2006. Stack-Based Approach (SBA) and

Stack-Based Query Language (SBQL).

http://www.sbql.pl

WfMC, 1995. Workflow Management Coalition: The

Workflow Reference Model, WfMC-TC-1003 issue

1.1

Kozankiewicz, H., Stencel, K., Subieta, K., 2005.

Implementation of Federated Databases through

Updatable Views. Proc. of the European Grid

Conference, Springer LNCS 3470, pp. 610-620

WEBIST 2007 - International Conference on Web Information Systems and Technologies

254