Mil-std-188-220a Parameter Optimization for Tactical Internet

The Tactical Internet (TI) architecture has evolved into the Tactical Internet Division and Below (TIDB) network. One of the major architectural changes is the addition of multicast routing in the Internet Controller (INC) (with the MIL-STD-188-220A protocol) to support both SINCGARS nets and EPLRS C2 CSMA channels. The TIDB design also employs a disciplined queue structure which will support the addition/extraction of queue entries based upon message precedence, type and age. The MILSTD-188-220A protocol specifies a number of link layer and radio parameters which are interrelated and have significant cross-correlations which make linear optimization techniques diflcult. We used the robust design method, with dynamic signalto-noise ratio (SNR) and L18 orthogonal array, to pe~orm optimization. The application of Taguchi ’s experiment method firther reduced the necessary simulation time. The minute analysis method was applied to select the optimum parameter set. INTRODUCTION This paper addresses the application of robust design methods to optimize MIL-STD188-220A [Ref. 1] protocol parameters for the SINCGARS ASII? radio networks. Traffic loading on the network was varied at different levels according to the robust design methods [Ref. 2] to provide sensitivity analysis of the protocol parameters chosen for optimization. A detailed Optimized Network Engineering Tools (OPNETTM) from MIL3, Inc. was developed for MIL-STD-188-220A protocol used by the radios. The selection of protocol parameters for optimization and the TIDB traffic loading levels are described. Taguchi design experiments and the results from data analysis were used to obtain the optimum selection. The results confirmed that completion of SA DOWN messages with optimum set improves 15 percent while average delays decreases 3.7 seconds when compared with the nominal parameter settings before optimization. This is followed by confirmation and verification experiments. MODELED MIL-STD-188-220A PROTOCOL The MIL-STD-1 88-220A protocol defines several different network access algorithms and a number of link layer types. In this paper, the network access algorithm and link layer were the L-NAD algorithm and Type-4. The L-NAD stands for Load Factor Net Access Delay. This algorithm dynamically adapts to traffic loads in terms of precedence and quantity. The L-NAD scheduler was used at TFXXI testing. The Type-4 link layer protocol is also known as decoupled acknowledgment (ACK). In this protocol, messages are sent from source to destination and then entered back into the source’s retransmission queue. The source requires an acknowledgment from each destination. If the ACK is received, the message is deleted from the queue. If the acknowledgment is not received within a specified time period, the message is “enqueued” for a retry. A limit of MaxTxLinkLayer is placed on the total number of transmissions. As the number of transmissions is increased, the probability of success of individual messages increases while the overall network throughput decreases. Another control parameter called “RtxQTimeIn” is expressed in the number of seconds each spends in the retransmission queue. For SA message broadcast, the acknowledgment is not needed and the message is sent only once without retransmission. ROBUST DESIGN METHODOLOGY The following steps [Ref. 2, 3] need to be taken to use a robust design experiment: Zdentifi Main Function. The TIDB requires a tactical radio network which provides robust performance mixed voice and data traffic during “stressed link” conditions. The function of a Tactical Internet (TI) is to deliver user data traffic under dynamic conditions with maximum throughput and minimum delay. Identifi Noise Factors. Noise factors are parameters which cannot be controlled by system analyst. A successful system must be designed to perform satisfactorily at all levels within a given range. A company size net of 16 0-7803-4902-4/98/$10.00 (c) 1998 IEEE members was used to generate various types of TIDB traffic in our study: (1) situation awareness up (SA UP) messages, (2) SA down (SA DOWN) messages, (3) battle command control (C2) messages and (4) multicast messages. The SA DOWN messages sent by the SA server were further divided into friendly reports (K5. 1), enemy reports (K4.99) and C2 supplement messages. The message size, rate and precedence are listed in Figure 1. MSG Type I Size I Rate Precedence (Byte) (msg/hr) SA UP 30 150 routine SA DOWN K5.1 221 2100 routine K4.99 54 84 routine 576 2.4 urgent Supplement C2 UniCast 576 4 1/3 each precedence MultiCast 576 4 113each precedence Figure 1. Medium TIDB Traffic All messages were exponentially distributed to generate script files. The SA DOWN messages were dominant (more than 85 percent of the total traffic). This traffic level of Figure 1 is denoted as Medium. These generation rates were halved and then doubled to produce traffic levels denoted as Low and High in Figure 2. The voice load was expressed as a percentage of the total load in the net (i.e, 60 Yo of the time, a station in the net will be transmitting voice information). Three different random seeds were then used to run the simulation as a check on the data. In this study we treated noise as described in Figure 2. Noise Factor Level 1 Level 2 Level 3 I. Voice Loading 0910 3070 60% J. Traffic Loading Low Medium High K. Random Seeds 123 1234 12345 Figure 2. Noise Factors and Levels Identify Control Factors. The parameters set by the system analyst are called control factors. Control factors selected for this study are those which are controlled by the configuration database. The most significant of these are the L-NAD scheduler and the CSMA algorithms. The continuous scheduler timer (T.) is recalculated periodically if the channel is not busy by voice: TC= SchedIntMinFix + random (SchedIntLimit) where SchedIntMinFix = fixed time offset SchedIntLimit = SchedIntLimitFactor * (Number of Active Stations/16) * F1.ld SchedIntLimitLow < SchedIntLimit < SchedIntLimitHigh The SchedIntLirnitFactor and SchedIntLirnitHigh were used as scheduler related parameters. Two fixed values were, then, used in conjunction with CSMA channel access algorithm. Offset was used to define the minimum number of time slots before access could be initiated. Limit defined the number of time slots for the message precedence. Two different sets of control factors were selected for voice/data mode and data mode since the voice/data mode was prioritized for voice and the data mode was prioritized for data. The operating assumption was that the voice load was very light for the data mode networks. The number of voice slots (NumVoiceSlots) was one control factor for voice/data mode. There were 54 milliseconds guard time between datahoice slots to avoid the voice and data collision. It was assumed that the priority precedence (NumPrioritySlots) were overlapped with the urgent precedence for the voice/data mode. The limits and offsets were overlapped in data mode. These overlaps were defined in the control level setting. The link layer parameters (RtxQTimeIn and MaxTxLinkLayer) were added to the control factors for voice/data mode, And the maximum number of outstanding I-Frame Queue size (MaxRtxQSize) was selected for data mode. SELECT SIGNAL FACTORS AND TEST CONDITIONS To minimize the number of simulation runs, we simplified the signal and noise factors as listed in Figure 2. Noise orthogonal array of Lg as proposed by Taguchi is shown in Figure 3. Test Conditions #1 #2 #3 #4 #5 #6 #7 #8 w I.Voice Load 1 1 1 2 2 2 3 3 3 J. Traffic Load 1 2 3 1 2 3 1 2 3 K. Seeds 1 2 3 2 3 1 3 1 2 Figure 3. Noise Orthogonal Array (Outer loop) The numbers under the columns labeled test conditions 1 through 9 indicate the noise factor levels. Test conditions 1,2 and 3 were at O % voice loading. Test conditions 3,4, and 5 were at 30 $ZO voice loading. Test conditions 7, 8, and 9 were conducted at heavy voice loading (60 70). The performance was expected to degrade as the test number increases. Test conditions 1, 2 and 3 were used for data mode. Test conditions 7, 8, and 9 were used for voiceldata mode. Control Factor Levels. The settings of control factors are called levels. Each level represents an operating condition or setting. Eight control factors were selected to optimize MIL-STD-188-220A protocol parameters. Figure 4 presents the P-Diagram of MIL-STD188-220A protocol optimization. 0-7803-4902-4/98/$10.00 (c) 1998 IEEE TAGUCHI EXPERIMENT RUN MATRIX Noise Factors Net Members TIDB Traffic Voice Loading 4 -Cosite Responce Signal Factor MSG Thruput Time MIL-STD188-220A voice Completion