Modeling Structural, Temporal and Behavioral Features of a Real-Time
Database
Nada Louati
1
, Rafik Bouaziz
1
, Claude Duvallet
2
and Bruno Sadeg
2
1
MIRACL-ISIMS, Sfax University, BP 1088, 3018, Sfax, Tunisia
2
LITIS, UFR des Sciences et Techniques, BP 540, 76 058, Le Havre Cedex, France
Keywords:
Real-time, Database, MARTE, UML-RTDB, Profile.
Abstract:
Real-time databases are different from traditional databases in that they have timing constraints on data and
on transactions upon the data. The design of this kind of databases must consider both temporal aspects of
data and timing constraint of transactions in addition to the logical constraints of the database. This paper
proposes a new UML profile that extends UML with concepts related to real-time databases design. These
extensions aim to accomplish the conceptual modeling of a real-time database according to three aspects:
structural, temporal and behavioral. Our propositions is based on MARTE (Modeling and Analysis of Real-
Time and Embedded systems) profile which provides capabilities of modeling concepts to deal with real-time
and embedded system features. The proposed profile is illustrated by a case study in the context of Ait Traffic
Control System.
1 INTRODUCTION
Real-Time DataBases (RTDBs) are typically used to
manage environmental data in computer control ap-
plications, such as air traffic control, automated man-
ufacturing, and military command and control (Ra-
mamritham, 1993). The design of such RTDBs dif-
fers from both that of real-time systems and that of
conventional database management systems. The de-
signers of RTDBs much consider both temporal as-
pects of data and timing constraints of transactions in
addition to logical constraints of the database
The design of RTDBs is performance-and
semantic-dependent. It must consider factors such as
sensor data, derived data and quality of data manage-
ment, temporal semantics in transactions scheduling
algorithmsand concurrencycontrol protocols, to meet
the timing constraints defined by the real-time appli-
cations (Idoudi et al., 2010). This paper proposes a
new UML profile that incorporates these concepts. It
is based on a subset of concepts inspired from the
HLAM, NFP, and TIME packages of MARTE. The
motivations behind these extensions are three-folds:
(i) to have new notations distinguishing quantitative
and qualitative features of RTDBs, (ii) to facilitate
the modeling of timing aspects of data and transac-
tions, (iii) to accomplish the conceptual modeling of
RTDBs. This profile not only captures the structural
aspects of a RTDB features, but also the temporal and
behavioral aspects.
The remainder of this paper is structured as fol-
lows: in Section 2, we present the related works. In
Section 3, 4, and 5 we respectivelyaddress the model-
ing of structural, temporal, and behavioral aspects of
RTDBs. In Section 6, we present a case study to more
illustrate our proposals. In Section 7 we conclude the
paper and we give some perspectives of our work.
2 RELATED WORK
In recent years, several UML profiles have been pro-
posed for modeling real-time systems to depict its
real-time constraints. Only a few of these UML pro-
files address RTDBs modeling.
In (DiPippo and Ma, 2000), DiPippo and Ma de-
scribe an UML package, called RT-Object, for spec-
ifying RTSORAC object. This UML package con-
tains real-time attributes, real-time methods, compat-
ibility functions and constraints. The RT-Object pack-
age is based on a past standard of UML which is
UML 1.3 Extension Mechanisms package. Further-
more, imprecise computation encapsulated within the
RTSORAC object model is described in the context of
Epsilon Serializability (Ramamritham and Pu, 1995),
119
Louati N., Bouaziz R., Duvallet C. and Sadeg B..
Modeling Structural, Temporal and Behavioral Features of a Real-Time Database.
DOI: 10.5220/0003975601190125
In Proceedings of the 14th International Conference on Enterprise Information Systems (ICEIS-2012), pages 119-125
ISBN: 978-989-8565-10-5
Copyright
c
2012 SCITEPRESS (Science and Technology Publications, Lda.)
Figure 1: HLAM package specification in MARTE.
and does not support the notion of quality of data in-
troduced in (Amirijoo et al., 2006).
Another work proposed the UML-RTDB profile
(Idoudi et al., 2008a) (Idoudi et al., 2008b) to ex-
press RTDB features in a structural model. It sup-
plies concepts for RTDBs modeling such as real-time
attributes, real-time methods and real-time classes.
However, the proposed extensions are based on the
Evolutionary Stereotype concept (Debnath et al.,
2003) which is not a standard way of extending UML.
MARTE (OMG, 2009) is an industry standard of
the OMG for model-driven development of embed-
ded systems (OMG, 2009). It aims at providing sup-
port for specification, design, validation, and verifi-
cation stages in real-time and embedded system de-
velopment. The richness of MARTE profile in terms
of concepts offers an interesting common modeling
basis to adequately specify many design features of
RTDBs. However, the notions of real-time data and
real-time transactions, which are the basis features of
RTDBs are missing in MARTE.
3 MODELING OF STRUCTURAL
ASPECTS
3.1 High Level Application Modeling in
MARTE
The HLAM package of MARTE proposes concepts
that enable to describe both quantitative and quali-
tative features of real-time applications at a high ab-
straction level. Figure 1 depicts the basic stereotypes
associated with the HLAM package. An RtUnit may
be seen as an autonomous execution resource that
owns one or more schedulable resources to handle in-
coming messages. RtUnits may provide real-time ser-
vices. The RtService stereotype has been introduced
for that purpose. The RtFeature stereotype is used to
annotate model elements with real-time features ac-
cording to set of RtSpecification associated with this
stereotype. It can be attached to multiple kinds of
modeling elements (i.e., behavioral features, actions,
messages, signals, and ports). For a full description
of all these stereotypes, the reader may refer to the
specification document of MARTE (OMG, 2009).
Figure 2: RTDB Profile Metamodel.
3.2 Modeling of RTDB Structural
Features
We propose new stereotypes expressing RTDB fea-
tures in a structural model on the one hand, and ac-
cording to the MARTE profile on the other hand. Fig-
ure 2 shows the extensions proposed to some meta-
classes belonging to the class diagram. In order to
take advantages of some MARTE concepts, the pro-
posed profile imports stereotypes from HLAM, NFP,
and VSL sub-profiles.
3.2.1 Stereotypes for Real-Time Data
Real-time attributes are divided into two types: sensor
attributes and derived attributes (Ramamritham et al.,
2004). Sensor attributes are used to store sensor data
which are issued from sensors. Derived attributes are
used to store derived data which are calculated from
sensor data. Thus, we define two stereotypes, sensor
and derived, in order to declare respectively sensor
and derived attributes in the class diagram. As illus-
trated in Figure 2, we define an abstract stereotype,
called realTimeProperty, that factorizes validity du-
ration property which indicates the amount of time
ICEIS2012-14thInternationalConferenceonEnterpriseInformationSystems
120
during which the attribute value is considered valid.
We use the MARTE NFP Duration datatype as a type
for the validity duration feature.
We characterize each sensor stereotype by two
properties : maximum data error and periodicUpdate.
Maximum data error indicates the maximum devia-
tion tolerated between the current attribute value and
the updated value. It is of type NFP Real. Maximum
data error represents a non-functional property spec-
ifying the upper bound of the error. We propose to
associate the Nfp stereotype of MARTE profile to the
maximum data error attribute. PeriodicUpdate refers
to a periodic operation of the class that owns the sen-
sor attribute. The role of this periodic operation is to
update the current value and the timeStamp fields of
the sensor attribute. It is of type Operation.
As shown in Figure 2, we characterize derived
stereotype by sporadicUpdate property, that refers to
a sporadic operation of the class that owns the derived
attribute. This sporadic operation is used to update the
current value and the timeStamp fields of the derived
attribute.
Each real-time attribute value is characterized by
a timestamp, which indicates the time at which it was
last updated. So, for each real-time attribute value
corresponds a timestamp, which distinguishes it from
other attribute’s values. Whereas, the values of the
validity duration and maximum data error fields are
the same for the same real-time attribute.
3.2.2 Stereotypes for Real-Time Transactions
A real-time transactions may be aperiodic, periodic,
or sporadic (Ramamritham et al., 2004). It has timing
constraints such as deadline and period. We define
three stereotypes, aperiodic, periodic, and sporadic
(cf. Figure 2) in order to declare respectively ape-
riodic, periodic, and sporadic operations in the class
diagram. Each of these stereotypes is characterized
by a deadline, which indicates the last time by which
the method execution must be completed. Thereby,
we define an abstract stereotype, called realTimeOp-
eration, which is used to annotate model elements
(i.e. Operations) with real-time features according to
set of RtSpecification associated with this stereotype.
RtSpecification possesses nine tagged values among
which: relDl, occKind, and priority. The relDl at-
tribute specifies the deadline of a method execution.
The occKind attribute indicates the arrival transaction
specification (i.e. periodic, aperiodic, or sporadic).
For a periodic (respectively aperiodic and sporadic)
transaction, the PeriodicPattern (respectively Aperi-
odicPattern and SporadicPattern) enumeration literal
is selected for ArrivalPattern attribute of the occKind
property. The priority attribute specifies the priority
Figure 3: Time Specification in MARTE.
order of a transaction. The realTimeOperation stereo-
type factorizes also two properties: concPolicy and
isAtomic. When the value of isAtomic is true, the
method execution is done as one individual unit. This
fact coincides with the atomicity property of a trans-
action. However, in RTDBs, this feature is relaxed in
order to allow the validation of a transaction even if
only a part of its actions have been executed (Agrawal
et al., 1994). So, in our work, we consider that the
value of the isAtomic property is always false. The
concPolicy property specifies the concurrency policy
of a transaction. The values of this property may be:
reader, writer or parallel. A writer transaction im-
plies that multiple calls from concurrent transactions
may occur simultaneously and will be treated as soon
as concurrency on data allows its execution. A reader
transaction implies that multiple calls from concur-
rent transactions may occur simultaneously and will
be executed simultaneously if there is no writer trans-
action using one or more data that the reader trans-
action needs. A parallel transaction is a transaction
whose actions do not use any data of the database in
reading mode nor in writing mode. In our work, we
consider that for an update transaction, which can pe-
riodic or sporadic, the concPolicy property is writer.
3.2.3 Real-Time Class Stereotype
The design of a RTDB, which is by definition a
database system, has to take into account the manage-
ment of many components such as queries, schemas,
transactions, commit protocols, concurrency control
protocol, and storage management (Stankovic et al.,
1999). In order to deal with time-constrained data,
time-constrained operations, parallelism, and concur-
rency property inherent to RTDBs, we introduced the
RealTimeClass stereotype. This stereotype specifies
that instances of a class will encapsulate real-time
data and real-time operations and a local concurrency
mechanism. Because of the dynamicnature of the real
world, more than one transaction may send requests
to the same object. Concurrent execution of these
transactions allows several methods to run concur-
rently within the same object. To handle this essential
property of RTDB systems, we associate to each ob-
ject a local concurrency control mechanism, named
local controller, that manages the concurrent execu-
ModelingStructural,TemporalandBehavioralFeaturesofaReal-TimeDatabase
121
tion of its methods. Thus, the object receives mes-
sages in its mailbox awaking its local controller that
checks the timing constraint attached to messages and
selects one message following a special scheduling al-
gorithm. The local controller verifies the concurrency
constraints with the already running methods of the
object. Then, it allocates a new thread to handle the
message when possible. When a method terminates
its execution, the corresponding thread is released and
concurrency constraints are relaxed. In our work, the
scheduling algorithm adopted by the mailbox is EDF
(EarliestDeadlineFirst). For that purpose, we import
from MARTE library the SchedPolicyKind enumera-
tion which defines the most common kind of schedul-
ing policy. Add to that, the concurrency control pro-
tocol adopted by the local controller is 2PL-HP (Two
Phase Locking-High Priority). In 2PL-HP, all data
conflicts are immediately resolved by aborting lower
priority transactions. Thereby, we enhance our profile
model library by a new enumeration type, called Con-
currencyControlKind, in order to define concurrency
control policies (i.e. 2PL-HP, PCP (Priority Ceiling
Protocol), SCC (Speculative Concurrency Control),
etc.). Our RealTimeClass stereotype overlaps with the
UML extensions presented by MARTE profile espe-
cially those relative to the mailbox and the local con-
troller. In fact, a RealTimeClass may be considered
as an autonomous execution resource, able to handle
different messages at the same time. It can manage
concurrency and real-time constraints attached to in-
coming messages. This has the same meaning as the
RTUnit stereotype defined in MARTE (cf. Figure 1).
Thus, we consider that the RealTimeClass concept as
a specialization of MARTE RtUnit concept. Thereby,
a RealTimeClass is a real-time unit but with a RTDB
modeling semantics. It represents the RTDB entity
which encapsulates time-constrained data and time-
constrained operations. It also deals with parallelism
and concurrency features.
4 MODELING OF TEMPORAL
ASPECTS
4.1 Time Specification in MARTE
The sub-profile TIME of MARTE has been intro-
duced to model timing aspects. Figure 3 depicts the
MARTE extension about time modeling in UML. The
Clock stereotype is a model element that represents
an instance of ClockType. The ClockType stereotype
is related indirectly to the Clock stereotype. Its prop-
erties specify the kind of clock (chronometric or log-
Figure 4: Clocks Specification in our RTDB Profile.
ical), the nature (dense or discrete) of the represented
time, a set of clock properties (e.g., resolution, max-
imal value, etc.) and a set of accepted time units.
The ClockConstraint is a stereotype of the UML Con-
straint concept. The clock constraints are used to
specify the time structure relations of a time domain.
A ClockConstraint is a constraint that imposes depen-
dency between clocks or between clock types.
4.2 Modeling of RTDB Temporal
Properties
Time is present in different RTDB features and more
specially in real-time data. Real-time data are divided
into two classes: sensor data and derived data. As pre-
viously mentioned in Section 3.2.1, we have defined
for each sensor data a periodic method (i.e. period-
icUpdate) which is periodically executed to update
the values of the corresponding sensor data. In the
same way, we have defined for each derived data a
sporadic method (i.e. sporadicUpdate) which is spo-
radically executed to update the values of the corre-
sponding derived data. In this section, we define for
the sensor and derived data clocks as well as their as-
sociated properties in a high level specification using
the TIME sub-profile of MARTE. The temporal unit
that we are looking for can be used in two ways: (i) to
reference the physical time and adopt a chronometric
clock for sensor data, (ii) to reference a logical time
that is incremented each time sensor data have been
updated.
The MARTE TimeLibrary provides a model for
the ideal time used in physical laws: idealClk, which
ICEIS2012-14thInternationalConferenceonEnterpriseInformationSystems
122
is an instance of the class IdealClock, stereotyped by
ClockType. The IdealClock represents the time evo-
lution. This time is expressed in seconds in the inter-
national system units. Starting with idealClk, we de-
fine new discrete chronometric clock (cf. Figure 4).
First, we specifies Chronometric (a class stereotyped
by ClockType) which is discrete, not logical (therefore
chronometric), and with a read only attribute (resolu-
tion). Clocks belong to timed domains. In Figure 4, a
single time domain is considered. It owns two clocks:
idealClk and sensorClk, an instance of Chronometric
that both uses the second (s) as a time unit; and whose
resolution is 0.01 s. The two clocks are a priori in-
dependent. A clock constraint specifies relationships
among them. The first statement of the constraint de-
fines a clock c local to the constraint. C is a discrete
time clock derived from idealClk by a discretization
relation. The resolution of this clock is 1 ms. The
next statement specify that sensorClk is subclock of
c with a rate p times slower than c. The fourth state-
ment indicate that sensorClk is not a perfect clock.
Flaws are characterized by non functional properties
like stability and offset. Its rate may have small vari-
ations (a stability of 10
5
implicitly measured on ide-
alClk). The last statement claims that the clocks is
out of phase, with an offset value between 0 and 5 ms
measured on idealClk.
Figure 4 depicts the definition of a logical clock
dedicated for derived data. We have created a class
DerivedDataClock with ClockType as stereotype.This
class has three attributes: maximalValue, offset and
resolution. MaximalValue specifies the maximal
value of the associated clock, value at which the clock
rolls over. The offSet property determines the initial
instant of the associated clock. The resolution at-
tribute defines the resolution of the associated clock.
Now that we have defined the clock, we need to in-
stantiate it. Figure 4 depicts the instantiation of the
clock DerivedDataClock called derivedClk. The unit
of DerivedClk is the sensor data update.
Sensor data are periodically updated causing the
update of each derived data that uses those sensor
data. This is equivalent to say that the derived data up-
date operation is activated everytick of the clock asso-
ciated with sensor data. We can deduce an affine rela-
tion between sensorClk and derivedClk as specified in
the clock constraint specified in Figure 4: derivedClk
isPeriodicOn sensorClk, period = pp, offset=t. Such
a constraint states that each pp
th
occurrence of sen-
sorClk there will be an occurrence of derivedClk and
the first occurrence of derivedClk appears at the t
th
instant of sensorClk.
5 MODELING OF BEHAVIORAL
ASPECTS
5.1 Behavior Specification in MARTE
State machines are often used in the real-time and
embedded domain. They allow the description of a
system behavior in terms of states and transitions be-
tween these states. MARTE introduces mainly two
time-related concepts that can be used to improve the
usage of the UML behaviors, such as state machines,
for developing real-time applications: timed process-
ing and timed events (cf. Figure 3). The TimedPro-
cessing stereotype enables modelers to specify dura-
tion for a behavior. The TimedProcessing stereotype
represents activities that have known start and finish
times or a known duration, and whose instants and du-
rations are explicitly bound to clocks. It can also ref-
erence events triggered when a behavior or processing
starts and ends. The duration can be specified using
the VSL language, which supports time expressions.
The TimedEvent stereotype represents events
whose occurrences are explicitly bound to a Clock.
It extends the TimeEvent concept of UML. It allows
to characterize the logical or physical clock on which
a time event relates.
5.2 Modeling of RTDB Behavior
A RTDB is a collection of real-time objects which are
used to manage time-critical dynamic systems in the
real world. RTDB behavior model building consist of
describing the behavior with state-transition diagram
for each real-time object of the RTDB. Stereotypes
TimedProcessing and TimedEvent are used to specify
behavior, duration of a behavior, and events bounded
to a clock. The tagged value queueSchedPolicy, de-
fined in MARTE, are associated to the state-transition
diagram in order to indicate the queue scheduling pol-
icy of a behavior. We apply the TimedProcessing
stereotype on the state machine itself and we spec-
ify the corresponding clock using the meta-attribute
on. We employ the TimedEvent stereotype in order to
characterize the logical and physical clocks on which
a time event relates.
6 CASE STUDY
Throughout, this section, we will use a running exam-
ple to illustrate the use of our profile. We illustrate
our proposal on an air traffic control system. The
aim of the air traffic control is to separate aircrafts,
ModelingStructural,TemporalandBehavioralFeaturesofaReal-TimeDatabase
123
Figure 5: Aircaft real-time class.
Figure 6: Aircraft behavior modeling.
to avoid collisions and to organize the flow of traf-
fic. It consists of a large collection of data describ-
ing the aircrafts, their flight plans, and data that re-
flect the current state of the controlled environment
(Locke, 2001). This includes flight information, such
as aircraft identification, speed, altitude, origin, des-
tination, route and clearances. In fact, each aircraft
has three sensor data which are speed, altitude and
location and two derived data which are path and
lane. Sensor data are periodically updated to reflect
the state of an aircraft. The derived data is calcu-
lated based on altitude, location, and direction val-
ues, in order to verify if the aircraft deviates from a
predetermined path or lane. All these data values are
published periodically by sensors supervising the air-
crafts controlled elements.
For sake of clarity, only the Aircraft class and
a subset of its methods are specified. Each Air-
craft in the airspace is stereotyped by realTimeClass.
We characterize the Aircraft class by three sensor at-
tributes: direction, location, and altitude. Each at-
tribute is periodically updated in order to closely re-
flect the real world state of the application environ-
ment. Thereby, we associate to each sensor attribute
a periodic method: updateAltitude(), which periodi-
cally updates the value and the timestamp of the alti-
tude. We characterize also the Aircraft class by two
derived attributes: lane which is calculated from loca-
tion and altitude values, and path, which is calculated
from location and direction values. Each derived at-
tribute has its own sporadic method: computePath(),
which sporadically updates the value and the times-
tamp of the path. Figure 5 depicts a simplified view of
the Aircraft class. We indicates that the Aircraft class
is the main unit of the application (property isMain is
set to true) and it starts a main rtService, called start().
Its isDynamic property is set to true. In this case, the
schedulable resources are created dynamically when
required. The attribute altitude has a validity dura-
tion equal to (3, s) and its maximumDataError is 5.
Additionally, it is updated periodically using the up-
dateAltitude() operation, which has a period equal to
(3, s). The concPolicy attribute of that operation is
writer and its isAtomic attribute is set to false. In
the same way, we characterize the derived attribute
path by a validityDuration and a maximumDataError
and it is sporadically updated by means of the com-
putePath() operation.
Figure 6 depicts a state-machine diagram that pro-
vides a simplified view of the Aircraft behavior. The
Aircraft has four states: TakeOff, Flying, Update, and
Landing. Periodically , it enters in the Update state
for updating the Aircraft sensors values. The sensors
values update has to be done with a period of 5 s and
lasts 2 s. Then it returns to the state which activated
the update transition. We apply the TimedProcessing
stereotype on the state machine itself. We use the for-
mer to indicate the scheduling policy associated with
the Aircraft behavior (i.e EDF). The latter is used to
assign the sensorClk clock to the model. All the el-
ements of the state machine have the same time ref-
erence. We apply the TimedEvent stereotype on the
UML timed event that triggers the UpdateTimeOut
transition.
ICEIS2012-14thInternationalConferenceonEnterpriseInformationSystems
124
7 CONCLUSIONS
The richness of MARTE profile in terms of con-
cepts offers an interesting common modeling basis
to adequately specify many design features of RT-
DBs. In This paper, we showed how the concepts
of the MARTE standard profile can serve to model
RTDB features. We used the HLAM package to de-
scribe both quantitativeand qualitativeproperties, and
the TIME sub-profile to specify temporal properties.
Moreover, we proposed UML/MARTE-based exten-
sions for a complete and powerful RTDB modeling.
Additionally, the proposed extensions not only cap-
ture the structural aspects of RTDB features, but also
the temporal and behavioral aspects. We have in-
vested some efforton ensuring that all concepts havea
well defined basis semantics. This has been illustrated
on a case study in the context of Air Traffic Control
System.
We are currently working on the the integration
of the RTDB development process in the context of
a Model Driven Architecture. This way, the complex
task of designing the whole RTDB is tackled in a sys-
tematic, well-structured and standard manner.
REFERENCES
Agrawal, D., Bruno, J. L., Abbadi, A. E., Krishnaswamy,
V., El, A., and Krishnaswamy, A. V. (1994). Relative
serializability: An approach for relaxing the atomicity
of transactions. In In Proceedings of the 13th ACM
Symposium on Principles of Database Systems, pages
139–149. ACM.
Amirijoo, M., Hansson, J., and Son, S. H. (2006). Specifi-
cation and management of qos in real-time databases
supporting imprecise computations. IEEE Trans.
Computers, 55(3):304–319.
Debnath, N., Riesco, D., Montejano, G., Grumelli, A., Mac-
cio, A., and Martellotto, P. (2003). Definition of a
new kind of uml stereotype based on omg metamodel.
In Computer Systems and Applications, 2003. Book
of Abstracts. ACS/IEEE International Conference on,
pages 49–54.
DiPippo, L. C. and Ma, L. (2000). A uml package for spec-
ifying real-time objects. Comput. Stand. Interfaces,
22:307–321.
Idoudi, N., Duvallet, C., Sadeg, B., Bouaziz, R., and
Gargouri, F. (2008a). Structural model of real-time
databases. In ICEIS (3-2), pages 319–324.
Idoudi, N., Duvallet, C., Sadeg, B., Bouaziz, R., and
Gargouri, F. (2008b). Structural model of real-time
databases: An illustration. In ISORC, pages 58–65.
Idoudi, N., Louati, N., Duvallet, C., Sadeg, B., Bouaziz, R.,
and Gargouri, F. (2010). A framework to model real-
time databases. International Journal of Computing
and Information Sciences (IJCIS), 7(1):1–11.
Locke, C. D. (2001). Applications and system characteris-
tics. In Real-Time Database Systems, pages 17–26.
OMG (2009). A UML Profile for MARTE: Modeling and
Analysis of Real-Time Embedded systems, version
1.0, formal/2009-11-02.
Ramamritham, K. (1993). Real-time databases. Distributed
and Parallel Databases, 1(2):199–226.
Ramamritham, K. and Pu, C. (1995). A formal characteriza-
tion of epsilon serializability. IEEE Trans. on Knowl.
and Data Eng., 7:997–1007.
Ramamritham, K., Son, S. H., and DiPippo, L. C. (2004).
Real-time databases and data services. Real-Time Sys-
tems, 28(2-3):179–215.
Stankovic, J. A., Son, S. H., and Hansson, J. (1999). Mis-
conceptions about real-time databases. IEEE Com-
puter, 32(6):29–36.
ModelingStructural,TemporalandBehavioralFeaturesofaReal-TimeDatabase
125