system scad

7/27/2019 SYSTEM SCAD

1/6

IEEE Transactions on Power Delivery, Vol. 11, No 3 , July 1996 1195THE ROLE OF MEDIUM ACCESS CONTROL PROTOCOLS IN SCADA

SYSTEMS.

Joaquin Luque and Isabel G6mezDepartment of Electronic TechnologyUniversity of Seville.

Absb-acc- This paper deals with the influence of the mediumaccess control (MAC) protocol in the performance of thecoriiriiunication channels used in the SCADA system(Supervisory Control And Data Acquisition). Three types ofMAC protocols are presented and analyzed: Polling,CSMAXD (Carrier Sense Multiple Access with CollisionDetection), and Token Bus. The simulation of each one ofthese protocols allows us to quantify its performance andestablish cer bi n criteria in order to be able to coinpare them.As a result of the analysis, the use of a MAC protocol isproposed for SCADA systems.

I, INTRODUCTIONThe use of a SCADA system for operating powernetworks is supported by a distributed architecture, withdozens or hund reds of comp uters, called R emote TerminalUnits (RTUs). These RTUs communicate with a controlcenter, transm itting the rneasures and the state of the powernetwork, accepting the conm ands which have to be carriedout within the system. Taking into account the largeamount of equipment and their geographical dispersion, thecommunication with the control center usually usesmultiplexed lines whe re narrowband channels are available.

Fo r this reason, th e lines used between the center and eachRTU are of low speed, running typically at 600 o r 1200bps (bits per second).On the other hand, in order to lower the totalnumber of required channels, several RTUs are usuallyconnected to each line, sharing the channels resources.This multipoint configuration (fig. 1) means that the

CONTROLCENTER

I I

Fig. 1. SCADA system architecture.

communication protocol which is being used, Gcludesprocedures which permits the coordination of th e access tothe physical medium which is being used. T hese proceduresare called Medium Access Control (MAC), and thecorresponding protocol is called a M A C protocol. Thereare a large number of protocols used in SCA DA systems.Many of them have been developed by the systemmanufacturers, some by the users, and only a smallpercentage of them correspond to the standardizing efforts.However, there is a general trend towards consid ering theprotocols as a seven layer structure following the OS1model [l ]where, in the case of power network protocols,one or more of these layers are empty. In figure 2, th eseven layer OS1 model is presented and compared to theEPA model (Enhanced Protocol Architecture) specificallyproposed by the IE C (International ElectrotechnicalCommission) [2] for the telecontrol of power networks. Inboth approaches, the control procedures for the MediumAccess Control are in the lower sublayer of the Data LinkLevel (layer 2).

96 WM 060-4 PWRD A paper recommended and approved by the IEEEPower System Comm unications Committee of th e IEEE Power EngineeringSociety for presentatlon at the 1996 IEEE/PES Winter Meeting, January 21-25, 1996, Baltimore, h.1D Manuscript submitted July 24, 1995; made diff eren t techniques can be

In order to reach mo re efficientSCADA protocols,the previously mentioned layers In other papers [3 ,4] w ehave presented how , studylng the evolution of the measuresin power networks, they can be transmitted in acompressed message, and so increasing the efficiency ofthe protocol. Also the Medium Access Control protocol canbe improved. During the next pages, well show how thisobjective can be achieved.

affecting one ofavailable fo r printing November 13, 1995.

0885-8977/96/$05.000 996IEEE


2/6

1196

Reservation

os1I Applicat ion I

Many

layerPresentat ion

Sess lonlayer

Network

PAApplicat ionlayer

Data linklayerPhysicallayer

11

Fig. 2. OS1 and EP A models.

11. THE MEDIUM ACCESS CONTROL PROTOCOLS.There are many Medium Access Controlprotocols, which m akes difficult the task of selecting the set

of protocols to be analyzed. However, the MAC protocolscan be classified into categories, presenting a similarbehavior those protocols belonging to the same class. Forwhich, instead of studying some certain protocols, we willanalyze some classes of protocols, which will, in tum, letus simplify the problem. In this way, the organization ofthe MAC protocols into three classes is classic (table l ) ,depending 011 the technique used: selection protocols,contention protocol and reservation protocols. Most ofthese protocols have been significantly developed due to thegrowth of Local Area Networks (LAN). The problems ofthe Medium Access Control which are found in the LANsare very similar to those which can be found in amultipoint line in a SCADA system, which is why we canmake good use of the results and experiences from theLANs field.

T a b l e 1.-RlAC Techniques.

TECHNIQUESelect on Token B usToken RingContent ion

IAlohaCSMACSMA/CD

Th e MA C protocols using the selection techniquesare based on the courtesy, in othe r wo rds, th e transmissionturn goes, in an ordered way, around the set of computersconnected to the same physical medium. W e can definethree main classes of selection protocols: polling, tokenpassing with a bus topology, and token passing with a ringtopology. We will look at the two first protocols later inmore detail. On the other hand, the token ring protocoldoesnt seem to be interesting for SCADA systems, seeingthat it supposes ring topologies with physical features w hichcan be found mainly in LANs.

MAC protocols using contention techniques arebased on the rule of the the strongest one. The computersconnected to the same physical medium fight betweenthemselves for transmission, producing collisions whic h, ofcourse, supposes the destruction of the messages. In thepast, a vast number of protocols of this type have beendescribed. Some of the most significant contentionprotocols are the Aloha, the CSMA and the CSMAICD.The last one is the one which presents best performanceand, what is more, it has go& lot of acceptance due to itsadoption in Ethernet LANs which ha s been standardized asthe I EEE802.3 [5].Which is why it will also be one of theprotocols chosen to be analyzed when used in SCADAsystems.

Lastly, MAC protocols with the reservationtechnique are based on splitting the transmission mediuminto different sub-channels, using physical mechanisms(Frequency Division Multiplexing, Time DivisionMultiplexing, etc.), o r logical ones. Each one of these sub-channels is used to establish point to point virtualconnections between the computers within the networks.The wide diversity of solutions which are grouped underthe name of reservation techniques, makes the study of ageneric type of protocol of this class, impossible. In orderto obtain valid consequences we would have to analyzeeach individual protocol. Additionally, this typ e of protocoldoesnt offer any significant benefit in any other area inwhich i t ha s been employed, its use being very scarce inLA N s , which as we said abo ve, will be used as a referencepoint.

For the reasons mentioned, o f all the many groupsof MAC protocols, we have selected three for analysis:pol l ing , CSMA/CD, and token bus. During the next fewpages, w ell describe its functioning and its application toS C A D A systems.


3/6

1197medium, two or more nodes could t ransmit a t the sametime, prod ucing a collision of their messages. When a nodedetects this situation (CD: Collision Detection), i t abor ts itstransnlission and waits for a random period of time until ittries the transmission again. Figure 4 represents thebehavior of this protocol, whose performance has beenwidely studied by many authors [7 ] [8 ] 191. On the oppositeto the polling protocol, the CSMA/CD doesnt use thechannel for the continuous questioning from the controlcenter, and so the efficiency is, in general, a lot higher,with typical values of between 80 to 90% .

111. TH E POLLING PROTOCOLTh e polling protocol is, maybe, the most simple ofthe MAC protocols and, without a doubt, the most widely

used in SCADA systems. For which, in manycircumstances, i t is the present situation starting fromwhich we can make improvements.In many SCADA systems this protocol takes aquestion-answ er form. T he control center polls the firstremote on the link; if the remote has inforination to send,

i t does so, an d if not, i t sends a null message. The controlcenter goes on the poll the second remote and so onsuccessively until i t has contacted all the RTUs on th e link,at which tim e it starts over again with the first RTU. Thisdialogue of questions and answers can be represented by atemporal diagram, such as the one in figure 3.

This protocols perform ance will be determined bya large num ber of parameters between which w e have jus tpointed out, the number of RTUs, the transmission speed,and tlic commuting delay (going from receiving totransmitting states). A detailed study of this performance ispresented in [6]. However, without having to make anycomp utation, we can see the wasting of channel time whichsupposes the co ntinuous questioning from the control centerto each RTU. This makes, in general, the efficiency quitelow, typically between 60-70%.

IV . THE CSMA/CD PROTOCOL.A completely different approach presents the

CSMA/CD protocol where, as we said before, the accessto the physical medium isnt done in an ordered waythrough a contention between the different RTUs. In thisprotocol, each node (control center and RTUs), whenwanting to transmit, checks the chaiuiel state and doesntstart the transmission until its free (CS: Carrier Sense).However due to the propagation delay through the physical

Polling Cyc l o wilhntdl rcspansa Polling Cydo wilhnon-null ronwnno

I

Fig. 3 . Tcmporal dingrnin of tlic polling protocol

Th e CSMA/CD protocol prescnts , from our pointof view, two inconveniences. Firstly, as i t was describedbefore, the transmission process includes a random processwhich makes the transmission delay of a message to beunbounded. In this way i t is impossible to deternune thetime of arrival of a certain message. T his unbou nded delaymakes the CSMAlCD protocol inappropriated for use inreal time systems, where the answering times must beguaranteed. However some variants of this protocol havebeen described which , withou t altering its performance toomuch, present a bounded delay [lo].

Th e second inconvenience of the CS MA/CD isaround the necessity of having a collision detectiontechnique. In general, this can be carried out without toomuch difficulty in certain physical media, as for example,the coaxial cables used in Ethernet LANs. However, theuse of these techniques in radio channels or in transmissionsystems which use the power lines, causes difficulties due,in part, to the different strengths of the received signals,and the present noise level. For this reason, when this typeof physical media is used, which is the case in most of the

nESTING

.Fig. 4 . Bchavior of thc C S M A i C D protocol .


4/6

1198SCA DA systems, w e must use another protocol, within the

same class, but without collision detection: the CSMA.Unfortunately, this protocol implies that, each time acollision is produced, the channel is wasted during thewhole transmission time and not only the time of collisiondetection. This makes the performance of the CSMA quiteinferior to those of the CSMA/CD.

V. T H E T O KE N BU S P R O T O C O LThe token bus protocol uses the same basicprinciples as the polling protocol, seeing that both belong

to the class of MAC protocols using selection techniques.This implies the absence of collisions. In this paper theclassic token bus protocol has been adapted to thepeculiarities of the SCA DA systems. In the following, thisadaption is described and analyzed. In the token busprotocol the nodes are ordered following a circularsequence which begins at the control center, goes throughall the RT Us, then returns to the control center. T he cyclebegins with the transmission of a message (either with orwithout information) from the control center to any RTU.This message serves as a token and after having beenreceived by the first RTU of the cycle, makes it send amessage (with or w ithout information) to the control centeror any other RTU. This second message serves as a newtoken which after been received by the second RT U in thecycle, starts up its corresponding transmission. Thisprocess is executed successively until again reaching thecontrol center, at which point i t goes back to start thecycle.

As can be seen , the described protocol is nothingmore than a polling protocol in which the answer from aRTU is used as well as a question to the following RT U inthe cycle. So the performance of this protocol can becalculated through sim ple adaptation s of those presented in[6]. In any case, and w ithout the need for any compu tation,it's clear that the performance offered by this protocol issimply superio r to that of the polling protocol, seeing thatall the process of continuous questioning from the controlcenter is eliminated. This gives i t a level of efficiency ofbetween 80 to 90 96.

The token bus protocol works very well in steadyconditions, but presents serious problems in abnormalsituations. The error in a token (message), the fault of aRTU or the inclus ion of a new RTU into the cycle breaksthe normal working dynamics of the protocol and needssome procedures for solving the abnormal condition. Ingeneral, these procedures apply contention techniques(CSM A-like) whic h, howev er, do n't seriously affect theoverall performance, seeing that these disturbed situations

ar e only produced every once in a while. Shown in thisway the complexity of the token bus protocol is clear,seeing that it needs a double technique: token passing forthe steady state, and contention techniques for the abnorm alconditions.

VI. C O M P A R I N G THE M A C PROTOCOLSIn the anterior sections, we have shown thenecessity of im proving the M A C protocols used in SCADA

systems, w e have justified the use of three of theseprotocols, and we have also described their advantages anddisadvantages. Now is the time to compare them and todraw up some conclusions. For this, we have set up ascene in which a control center and various RT Us, b uildingup a SC ADA system, ar e joined together. In this systemthe RTUs send two types of information to the controlcenter: periodic information (measures) and events(alarms). T he control center can also send comm ands to theRTUs. This scene has been simulated for the threeprotocols under study and we have obtained results fordifferent numbers of RTUs connected to the channel.

In figure 5 , the behavior of the efficiency of theprotocols can be seen as a function of the numbe r of RTUs.As was expected, the polling protocol is clearly inferior tothe other two. On the other hand the performances of thetoken bus and the CSMA/CD protocols are quite similar,with the CShIlAICD presenting better behavior with fewRTUs (fewer number of collisions), while the token busslightly overtakes the CSMA/CD for very saturatedchannels (lots of RTUs). This behavior repeats itself forother parameters studied, such as the measurement refreshcycle (fig. 6) or the delay in the transmission of an event(fig. 7). The last of these studies shows, however, anopposite effect. In figure 8, the delay in the transmission ofa command for each one of

Efflclency' I

1 3 5 7 9 11 13 15 17 I9 21 23 25 27 29Number of Remoles

Fig. 5 . Protocols ef f ic iency


5/6

1199Measurement refreshwde

0' I1 3 5 7 9 1 1 13 15 17 10 21 23 25 27 29Number ofRemoles

Fig. 6. Pro toco ls measurement re f resh cyc le .

Event Transmlsslon Delay20 I

I %*-- I

1 8 1 1 1 1 ' 1 (1 3 5 7 0 11 13 15 17 10 21 23 25 27 20

Number of f l Bm O l eSFig. 7. Pro toco ls w e n t transmiss ion delay.

Command Trawmlsslon Delay

-Token Passing-- CSMAICD-*. Polling

-Token Passing4-CSMAJCD* Polling

-Tokon Passing* -CSMAJCD.-.Polling

Number of f lemotesFig. 8. Protocols cooi tnand transmiss ion de lay.

the protocols is shown. In this case, the polling protocolis clearly superior to the other two, which in turn, have avery similar behavior. This is due to the fact that, in thepolling protocol, the permission for transmission is grantedto the control center after each transmission while, for theother two protocols, the control center is only ano ther nodein the line.VII. CHOOSING A MAC PROTOCOLWith these results at hand, w e are able to select a

MAC protocol for S CAD A -applications. From what w ehave seen, the polling protocol is clearly inefficient, whichmeans discarding it, i fw e aren't dealing with an applicationin which the delay in the transmission of a command iscritical. So the decision is hetween the token bus and theCSMAlCD protocols. However the performances of bothprotocols a re quite similar, in other words, othe r featureshave to be taken into account.

In this sense, we have to firstly point out theimpossibility to deternline the time of arrival of a certainmessage which presents the CSM A KD protocol, althoughas was said before, it is possible to find some variants ofthis protocol having bounded delays. Greater difficultyintroduces the collision detection when one uses certainphysical media frequently employed in SCAD A systems.For this we would have to go back to the CSMA version(without collision detection), decreasing in this way theoveral performances.On the other hand, w e pointed out the difficulty ofusing the token bus protocol for the necessity of managingthe abnormal conditions through the application of

contention techniques. This complexity of the token busprotocol has been argued in other studies seeking the use OFpower lines as the physical media in LANs [111, to rejectthe token bus protocols and to apply the CSM A. How ever,in the SCADA applications, seeing as a main node exists(the control center), we can design a polling protocol forthe management of the abnorm al condil ions coordinated bythe control center. This significantly simplifies theresolution of the abnormal conditions, and eliminates themain disadvantage of the token bus protocol, maintainingis superiority over the alternative protocol (the CSMA).

VIII. CONCLUSIONSIn this paper we have presented the convenience ofoptimizing the Medium Access Control (MAC) protocols

ured in SCADA systems. We have selected for the study,In a ~ u c t i f i e d ay, the poll ing, C SM AIC D, and token bu sprotocols. We have described each one of them and


6/6

1200compared their performances. From the analysis, we [4]propose the token bus protocol, with th e describedadaptations, as the best MAC protocols for most of th eSCAD A applications.

[51IX. BIOGRAPHIESJoaquh Luque.- received his degree in Industr ia l Engineering in1980 and hi s Doctorate Ind ustrial Engineering in 1986 from th eUniversity of Seville (Spain). Sincc 1980, h e h a s workcd fo rseveral companies in th e area of S C A D A Sys tems fo r electricalnctworks, p articipating in some of t h e primary EM S projects inSpain . H e is currcntly a professor of electronic cnginccring a t theUnivers i ty in Scville, and he is a m e m b e r of the IEEE.

161

Isabcl G6mez.- received her Physics (Electronics) degree in 1989from the University of Seville (Spain). She has been a professorof electronic engineering at the University in Seville since 1990,where she is doing research o n problems in computercommunica t ions fo r the con t ro l of electrical powcr networks.

[7]

P I. REFERENCESItitertiational Orgrriiizatioii f o r Staidardiznrioii:"Iirfo~~nratioiirocessing systems- Opeti SystemsItitercoririectiorr- Basic Rcfet-etice M o d e l " , I S 0Standard 7498, 1984.

[91

[ lo1nternational Electrotechnical Commission TC57,IfTelecotitrol Eq uipinetit atid Systems. Part 5.I:Tratrrrnission Frame Formats I, IntemationalStandard. Geneva 1990.

[ I l l. Luque, J.I. Escudero: "Measurement

Transmission in Power Control Centers", IEEEMELECON'91, Ljubljana, May 19 9 I .

J. Luque, J.I. Escudero, I. Gdmez, "DifferentialTransmission of Evolution Measures in SCADASystems", IEE Proc. -Genet+.Tram. Distrib., Vol.141, No . 3 . May 1994The Institute of Electrical and ElectronicsEngineers, Carrier Sense Multiple Access withCollision Detectiott (CSMAlCD). Access Methodaid Physical Layer Specijkatio ns, AmericanNational Standard ANSUIEEE Std, 802.3. 1985.J. Luque, I. Gdmez, -J.I. Escudero, "Determiningthe Channel Capacity in SCADA Systems UsingPolIing Protocols", IEEE Power EtigineeritigSociety, Summer Meeting, July 1995.H. Takagi. L. Kleinrock, "Throughput Analysisfor Pers is tent CSMA Systems", IEEETrnnsactioris or1 Communications, Vol. Corn-33,No. 7. July 1985.F.A. Tobagi, V.B. Hunt, "Performance Analysisof Carrier Sense Multiple Access with CollisionDetection", Comp uter Networks 4, 1980.S . S , Lam, " A Carrier Sense Multiple AccessProtocol for Local Netwo rks", Computer Networks4, 1980.H. Takagi, S. Yamada, S . Sugawara, "CSMAICDwith Deterministic Contention Resolution". IEEEJ . Selected A r e a oti Coniniunictrtioru,Vol. SAC-1, No. 5, July 1985.J. Onunga, R.W . D onaldson, "Distribution LineCommunications Using CSMA Access Controlwith Priori ty Acknowledgements" , IEEETrailsactioris ot i Power Delivery, Vol. 4. No.2,April 1989.

system scad

Documents