INTEGRATED DESIGN OF EVENT STREAM SERVICE SYSTEM

ARCHITECTURE (ESSSA)

Kyungeun Park, Yanggon Kim

Department of Computer & Information Science, Towson University, 7800 York Road, Towson, MD 21252, USA

Jinkyu Lee

, Juno Chang

ReegySys, Inc., 456 Dogok-dong, Gangnam-gu, Seoul, 135-270, Republic of Korea

Div. of Media Technology, Sangmyung University, 7 Hongji-dong, Jongno-gu, Seoul, 110-743, Republic of Korea

Keywords: Event Stream Management, Event Query Language (EQL), Radio Frequency Identification (RFID),

Ubiquitous Sensor Network (USN), Object-Relational Spatial Database Management System.

Abstract: This paper proposes the integrated design of an event stream service system architecture (ESSSA) which

can be used as a gateway to capture and process enormous amounts of event streams continuously generated

by radio frequency identification (RFID) or variable sensors over ubiquitous sensor network (USN) installed

in a wide variety of locations, transfer the refined events to the various event-driven applications connected

to this architecture, and allow the corresponding actions to be taken by the applications. The events

continuously delivered from versatile origins and their streams are inherently uncertain, unbounded, and

time-varying. Their arrival rate might severely fluctuate in some cases. Accordingly, this design focuses on

the elaborate handling of the event streams by facilitating robust event gathering scheme as a front-end

gateway in combination with a back-end event processing engine. The paper also contains the time-varying

event query processing schemes of the ESSSA by extending the object-relational spatial database

management system, ZEUS. The ESSSA can be applied as a framework for interfacing RFID or USN

sensor based applications such as an enterprise resource planning system (ERP), a supply chain

management system (SCM), a warehouse management system (WMS), or a centralized command and

control system.

1 INTRODUCTION

Today’s highly networked environments encouraged

by rapid growth of highly advanced wireless

telecommunication and corresponding software

technologies encourage us apply them to real world

and take fully advantage of the conveniences and

efficiencies they present. To do so, it necessarily

follows that we have to consider the effective way to

handle continuously generated events or signal

streams in such ubiquitous computing environments.

Events streams are to be captured from the event

origins, filtered by the predefined conditions,

maintained within the efficient events archive and

delivered to the appropriate service applications

depending on the event status.

Technical considerations regarding on the related

issues are required to design the event stream service

system architecture (ESSSA) that can be used as an

infrastructure for handling enormous amounts of

event streams generated by all sorts of event origins

which are equipped with radio frequency

identification (RFID), sensors of ubiquitous sensor

network (USN), or conventional wireless

telecommunication facilities. In this paper, the event

capturing system built on top of the main memory is

to support real time transactions and enables inflow

of the enormous events to be captured and evaluated

instantaneously. Disk based conventional database

management system (DBMS) is used as a backup

repository for the post-analysis to the once

accumulated event history data.

Event data discussed in this paper is inherently

uncertain, unbounded, and time-varying and their

arrival rate from various event origins might

severely fluctuate in some cases. For this reason,

Park K., Kim Y., Lee J. and Chang J. (2007).

INTEGRATED DESIGN OF EVENT STREAM SERVICE SYSTEM ARCHITECTURE (ESSSA).

In Proceedings of the Second International Conference on e-Business, pages 51-56

DOI: 10.5220/0002113400510056

 SciTePress

elaborate consideration has to be taken into account

to capture and maintain the events, evaluate the pre-

defined event rules and react according to the

corresponding status of the events.

On receiving the events, the Event Collector of

the ESSSA pre-analyzes and classifies the categories

to which individual event belongs. The recognized

events are piled on the associated event queues

which are constructed by event stream within the

main memory, while the Event Query Manager,

important sub-component of the ESSSA parses the

event queries requested by the service applications

and extracts the event monitoring rules. They are

registered to the event monitoring rule tables in

order for the continuously incoming events to be

evaluated immediately as soon as they are identified

as satisfying events to a rule of interest.

Simple event is interrelated between events and

may imply different meaning according to the

context and the arrival of time. In this paper, we

extended the spatial query language supported by

ZEUS, object-relational spatial DBMS (Park, et al,

1998), to allow the ESSSA to define events and

manipulate their behavior and the relationship

beteen them. As the events received from ubiquitous

equipments may exist in different spatial locations,

using the spatial fearures of ZEUS system makes the

target system more powerful and easily applicable to

the real world applications.

It is required to provide a standard interface that

facilitates seamless connection across add-on service

applications. To resolve these requirements, service-

oriented architecture (SOA) interface technology is

selected as an interface infrastructure. This will

provide the plug-and-play interfaces and accelerate

the extension of the existing service framework with

the new services considered to be attached.

This paper is composed of 5 sections. This

section introduces the research background and the

fundamental technologies applied to the design of

the ESSSA. In section 2, the previous related

research and our current work on event stream

processing is presented. Section 3 discusses about

the main components of the ESSSA in detail. The

fourth section describes the issues related to the

processing event streams in the ESSSA. Finally, we

conclude the paper with a discussion of the proposed

system and current and future implementation issues.

2 RELATED RESEARCH AND

OUR WORK

2.1 Related Research

In handling events, a lot of research efforts have

been made in the last decade to formalize the event,

the behaviour of the events, and the relationship

between them.

The study on event streams was drived from the

concern for handling time-varying data within time-

series management system. Then, the active

database rose to the surface of event-based temporal

reasoning. (Motakis, Zaniolo, 1997). Active

database has been frequently compared with data

stream management system. Active database uses

triggers to support automatic response to events.

The action may change the value of the database,

whereas, the rapid materialization of ubiquitous

computing environments revitalized the events

stream handling issue. Data stream management

system demands complex events processing schemes

to treat inflow of events from the sources and

supports the continuous queries over the running

streams. (Luckham, 2001), (Rizvi, 2005).

Many researchers have analyzed events and

distinguished the nature of the event queries over

event streams in time from the traditional queries on

the relations. (Motakis, Zaniolo, 1997), (Zimmer,

Unland, 1999), (Luckham, 2001), (Arasu, Badu, &

Widom, 2003).

The most important difference is that the event

initiates the event query, while the ordinary query

which is used in the traditional DBMS is self-

activated. (Chandrasekaran, et al, 2003). Time-

varying event query needs to keep the event history

in event repository and sliding window mechanism

to constraint the number of event at a particular

point of time has been chosen as a soultion.

(Ghanem, et al, 2007).

Also, semantic formalization such as basic

definitions and classification of events, their time-

varying relationships, and the events rules have been

accomplished from different perspectives. (Motakis,

Zaniolo, 1997), (Zimmer, Unland, 1999), (Arasu, et

al, 2003), (Chandrasekaran, et al, 2003), (Rizvi,

2005), (Bry, Eckert, 2006), (Bry, Eckert, &

Pătrânjan, 2006).

Many event stream management systems such as

TelegraphCQ (Chandrasekaran, et al, 2003),

STREAM (The STREAM Group, 2003), and REMS

(Hwang, Cheong, Kim, & Lee, 2005) have been

developed as the solutions to the challenges

discussed so far.

ICE-B 2007 - International Conference on e-Business

2.2 Design of the Event Stream Service

System Architecture (ESSSA)

Figure 1: The Event Stream Service System Architecture.

Figure 1 shows four major sub-systems: the

Event Collector, the Event Query Manager, the

Event Archive Manager, and the Service Manager.

The ESSSA considers fundamental issues

described in the previous section and targets to

handle event streams from massive RFID tags,

sensor data of USN, and any other spatiotemporal

information of any moving object.

To satisfy this objective, the ESSSA is to extend

the object-relational spatial DBMS, ZEUS

(KTDATA, 2005), from the lower storage

management level to the higher query processing

level.

3 MAIN COMPONENTS

3.1 Event Collector

At the front-end of the Event Collector is the Access

Gateway, which receives immense amounts of event

streams that have been classified and recognized

through the event pre-processors such as object

naming service (ONS) server, physical markup

language (PML) server, and Savant server as defined

in EPCglobal architecture framework standard

(EPCglobal, 2005). The events are originated from

various RFID readers, sensors in USN, wireless

communication network entities.

The Event Collector also includes the Event

Stream Analyzer, the Event Stream Refiner and the

Event Stream Queue Manager. They categorize

events, input them in the event stream queue, and

filter out duplications among adjacent readers or

meaningless events holding the unchanged value.

The events refined in this level are ready to be

monitored and evaluated in the Event Query

Manager and broadcast to the external service

applications such as enterprise resource planning

system (ERP), supply chain management system

(SCM), warehouse management system (WMS),

tracking system, or centralized command and control

system.

3.2 Event Query Manager

The Event Query Manager processes event queries,

which perform the event rules requested from the

service applications and provide them with the

proper services in time. The Event Query Analyzer,

the Event Rule Manager, the Event Query Optimizer,

and the Event Query Processor are composing the

Event Query Manager.

The Event Query Analyzer parses event query

statements written in the extended structured query

language (SQL) query language and extracts the

event rules and registers them to the appropriate

event monitoring rule tables.

The Event Rule Manager gathers event rules

from event monitoring rule tables which are divided

into Post-Event Monitoring Rule Table and Pre-

Event Monitoring Rule Table.

The Event Query Optimizer makes a plan for the

execution of the complex event query. Finally, Event

Query Processor processes the event query

according to the query plan.

3.3 Event Archive Manager

The Event Archive Manager consists of the Event

Event Storage Manager. The Event Register stores

events coming from the Event Collector to the event

repository.

The Event Triggering Monitor synthetically

monitors registered events, determines if the events

are matched with the predefined triggering rules, and

reports them as query results. The associated

business service applications are then to be informed

of the query results through the Service Manager.

The Event Storage Manager maintains the events

by constructing spatiotemporal index on them

depending on their (symbolic/physical) locations and

time of occurrences, keeps a snapshot of the events

data of the main memory repository, and performs

swapping between the main memory repository and

the secondary backup DBMS.

Event Collector Event Archive Manager

Access Gateway

Event Stream Refiner

Event Register

Event Storage Manager

(Main/Secondary Mem ory)

Secondary

Memory

Main

Memory

Event Triggering Monitor

Channel

Adaptor

Channel

Adaptor

Service Manager

Integrated Service Gateway

Service Management

Database

ERP SCM WMS

Moni toring

System

RFID

Reader

ONS PML

Savant

Channel

Adaptor

Channel

Adaptor

Event Stream Analyzer

RFID

Reader

RFID

Reader

Event Stream Queue Manager

Service

Organizer

Event Query Manager

Event Query Analyzer

Event Query Processor

Event Query Optimizer

SOA-based Open Interface

USN

Wireless

Service

Provider

Rule-based

Event

Handler

Service

Monitor

Event Rule Manager

INTEGRATED DESIGN OF EVENT STREAM SERVICE SYSTEM ARCHITECTURE (ESSSA)

3.4 Service Manager

The Service Manager contains the Integrated Service

Gateway, the Service Organizer, the Rule-based

Event Handler, the Service Provider, and the Service

Monitor. The Integrated Service Gateway allows

external service applications to interact the ESSSA

with SOA-based open interfaces. The Service

Organizer registers services in combination with the

event conditions. The Rule-based Event Handler

transforms registered services into event queries and

requests the associated query results. The Service

Provider provides external business service

applications with web services to broadcast the event

occurrences and their status in real-time. The

Service Monitor controls the enrolled service

conditions and then the ESSSA will identify the

occurrence of the event and provide the related

information through Service Provider.

4 EVENT PROCESSING

This section mainly discusses the key concepts and

major extension details of the ESSSA to effectively

process the event streams. They are composed of

event collecting process, query language extension,

and event querying process.

4.1 Event Collecting Process

In order to handle the massive event streams from

many different origins, this process acquires,

analyzes and refines, and manages event stream

queues by phases.

Figure 2: The Event Collecting Process.

As depicted in Figure 2, this architecture

assumes that events are transmitted from RFID tags,

gathered through ONS, PML and Savant server and

sent to the Access Gateway via Channel Adaptor of

Event Collector. Initially acquired events are

delivered to the Event Stream Analyzer, which

breaks down the tag information into separate

meaningful values such as the classification and the

identification information of the tagged object. The

electronic product codes (EPCs) system for RFID

tag is used to uniquely identify a single object.

Generally, the EPC tag contains a header, the EPC

manager’s information, object class, manufacturer,

product identification information, serial number,

and so on. (EPCglobal, 2005). The Event Analyzer

classifies the events and inputs them into the

separate event stream queues.

The Event Stream Refiner filters out duplications

among adjacent readers or removable events, which

hold the same value as the one of the previous event.

The events refined in this level are ready to be

monitored and evaluated based on the Event Query

Manager according to the event rules. The query

results are then replied to the Service Manager to be

delivered to the external add-on service applications.

In figure 2, events starting with “E1:1” are

duplicated and the Event Stream Refiner eliminates

the last one. In some cases, we should not cut the

duplicate to keep track of the status of the object.

Thus, an event query language (EQL) has to support

capabilities to define these constraints to the rules

depending on the specific characteristics of events.

4.2 Query Language Extension

The extension aims to accommodate characteristics

of events and represent them. Specifically, event

handling operators provides the service applications

with fundamental means to express various event

rules. This section describes the extension details by

presenting simple query statements, which express

event related tasks. The base query language for

extension is from the spatial query language of

object-relation DBMS, called ZEUS. (Park, et al,

1998), (KTDATA, 2005).

The following section describes detailed features

of the query language extension for event processing.

4.2.1 Definition of Event Relation

The syntax for event definition is a variant of the

relations definition in conventional DBMS.

create event PassGate (epcID

char(100), gateID char(15),

location point);

create event StoreToShelf (epcID

char(100), shelfID char(15),

noBox integer);

create event LiftFromShelf (epcID

char(100), shelfID char(15),

noBox integer);

Event Collector

Event

Archive

Manager

Access

Gateway

Event Stream

Refiner

RFID

Reader

ONS PML

Savant

Channel

Adaptor

Event

Stream

Queue

Manager

RFID

Reader

RFID

Reader

Event Stream

Analyzer

E1:1 01 0004BFA GAF2BB 02379DA1F

E1:2 01 0004BFA GAF2BB 02379DA20

E1:3 01 0004BFA GAF2BB 02379DA21

E1:1 01 0004BFA GAF2BB 02379DA1F

Event Queue : E1

E1:3E1:2E1:1 E1:1

E2:1 02 0004BFA GAF2BB 02379DA20

E2:2 02 0004BFA GAF2BB 02379DA21

E2:1 02 0004BFA GAF2BB 02379DA20

Event Queue : E2

E2:1E2:2E2:1

Event Queue : E1

E1:3E1:2E1:1

Event Queue : E2

E2:2E2:1

Main

Memory

ICE-B 2007 - International Conference on e-Business

create event LoadedToTruck (epcID

char(100), truckID char(15),

noBox integer);

In the above examples, the keyword “event” is

specified to alert us that a tuple of an event relation

implicitly contains an “event_time” attribute typed

as “timestamp” to hold the time of the event as well

as the explicitly indicated attributes. Additionally,

the system performs the event-related prerequisite

tasks such as preparing the event queues and

registering it to an anticipating event list.

4.2.2 Relationship Operators between

Events

Event relationship operators are classified into three

groups: temporal relationship, logical relationship,

and aggregation relationship.

Generally, temporal relationship operators

distinguish an event’s temporal order and act

depending on the individual operator. The following

shows the way that temporal operators are used in

“where” clause of traditional “select” query

statement:

select p.epcID, p.event_time,

s.event_time, s.shelfID

from PassGate p, StoreToShelf s

where PassGate happen before

StoreToShelf within

[p.event_time,

p.event_time+10];

select p.epcID, p.event_time,

l.event_time, l.shelfID

from PassGate p, LiftFromShelf l

where PassGate happen after

LiftFromShelf;

The first SQL statement finds the object, which

has passed the gate and is stored on the shelf for up

to 10 minutes. The second one recognizes the object

passing the gate after it has been lifted from the shelf

without any time limitation.

Another type of event operator is to simply

figure out the logical relationship between events

such as conjunction or disjunction. Furthermore,

query language supports negation of an event to

ensure an event does not happen.

select p.epcID, p.event_time,

s.event_time, l.event_time,

l.truckID, l.noBox

from PassGate p, LoadedToTruck l

where PassGate and LoadedToTruck

within [t, t+10];

select s.event_time, s.shelfID

from StoreToShelf s,LiftFromShelf l

where StoreToShelf or not

LiftFromShelf within

[t, t+30];

Moreover, extended query language provides the

aggregate operators, which gather the events of

interest and evaluate the operations: count( ), sum( ),

and average( ).

select count(PassGate),

from PassGate p

where p.epcID like ’01 0004BFA%’;

select sum(l.noBox),

from LoadedToTruck l

where LoadedToTruck within

[t, t+10];

The above two event query statements calculate

the number of objects having tags, which start with

‘01 0004BFA’ and the number of whole boxes

loaded to a truck within a specified time period by

using the aggregate operators.

4.3 Event Querying Process

In the Event Query Manager, the Event Query

Analyzer begins parsing the event query statements

and registers the extracted event rules to the

corresponding event monitoring rule tables, which

are handled by the Event Rule Manager. The event

rules that are supposed to be monitored are

thoroughly examined in order to schedule and

arrange the events within the event monitoring map

table (EMMT).

Table 1: The Event Monitoring Map Table.

As one single event can be related with multiple

event queries, the EMMT is introduced to

synthetically handle the events to be monitored. The

event map is associated with three sub-components

of the Event Query Manager. First, the Event Rule

Manager registers the event and builds up the map,

secondly, the Event Query Optimizer periodically

investigates the correlations between events, and

lastly, the Event Query Processor performs the

actual queries.

Event

Successor

Predecessor

E6, E3

Service ID

S1, S6

S2, S7

S3, S8

S4, S9, S10

INTEGRATED DESIGN OF EVENT STREAM SERVICE SYSTEM ARCHITECTURE (ESSSA)

By coupling the main memory repository scheme

and the disk-based DBMS, we intend to take

advantage of the benefits provided by both systems.

The continuously generated massive events need to

be processed in real time. In addition, the time-

varying results of the event query should be

produced to satisfy the requirements of the various

service applications based on the complex event

queries.

Figure 3: The Event Querying Process.

Figure 3 shows the event querying process in

combination with the sub-components of the ESSSA.

5 CONCLUSIONS

This research results from the increasing

requirements relevant to the emergence of the

innovative sensing technologies related with RFID,

USN, and any other wireless telecommunication

facilities. Consequently, the elaborate handling of

the massive amounts of events are regarded as the

key success factor of the event stream service

framework, which coordinates various event-driven

real world applications.

In this paper, we have designed the ESSSA

based on the object-relational spatial database

management system, ZEUS. We extended the spatial

query language in order to treat the time-varying

nature of the event streams. Furthermore, as an event

stream is related with multiple queries and

enormously increased over time, the efficient event

sharing and the flexible switching mechanism needs

to be thoroughly considered when designing and

implementing the main memory repository.

The event query language suggested in this paper

is to be formalized to guarantee the completeness of

the functionalities. Moreover, the query optimization

procedures of the existing system should be refined

to support the event queries in the future work.

REFERENCES

Arasu, A., Badu, S., & Widom, J., 2003. CQL: A

Language for continuous queries over streams and

relations. In DBPL.

Bry, F., Eckert, M., & Pătrânjan, P., 2006. Querying

composite events for reactivity on the Web. In Proc.

Intl. Workshop on XML Research and Applications,

number 3842 in LNCS, pages 38-47. Springer.

Bry, F., Eckert, M., 2006. A-High-level query language

for events. In Proc. of the IEEE Services Computing

Workshops (SCW’06).

Chandrasekaran, S., Cooper, O., Deshpande, A., Franklin,

M. J., Hellerstein, J. M., Hong, W., Madden, S.,

Raman, V., Reiss, F., & Shah, M., 2003.

TelegraohCQ: Continuous dataflow processing for an

uncertain world. In CIDR 2003, First Biennial

Conference on Innovative Data Systems Research,

Asilomar, CA, USA.

Ghanem, T. M., Hammad, M. A., Mokbel, M. F., Aref, W.

G., & Elmagarmid, A. K., 2007. Incremental

Evaluation of Sliding-Window Queries over Data

Streams. In IEEE Trans. on Knowledge and Data

Engineering, vol. 19, no 1.

Hwang, J. G., Cheong, T. S., Kim, Y. I., & Lee, Y. J.,

2005. Trends of RFID Middleware Technology and Its

Application. In Telecommunication Trend Analysis

and Survey, vol 20. no 3.

Luckham, D., 2001. The Power of Events: An Introduction

to Complex Event Processing in Distributed

Enterprise Systems. Addison-Wesley Longman

Publishing Co., Inc. Boston, MA, USA,

Motakis, I., Zaniolo, C., 1997. Temporal aggregation in

active database rRules. In Proc. Inlt. Conf. on

Management of Data(SIGMOD). ACM Press.

Park, K. E., Lee, J. K., Lee, K. J., Ahn, K. H., Lee, J. W.,

& Kim, J. S., 1998. The development of

ZEUS(GEUS): A spatial DBMS tightly integrated

with an object-relational database engine. In URISA

1998 Annual Conference Proceedings, Charlotte, NC.

Rizvi, S., 2005. Complex event processing beyond active

databases: Streams and uncertainties. Technical

Report No. UCB/EECS-2005-26.

The STREAM Group, 2003. STREAM: The Stanford

stream data manager. In IEEE Data Engineering

Bulletin. vol. 26, no. 1.

Zimmer, D., Unland R., 1999. On the semantics of

complex events in active database management

systems. In Proc. Intl. Conference on Data

Engineering (ICDE), pages 392-399. IEEE Computer

Society Press.

KTDATA, 2005. User Manual of ZEUS DBMS, vol. I, II.,

http://ktdata.co.kr.

EPCglobal, 2005. The EPCglobal Architecture Framework.

From http://www.epcglobalinc.org/standards/Final-

epcglobal-arch-20050701.pdf.

Event Collector

Event Queue : E1

E1:3E1:2

E1:1

Event Queue : E2

E2:2E2:1

Event