KR20130022784A - A resource allocation method for the assured service in differentiated services through networks within a vessel - Google Patents

Abstract

A method of allocating a resource for an Assured Service using a DiffServ scheme for in-ship network is disclosed. A resource allocation method for an Assured Service of an in-vehicle network DiffServ method according to an aspect of the present invention, the maximum guaranteeing for each subclass according to the required buffer size by differentiating the amount of bandwidth allocated for each Assured Service subclass Differentially setting a delay time, setting a RIO variable value based on a buffer size determined according to a network topology and a ratio of bandwidth allocated to each subclass, and controlling access by applying the Rio variable value And setting the size of the output link bandwidth on which the reference is determined.

Description

A Resource Allocation Method for the Assured Service in Differentiated Services through Networks within a Vessel}

The present invention relates to a DiffServ (Different iated Services) scheme as a structure for a next-generation Internet capable of guaranteeing QoS required by a user. More specifically, the present invention relates to a PHB (Per B) for an IP packet using a DiffServ Code Point (DSCP). It relates to a resource allocation method for Assured Service of DiffServ type by defining Hop Behavoir.

In order to guarantee minimum rate QoS for TCP flow using AS in DiffServ method for in-vehicle network, access control is required, and protection and yield guarantee for in-profile packet traffic contracted by user is required.

However, the concept of RIO variable setting that can maximize the yield and link utilization for in-profile traffic under access control is still in the conceptual state.

Accordingly, a technical problem of the present invention is to provide a service for a DiffServ type AS for in-vehicle networks, and to provide a maximum delay time for each subclass according to a required buffer size by differentiating the amount of bandwidth allocated for each AS subclass. It is an object of the present invention to provide a resource allocation method for an Assured Service of a DiffServ method for in-vehicle networks that makes a difference to be relatively different.

Another object of the present invention is to set the RIO variable value and access control based on the buffer size determined according to the network topology and the ratio of bandwidth allocated to AS subclasses to maximize the yield and link utilization for in-profile traffic. It is to provide a resource allocation method for setting the size of the output link bandwidth to be a reference of.

 The object of the present invention is not limited to the above-mentioned object, and other objects that are not mentioned will be clearly understood by those skilled in the art from the following description.

According to an aspect of the present invention, a resource allocation method for an Assured Service of an in-vehicle network DiffServ method according to an aspect of the present invention differentiates the amount of bandwidth allocated for each Assured Service subclass, and for each subclass according to the required buffer size. Based on the buffer size determined by the ratio of bandwidth allocated by network topology and Assured Service subclasses to relatively differentiate the guaranteed maximum latency and maximize yield and link utilization for in-profile traffic. RIO) setting the variable value and setting the size of the output link bandwidth that is the reference of the connection control.

Here, the traffic characteristics of the AS In-profile packet defined in the DiffServ method for the in-ship network are the traffic conditioner in the inlet router of the DiffServ domain connected with the average transmission rate and the maximum transmission rate reported by the user. The burst length is defined by three factors related to the size of my token bucket.

According to the traffic characteristics of the AS in-profile packets, if the amount of bandwidth allocated by the j th subclass of the AS is determined as u _j r _i using the over-provisioning factor u _j value, the maximum delay time for each sub class is determined. Can be relatively differential.

In addition, the step of setting the Rio variable value is the average ratio of the buffer space in which the in-profile packet and the out-of-profile packet is stored in the AS queue of the DiffServ router for in-vessel network 1: 1: (β <1 The Rio variable value is set to satisfy the relationship

As described above, according to the present invention, the maximum delay time guaranteed for each AS subclass is differentiated by differentiating the amount of bandwidth allocated for each subclass of the AS of the DiffServ method for in-ship network. In order to maximize in-profile traffic yield and link utilization, the RIO variable value is set based on the buffer size determined by the network topology and the ratio of bandwidth allocated to AS subclasses. We proposed a method for setting the size of. Using the proposed method, we can derive the service capacity for each AS subclass according to the network topology, and apply the proposed RIO variable value in consideration of the service capacity to calculate the in-profile traffic yield and link utilization. It can be maximized.

1 is a diagram illustrating a Rio buffer management scheme according to an embodiment of the present invention.
2 is a diagram illustrating AS traffic in a DiffServ router for in-vehicle networks considering a network topology according to an embodiment of the present invention.
3 is a diagram illustrating a simulation model for the Rio parameter setting method of FIG.
4 is a diagram illustrating the number of IN packets and the number of IN packets that arrive at the C1 router of FIG.
5 is a diagram showing the total number of packets passed by the C1 router of FIG.
FIG. 6 is a diagram illustrating the number of In-Profile bytes received at the R2 host of FIG. 3.
FIG. 7 is a diagram illustrating a change in an average queue length in the C1 router of FIG. 3.
FIG. 8 is a diagram illustrating the number of received In-bytes of an R2 host according to the number of transmitting hosts of FIG. 3.

Advantages and features of the present invention and methods for achieving them will be apparent with reference to the embodiments described below in detail with the accompanying drawings. The present invention may, however, be embodied in many different forms and should not be construed as being limited to the embodiments set forth herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of the invention to those skilled in the art. Is provided to fully convey the scope of the invention to those skilled in the art, and the invention is only defined by the scope of the claims. It is to be understood that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting of the invention. In the present specification, the singular form includes plural forms unless otherwise specified in the specification. It is noted that the terms "comprises" and / or "comprising" used in the specification are intended to be inclusive in a manner similar to the components, steps, operations, and / Or additions.

Hereinafter, a method of allocating a resource for an Assured Service of a DiffServ scheme for an in-vehicle network according to an embodiment of the present invention will be described with reference to FIGS. 1 to 8.

First, the amount of bandwidth allocated for each Assured Service subclass is differentiated to differentiate the maximum delay time guaranteed for each subclass according to the required buffer size.

Specifically, if the amount of bandwidth allocated from the j th subclass of the AS is determined by u _j r _i using the over-provisioning factor u _j value according to the traffic characteristics of the AS in-profile packets, the maximum delay provided for each sub class. Time can be relatively different.

As shown in Equation 1 below, when the bandwidth of u _j r _i is allocated to each subclass with respect to the average transfer rate r _i required by the user, the maximum delay time d _{max, j} guaranteed for each subclass is determined according to the traffic information per flow. Differential is achieved.

[Equation 1]

In the system proposed from Equation 1, the quality of service (QoS) guaranteed to the AS user is the maximum delay time of (b _i -1) l _i for in-profile packets. 2 is a diagram illustrating a network topology of a DiffServ internal router for an in-vehicle network. When there is no resource reserved for PS, the output link bandwidth of the DiffServ internal router for an in-vehicle network that the AS can use is referred to as L _o . Defined. Also _, when the amount of resources reserved for BA (Behavior Aggregate) of flows belonging to the PS at time u _k is R _{ps, k} , the output bandwidth available to the AS class is defined as L _k as in Equation 2.

In the same way, in the previous DiffServ router for intra-vehicle network, which uses the jth input link of the n input links connected to the in-vehicle DiffServ router for the output link, the bandwidth L _j available to the AS class is expressed by Equation 2. Is defined.

From Equation 3, λ _top represents the ratio between the input bandwidth and the output bandwidth of the router according to the network topology, and the sum of each R _{ps , j} is equal to R _{ps , k} . If λ _top is greater than 1, the resource reservation amount R _{ps , k} for BA of flows belonging to PS continues to increase, and if it is less than 1, it continues to decrease.

&Quot; (2) &quot;

&Quot; (3) &quot;

Out-of-profile packets of AS traffic are generated when the amount of packets generated by the user's TCP traffic is greater than the amount of tokens remaining in the traffic conditioner of the inlet router of the DiffServ domain for the onboard ship network connected to the user. Packets are generated by the amount of tokens remaining. Since the output link bandwidth of the router should be made available to most of the in-profile packets, in the present invention, in-profile packets and out-of-profile packets are stored in the AS queue of the DiffServ router for the in-vehicle network. Set the RIO variable values so that the average ratio is 1: β (β << 1). If the average ratio of the buffer space in which the in-profile packet and the out-of-profile packet are stored in the AS queue of the DiffServ router is 1: β (β << 1), Out-of-profile packets will use the output link bandwidth L _k at a ratio of 1: β (β << 1).

In other words, in-profile packets use the bandwidth of and out-of-profile packets use the bandwidth of L _k β (1 + β). In addition, in order to make the average ratio of the buffer space sizes occupied by in-profile packets and out-of-profile packets in the AS queue to be 1: β (β << 1), as shown in Equation 4, access control considering u _j The sum of the average rates of in-profile packets generated in the flows through the router should not exceed the bandwidth of L _k / (u _j (1 + β)).

&Quot; (4) &quot;

Since the DiffServ method for in-vehicle networks does not perform flow information management, when the maximum value l _{i , max} among the burst lengths l _i of the flows reporting the length of an arbitrary time interval τ is set, u _j is as shown in FIG. 2. In the case of one AS queue of 1, the maximum amount of in-profile packets that can arrive during τ from previous routers by n flows via the current router is the total bandwidth of L _k / (1 + β) through access control. In case of reservation, L _k b _{i , max} l _{i , max} / ((1 + β) packetsize)) as shown in Equation 4, and the maximum arrival amount of out-of-profile packets is βL _k b _{i , max} l _{i , max} / ((1 + β) packetsize)) Here _{, the} values b _{i , max} , l _{i , max} represent the largest value among the reported traffic information b _i , l _i .

In addition, the maximum amount of packets remaining in the AS queue for τ (= l _{i , max} ) by the amount of output link bandwidth, L _k / (1 + β) and L _k β / (1 + β), is in-profile. The packets are L _k (b _{i, max} -1) l _{i, max} / ((1 + β) packetsize) and the out-of-profile packets are βL _k (b _{i , max} -1) l _{i, max} / ( (1 + β) packetsize). In general, however, λ _top is greater than or equal to 1 and less than b _{i , max} . Therefore, if λ _top is less than b _{i , max , the maximum} number of in-profile packets that can arrive during τ is L _k λ _top l _{i , max} / ((1 + β) packetsize)), and Out-of-profile packets The maximum amount of these is βL _k λ _top l _{i , max} / ((1 + β) packetsize)). The maximum remaining amount in the AS queue during τ is the in-profile packets L _k (λ _top -1) l _{i, max} / ((1 + β) packetsize)) and the out-of-profile packets are βL _k ( λ _top -1) l _{i, max} / ((1 + β) packetsize))

These maximum amounts are the amount of packets that must be protected in RIO buffer management. In the RIO method, max_in and max_out set the maximum available buffer size for the in-profile packet and the entire packet, respectively. Assuming no in-profile packets arrive, all out-of-profile packets should be available. Equation 5 shows a value calculated in consideration of this case. In addition, if RIO protects the buffer space larger than the total buffer size determined by the value obtained from Equation 5 and the PS reserved resource amount, in-profile traffic yield may not be guaranteed due to increased out-of-profile packets. In addition, it is possible to set the size of the reference output link bandwidth of the access control of the in-profile traffic from the β value of the equation (5).

[Equation 5]

As a result, the bandwidth of u _j r _i is allocated in the AS subclass with the over-provisioning factor u _j to the flow requiring the average transfer rate r _i in the DiffServ method for in-vehicle network (b _i- u _j ) l _i / u _j In order to guarantee the maximum delay time, the DiffServ router for in-vehicle network must control access according to the value calculated by the network topology and PS reservation resource amount. In the RIO buffer management method, max_in, max_out, min_in, and min_out should be set according to the buffer size required for in-profile packets and out-of-profile packets. At this time, the max_in value is set in consideration of the delay time according to the maximum buffer size of the in-profile traffic, and the max_out value is considered in consideration of the link utilization efficiency.

Set to a value. The values of min_in and min_out are set in consideration of the average queue length, and are generally set to 1/2 of max_in and max_out.

In the present invention, as a service providing method for the DiffServ type AS for in-network networks, the maximum delay time guaranteed for each subclass is differentiated by differentiating the amount of bandwidth allocated to the queue and subdividing the queue for each subclass. In order to maximize the yield and link utilization for in-profile traffic, access control is performed such that the output link bandwidth allocated to in-profile traffic is L _k / (u _j (1 + β)). According to the network topology and the size of the buffer space determined by the bandwidth allocated for each subclass of AS, the RIO variable values are presented as shown in [Table 1].

[Table 1]

In the present invention, simulation is performed considering only one AS subclass with u _{j equal} to 1. On the other hand, in order to guarantee the minimum rate QoS to the TCP flow in the network using the TCP transport protocol, the Token Loss problem in the Traffic Conditioner must be solved. Token Loss is a phenomenon in which the token in the token bucket is lost because Ack does not arrive for the transmitted packet even though there are remaining tokens in the Traffic Conditioner.This causes in-profile traffic from the sending host as much as the average transfer rate contracted by the user. You won't be able to. Therefore, in order to guarantee the minimum rate QoS for TCP flows, Token Loss is suppressed as much as possible so that in-profile packets occur at a level close to the average transmission rate contracted by the user. As much as possible to compensate for the loss of the average rate. On the premise of this, the network router sets buffer management criteria and performs access control based on information such as average transfer rate and maximum transfer rate of flows. The purpose of the buffer management approach is to improve the yield of high priority in-profile traffic and overall link utilization.

The number of hops with different round trip times (RTTs) for each sending host and the possibility of packet discard depending on the routed route, and in other situations, hosts with large hop counts are more likely to be discarded by RIO's avg_in and avg_total, Hosts with a large RTT are more likely to experience Token Loss due to packet dropping. That is, the traffic generation amount of flows becomes unfair. In this environment, traffic originating from the sending host above the contracted average transmission rate without Token Loss ultimately results in in-profile traffic of a host with a large number of hops and RTTs when there are multiple hosts reporting the same average transmission rate. In-profile traffic generation of hosts with small RTTs and hops is the same, and research on this is underway. In the present invention, the solution for the Token Loss problem was not applied, but as shown in FIG. 3, the traffic profile of the average transmission rate and the maximum transmission rate equal to the RTT of each transmitting host and the number of passing hops are equal, so that all hosts are equal. The incidence of in-profile traffic and fairness among flows were minimized and the probability of occurrence of token losses was minimized.

3 illustrates a network topology model for analyzing the performance of the RIO variable setting method under the condition that access control is performed in consideration of the network topology.

FIG. 3 is a simulation model for evaluating the performance of the proposed RIO variable setting method for the general case in which b _{i and max} values are larger than λ _top in Table 1, where λ _top = 3. In FIG. 3, E1 and E2 routers represent DiffServ inlet routers for onboard networks, and C1 represents DiffServ internal routers. The size of the token bucket from Equation 1 assuming that each transmitting host S _n transmits an average of 0.266 Mbps and a maximum of 1.064 Mbps to the receiving host R _n represented by the same number, and l _{i and max} are 33 ms.

Is 27packets and packetsize is set to 125 bytes. As a result, traffic from an average of 2.66 Mbps and up to 10.64 Mbps from 10 hosts will pass through the 12 Mbps link. The bottleneck link bandwidth is the output link of the C1 router and is set to 4Mbps, and the actual buffer size of the C1 router is set to 350 packets. Accordingly, FIG. 3 shows a situation in which access control is performed on an AS considered according to a value obtained by ½ from Table 1 of the present invention since traffic having an average data rate of 2.66 Mbps passes through a bottleneck link having a 4 Mbps bandwidth. In the network environment of FIG. 3, when the proposed RIO variable setting method is applied to the internal router of C1, b _{i , max} is calculated as 4 from the maximum transmission rate and the average transmission rate of the transmitting host, and the maximum delay time guaranteed is 0.1sec from Equation 1. The total protected buffer space size required is 266 packets from Table 1. According to β value of 1/2, in-profile packets use 178 buffer space and 2.66Mbps bandwidth, and out-of-profile packets use RIO max_in and max_out to use 88 buffer space and 1.34Mbps bandwidth. 178 and 266 respectively.

The E1 and E2 routers also use the RIO method, but since the input link bandwidth is smaller than the output link bandwidth, no packets remain in the buffer, so the effect of the configured RIO variable value does not need to be considered. Each sending host uses TCP Reno and the RTT is set to 16ms. The value of W _q used in the RIO method is 0.002 for In-profile traffic and Out-of-profile traffic.

and

Values of 0.02 and 0.05 were used, respectively. Table 2 shows the values of each variable set in the simulation in FIG. 3.

[Table 2]

FIG. 4 illustrates how the total number of in-profile packets and the number of packets that have been passed for 100 seconds from 10 transmitting hosts are changed according to the configuration values of the RIO variables max_in and max_out in the C1 router of FIG. 3.

The results of FIG. 4 show that the total amount of in-profile traffic arriving at the variable values max_in = 178 and max_out = 266 set in the proposed method is about 2.55 Mbps, and the amount of in-profile traffic arrived is very close to the maximum value. Can be. On the other hand, since the amount of in-profile traffic arrived differs from the amount of traffic passed, it can be seen that the discarding of in-profile packets is occurring, and in-profile traffic of 0.255Mbps is generated for each host in the proposed variable value. Therefore, it can be seen that a minimum token loss of 0.011Mbps (4.3%) occurs on average.

FIG. 5 shows the total number of packets passed through the C1 router according to each RIO variable configuration value, that is, link utilization. In the RIO variable value set in the method proposed in FIG. 5, traffic of about 3.995 Mbps passes through, indicating a link utilization rate of 99.875%, which is very close to the maximum value.

By applying the proposed RIO variable value setting method under the condition that access control is performed by the method proposed in FIGS. 4 and 5, the total amount of in-profile traffic arriving from the hosts allowed to access and the total link utilization rate can be maximized. From the results, it is expected that the yield of each flow can be maximized.

FIG. 6 shows the in-profile traffic yield at the receiving host according to each setting value of the RIO variables max_in and max_out in the simulation topology of FIG. 3 and is represented by the number of bytes received at the R2 receiving host for 100 seconds. In case that the max_in value is 178 and max_out is 266, the number of bytes of In-profile packet received in 100 seconds of one host is 3224165 bytes, which is about 0.258Mbps, and the total number of bytes received is 5038040. The overall yield is about 0.403 Mbps, accounting for 1/10 of the output link 4 Mbps of the C1 router.

In the result of FIG. 6, if max_out is fixed to 266 and max_in is increased, the number of received In-profile packets increases until max_in is 178. However, if greater than 178, max_in does not increase because the number of received In-profile packets does not increase. It is reasonable to set 178, and if max_in is fixed to 178 and max_out is increased, the maximum number of in-profile packets received at 266 is greater, and if it is greater than 266, the out-of-profile packets increased due to fewer packet discards. As occupied, the number of in-profile packets received decreases. Therefore, it can be seen that the max_in and max_out values show the best performance in the yield of in-profile traffic per flow.

It will be understood by those skilled in the art that the present invention may be embodied in other specific forms without departing from the spirit or essential characteristics thereof.

The above described embodiments are to be understood in all respects as illustrative and not restrictive. The scope of the present invention is shown by the following claims rather than the detailed description, and all changes or modifications derived from the claims and their equivalents should be construed as being included in the scope of the present invention.

Claims (2)

Resource allocation method for Assured Service of DiffServ type for in-ship network,
Differentiating an amount of bandwidth allocated for each Assured Service subclass to differentiate a maximum delay time guaranteed for each subclass according to a required buffer size;
Setting a RIO variable value based on a buffer size determined according to a network topology and a ratio of bandwidth allocated to each subclass; And
Setting a size of an output link bandwidth as a reference for access control by applying the Rio variable value
Resource allocation method for the Assured Service of the DiffServ scheme for in-vessel network comprising a. The method of claim 1, wherein the setting of the Rio variable value comprises:
The Rio variable value is set to satisfy the relationship that the average ratio of the buffer space in which in-profile packets and out-of-profile packets are stored is 1: β (β <1). Comprising the step of setting
Resource allocation method for Assured Service of DiffServ method for in-vehicle network.

Images