KR20070001547A - Virtual networking method using diffserv-over-mpls te - Google Patents

Abstract

A virtual networking method using a DiffServ-over-MPLS TE is provided to constitute each virtual network at each class-type by using an L-LSP in a DiffServ-over-MPLS TE model. One physical network is constituted with several virtual networks. An L-LSP constituting each virtual network has different characteristics in bandwidth, delay and protection mode in order to guarantee QoS(Quality of Service) per class-type. Virtual overlay networking can increase resource utilization by enabling bandwidth borrowing among virtual networks per each class-type. A packet transmitted through the virtual network goes through Meter/Marker according to a traffic parameter, and then waits in a corresponding queue. Each queue is allocated to a specific user group.

Description

Virtual Networking Method using Diffserv-over-MPLS TE}

1 is a concept of a virtual network

2 is a test bed

3 is a result of generating traffic

4 is the QoSVNServer class

5 is the ConnectionAdmissionControl class

Figure 6 QoSDiffServVirtualNetworking class

7 is the TEConnectivity class

8 is a TEConnectivity class

9 is PacketFlow class

10 is MPLS_ASNetwork class

11 is MPLS_ASBRNode (node) class

12 is a TEEnabledInterfaceLink class

13 is Virtual Overlay Networking Class-diagram (1)

14 is Virtual Overlay Networking Class-diagram (2)

15 shows Configure MOs

Figure 16 Create QoS Virtual Network

Figure 17 Create TE Connectivity

18 shows Add Packet Flow

In the DiffServ-over-MPLS TE model, it is possible to configure a virtual network using L-LSP. Virtual network is composed of virtual network for each class-type by using virtual leased line, so it is easy to guarantee quality of service (QoS) for each class-type and to increase resource utilization.

In the present invention, the L-LSP in the DiffServ-over-MPLS TE model is used as the virtual leased line.

 The concept of a virtual network is shown in FIG.

As shown in FIG. 1, several virtual networks are configured in one physical network. The L-LSP that constitutes each virtual network has different characteristics in bandwidth, delay, protection mode, etc. to guarantee QoS for each class-type. For example, in case of NCT, it has a protection mode of 1 + 1 or 1: 1, but in case of AF, it can be set to have a protection mode of M: N.

Another advantage of Virtual Overlay Networking is that it can increase resource utilization. When each virtual network is configured as a physical network, other specific class-type networks may have a lot of surplus bandwidth, but other class-type networks may run out of bandwidth and cause packets to drop.

However, when this is configured as a virtual overlay network, bandwidth borrowing is possible between virtual networks of each class-type, thereby increasing resource utilization.

Packets transmitted through the virtual network are first determined which class-type they belong to through classification, and go through a meter / marker according to a predetermined traffic parameter. After passing through the meter / maker, the virtual network is determined by the class-type and destination address, and waits in the corresponding queue.

In a virtual network, there can be multiple queues for each LSP, and each queue is assigned to a specific user group. That is, packets of the same class-type are transmitted to the virtual network with different priorities according to which user group they belong to and are provided with differentiated QoS.

. Router Configuration for Virtual Overlay Networking

In order to implement Virtual Overlay Networking, traffic classification, metering / marking, and scheduling functions are required in edge router. In addition, bandwidth can be allocated when setting the MPLS TE-LSP in the Cisco router, but the set bandwidth does not affect the actual traffic transmission to manage the amount of bandwidth allocated to each port in the control plane. Do not.

Therefore, in order to guarantee the allocated bandwidth for traffic passing through each LSP, it is necessary to configure scheduling in all routers along the path through which the LSP passes. At this time, since the packet passing through the core routers is encapsulated by the MPLS shim header, the fields in the MPLS header cannot be used for classification. When marking in the IP DSCP field on the Cisco router, the upper 3 bits are copied to the MPLS EXP field by default. Therefore, this can be used to classify packets in the core router.

Set up Virtual Overlay Networking based on the above mentioned items, and configure the experimental network as shown below to see if the bandwidth allocated for each class-type is guaranteed properly, and LSP for NCT, EF, AF4 class-type. Was set.

<Table 1> Setup example for SLA and excess traffic

<Table 1> shows the policy for SLA and excess traffic for each class-type. After such a preset, detailed settings for the Virtual Overlay Network are shown below.

<Table 2> Access-list, class-map definition for Classification

<Table 2> is configuration for classification at input port of edge router. As defined in <Table 1>, range of port number for each class-type is set. In the test, since the source and destination of the traffic are defined, when defining the access-list, the source and destination information is not set separately.

However, in actual application, access-list should be created considering both source and destination.

<Table 3> Policy-map for Metering / Marking

<Table 3> shows the policy-map for metering / marking. Traffic classified through the configuration as shown in <Table 2> is metered and marked according to the previously set SLA. In order to use the configured policy-map for metering / marking, it must be applied to the port where traffic comes in. The command to apply the policy-map set in Step 15 to Step 20 of <Table 3> to port.

The traffic after metering / marking enters the queue for each class-type and is sent to the next node by scheduling. Scheduling must be set to ensure the bandwidth allocated to each LSP. In this case, class-type is classified using the MPLS EXP field. That is, the packet marked with 0 in the DSCP field is traffic exceeded without keeping the SLA, and the traffic is processed in the same manner as the BF class-type because there is no need to guarantee QoS for the traffic.

<Table 4> Classification for Scheduling

<Table 4> shows classification for scheduling. In this case, traffic of each class-type is classified by MPLS EXP field for each class-type and classified based on that. At this time, since packets marked with values corresponding to each class-type are in-profile, QoS (bandwidth) should be guaranteed for this. Scheduling settings to guarantee bandwidth for these packets are shown in <Table 5>.

<Table 5> Policy-map for Scheduling Setting

<Table 5> shows scheduling setting to guarantee bandwidth by each class-type. In case of NCT class-type, priority keyword is used, and bandwidth keyword is used for the remaining traffic. Priority keyword uses LLQ (Low Latency Queue) in Cisco router to guarantee the specified bandwidth, but drops all traffic when more traffic comes in. The bandwidth keyword uses Class-based Weighted Fair Queuing (CBWFQ) in a Cisco router, using the bandwidth specified as weight.

Up until now, when the LSP was configured, the settings were used to control the traffic passing through the LSP. In order to transmit specific traffic through the desired LSP, route-map setting like <Table 6> is necessary.

<Table 6> Route-map setting

Since only one route-map can be set in one interface, in order to map several traffic to several LSPs, the sequence numbers must be set differently as shown in <Table 6>.

All the settings so far are set at the edge router 7204_H. As mentioned above, in order to guarantee the bandwidth allocated to each LSP, scheduling should also be set for routers located in the intermediate path. At this time, the contents to be set are the same as in <Table 4> and <Table 5>.

After the above configuration, the traffic was generated as shown in <Table 7> in order to check whether the bandwidth is properly guaranteed, and the result is shown in FIG. 3. It can be seen from FIG. 3 that the bandwidth allocated to each class-type is well guaranteed.

There are two parts to configuring a virtual overlay network. The first is to decide which class-type virtual overlay network to configure for the network, and then set up an LSP to form a virtual link of the virtual overlay network. The second is to set a virtual overlay configured when a user traffic request is made. It is configured to be transmitted using network.

4 shows a QoSVNServer class. The QoSVNServer class is a class for communicating with the Mobile GUI. All requests from the Mobile GUI and responses to the Mobile GUI are made through this class. The QoSVNServer takes a pointer to the QoSDiffServVirtualNetworking class and the ConnectionAdmissionControl class and calls the appropriate function of that class for requests from the GUI.

5 shows a ConnectionAdmissionControl class. The ConnectionAdmissionControl class is a function that controls admission to a general user's connection request. It is executed when a user's connection request comes in.

6 shows a QoSDiffServVirtualNetworking class. QoSDiffServVirtualNetwor-king class manages multiple Virtual Networks in one AS domain as a whole. The QoSDiffServVirtualNetworking class has a pointer list of QoS_VirtualNetwork class for each class-type and a pointer of MPLS_ASNetwork class.

When constructing MOs to be managed by NMS for the first time, instancing MPLS_ASNetwork class and instancing QoS_VirtualNetwork when request to create QoS_VirtualNetwork is received from GUI. All requests from the GUI are then passed to the MPLS_ASNetwork or QoSDiffServVirtualNetworking classes through this class.

QoS_VirtualNetwork class is a class managing Virtual Network for one class-type. Since a virtual network for one class-type is actually composed of several TE-LSPs as links, the QoS_VirtualNetwork class is managed with several TEConnectivity classes. 7 is a class diagram for QoS_VirtualNetwork.

The TEConnectivity class is a class for managing TE connectivity (TE-LSP) that constitutes a virtual network. To this end, TEConnectivity class has not only the characteristics of TE-LSP but also various values used to configure and manage TE-LSP in real routers as member variables.

In addition, the TEConnectivity class is managed with the PacketFlow class to manage the various packet flows provided with the corresponding TE-LSP. 8 shows a TEConnectivity class.

The PacketFlow class is a class for managing packet flow of a user who is receiving a service through a virtual network. To do this, the PacketFlow class manages the configuration related values of the user's packet flow in the router as a member variable. 9, the details of the PacketFlow class can be seen.

The overall structure for managing Virtual Overlay Networking is located in NMS. NMS is in charge of managing the entire AS network. For this purpose, the MPLS_ASNetwork class exists, and the MPLS_ASNetwork class has information on nodes and links in the AS network. In this system, the functions of EMS and NMS are not clearly distinguished, but for future expansion, the MOs are newly managed by classifying them from MPLS layer network. The MPLS_ASNetwork class must know how to communicate with the various EMSs. All communication with the EMS is done through the MPLS_ASNetwork class.

The NMS initially obtains MO information from EMSs and newly configures and manages the MO. All settings via NMS are passed back to the router via EMS.

 10 shows the MPLS_ASNetwork class. Obtaining MO information from MPLS Layer Network sends and receives in XML format for future expansion, and requests to MPLS Layer Network use APIs defined in MPLS Layer Network. (Figure 2.4.3.8) and (Figure 2.4.3.9) are classes for Node and Link using information received from MPLS Layer Network.

After setting the TE-LSP, to ensure the allocated bandwidth, the scheduler should be set for all nodes that pass the LSP. To do this, NMS needs to know and manage what kind of class-type LSP is passed for each interface and how much bandwidth is allocated to this LSP. For this purpose, after one TE-LSP is configured, the LSP is notified to all interfaces (TEEnabledInterface class) that it passes.

First, when MPLS_ASNetwork is created, request RouterIDList managed by EMS to MPLSLayerNetwork in EMS. When you receive the Router ID List, create MPLSRouter as many as the number of routers and repeat the followings. After creating MPLSRouter, request the router's information again and set it in the created MO and request the Interface List in the Router.

Then, create TEEnabledInterface classes as many as the number of interfaces, receive information about each interface, and set the information to the corresponding interface. When all of these tasks are completed, the NMS obtains information about all MOs of the EMS and configures the MO.

In order to use the Virtual Overlay Networking function, the administrator must first decide which class-type Virtual Overlay Network to configure on the AS Network and which Edge Router to configure the Virtual Overlay Network on. To do this, a request to configure a Virtual Overlay Network is sent from the Mobile GUI, which sends a list of edge routers (ASBRs) currently on the AS network to the Mobile GUI.

Using this, administrator decides class-type and list of edge router of Virtual Overlay Network to create QoS_VirtualNetwork class according to the decision. QoS_VirtualNetwork class is created one for each class-type and has information about which edge routers it is composed of. At this point, the edge router becomes the node that constitutes the virtual network from the perspective of the virtual overlay network.

After the administrator decides on the class-type and edge router, the administrator must set up the TE-LSP between these edge routers. TE-LSP is a link constituting a virtual network from the viewpoint of a virtual overlay network. In order to configure TE-LSP, accurate information about the class-type and source / destination to which TE-LSP is to be configured first is needed.

When the administrator decides the class-type information, the list of edge routers that can be configured in the QoS_VirtualNetwork is delivered to the Mobile GUI. The administrator selects source and destination from the edge router list and sets other necessary information and requests TE-LSP (TE_connectivity) configuration to QoS_VirtualNetwork.

QoS_VirtualNetwork requests TE-LSP configuration with MPLS_ASNetwork because TE-LSP requested to actual router must be set. MPLS_ASNetwork calls functions for router configuration to EMS (MPLSLayerNetwork) and informs that TE-LSP has been set on all interfaces through which TE-LSP passes to manage the configured TE-LSP and updates the information of MO. After all these steps, create TE_Connectivity class.

After all preparation for the Virtual Overlay Network, the packet flow class should be created to inform the MO that the user traffic passes the specific TE-LSP (TE_Connectivity) and set in the router in order to receive the actual packet flow. Select TE-LSP (TE_Connectivity) through which user's packet flow passes, set flow information, and request AddPacketFlow to the corresponding QoS_VirtualNetwork. Check if there is enough bandwidth to allow this packet flow. Create a PacketFlow class and request MPLS_ASNetwork to set this information on the router.

MPLS_ASNetwork configures the router to provide QoS to the user's packet flow by calling various functions.

According to the present invention, it is possible to configure a virtual network using L-LSP in the DiffServ-over-MPLS TE model, and the virtual network is to configure a virtual network for each class-type using a virtual leased line. It is easy to guarantee the quality of service (QoS) for each class-type and increases the resource utilization.

Claims (1)

Virtual Networking Method Using Diffserv-over-MPLS TE

Images