KR100801575B1 - Fault Recovery Method according to the traffic character - Google Patents

Abstract

The present invention discloses a failure recovery method according to traffic characteristics that differentiates resilience in the event of a failure according to traffic in consideration of the importance of service quality provision level and the user's cost burden.

The method of the present invention is a multi-protocol Label Switching (DiffServ-over-MPLS) global network supporting method for supporting differential services, which is composed of a plurality of nodes, in a service level agreement (SLA). By setting the error recovery option by the network, and segmenting the global network into a predetermined number of subnetworks in consideration of the number of nodes and the number of links, the failure recovery is differentiated according to traffic characteristics, and the propagation delay is minimized when detecting a failure. Subnetworks can cope with node failures by overlapping a few nodes. The failover options include a 1 + 1 path protection scheme for transmitting the same data through different paths, a 1: 1 path protection method, It is a 1: N, M: N path protection method, the 1: 1 path protection method is composed of a standby mode (stand-by mode), a path option mode (path-option mode) and fast rerouote (Fast rerouote).

Traffic, failover, propagation delay, segmentation, node nesting

Description

Fault recovery method according to the traffic character}

1 is an example of typical encapsulated voice traffic;

2 is an example of FIS propagation delay in a general global network;

3 is an example of a propagation delay of a failure indication signal in a segment-based network according to the present invention;

4 is an example of a 1 + 1 path protection configuration according to the present invention;

5 is an example of a segment based network according to the present invention;

6 is a class diagram of a failure module management object according to the present invention;

7 is a sequence diagram illustrating a failover procedure in accordance with the present invention.

The present invention relates to a network management technology, and more particularly, to a failure recovery method according to traffic characteristics that differentiates the recovery ability when a failure occurs according to traffic in consideration of the importance of service quality provision level and the user's cost burden.
In general, differentiation / differentiation in failure management function provides differentiation in resilience in the event of a failure in consideration of the importance of service quality provision level and user's cost burden.

The reason for the differentiation and differentiation in the fault management function is that the amount of resources consumed and the recovery time are different according to the failure recovery scheme. Therefore, traffic having a high priority provides a fast failure recovery time even if the resource consumption is high. On the contrary, for the traffic with low importance, providing a resource recovery method with low resource consumption is effective in providing a quality service negotiated with the service user.
However, the technology related to quality assurance network management in the DiffServ-over-MPLS (MPLS) environment supporting the differential service has not been proposed yet.

The present invention has been proposed to meet the necessity as described above, and provides a failure recovery method according to traffic characteristics that can increase the efficiency of fault management by providing a failure recovery scheme that is differentiated according to each traffic characteristic. The purpose is.

In order to achieve the above object, the present invention provides a quality level network management method in a multiprotocol label switching (DiffServ-over-MPLS) global network environment supporting a differential service. Set the failure recovery option by contract (SLA), segment the global network into a predetermined number of sub-networks in consideration of the number of nodes and the number of links, differentiate the failure recovery according to traffic characteristics, and propagate the delay upon failure detection. It characterized in that to minimize.
The subnetwork is configured to cope with node failure by overlapping a few nodes, and the failure recovery option includes a 1 + 1 path protection method for transmitting the same data through different paths, a 1: 1 path protection method, It is a 1: N, M: N path protection method, the 1: 1 path protection method is composed of a standby mode (stand-by mode), path option mode (path-option mode) and fast rerouote (Fast rerouote).
Hereinafter, exemplary embodiments of the present invention will be described in detail with reference to the accompanying drawings.
First, the failover requirements according to the traffic characteristics are as follows.

In order to provide real-time multimedia services to users, a network management scheme suitable for the characteristics of multimedia services and a technique for guaranteeing the quality of services according to user requirements are required. In the present invention, in order to provide a real-time multimedia service, it is analyzed how much performance should be guaranteed for each multimedia service.

In general, voice data (VoIP) is encapsulated by RTP as shown in FIG. 1, which uses the same form for all messages. Because it supports application layer frames, this message type requires that messages be organized as required by various interpreters and specific applications.
Referring to FIG. 1, an RTP packet consists of a 12-byte header followed by a payload. According to the audio conference scenario of RFC 1889, which originally described the use of RTP, suppose that the data to be transmitted is sent audio data in semantic for 20 ms, a segment of the audio data is placed in front of the RTP header and the RTP message is transmitted by UDP packet. Is encapsulated. The RTP header indicates the audio conference encoding (PCM: Pulse Code Modulation) used. That is, the audio segment is lost by the failure time.

In order to transmit voice data through the Internet, a voice codec is required, and the level of bit errors that can be handled by the codec when an error occurs in a voice packet varies depending on the codec. Voice coding techniques used for VoIP should guarantee the performance of the algorithm itself for packet loss ratios of 1% (BER), 3% (air condition), and 5% (VoIP Pattern). Therefore, the network manager can use the service without affecting the failure when the failover is completed within the boundary of the error codec's own error recovery for the failover performance.
In fact, when voice data is sampled at 8000Hz, the fault rate is 64kbps per second, and the failure time is 10ms for 1% packet loss among 64000bit voice data per second. 3ms is 30ms and 5% is 50ms. When one second of voice data is used, a failure occurrence time of 10ms is obtained at 1%. Therefore, when a failure occurs within 50 ms, the user does not feel a failure in providing a real-time multimedia service.

delete

If a failure occurs, packet loss occurs. If a packet header is lost, the entire voice data must be discarded. However, the impact on the quality of voice data when each packet is lost will be solved by a study related to the quality of voice encoding technique.In the real-time multimedia environment, the transmission network does not generate more than 5% packet loss per unit time. Provides criteria for disaster recovery.
Segment-based failover method is as follows. The time required for the automatic recovery takes the time to process the fault and the message delay time for the fault detection time and the alarm transmission as shown in Equation 1 below.
T _restoration = T _detection + T _notification + T _operation + T _recovery

delete

delete

delete

Here, T _restoration is the total time required for recovery, T _detection is the time required for fault detection, T _notification is the transmission delay time required to transmit and analyze fault information, and T _operation is used to switch the path or link of the fault. The time taken for T _recovery is the time taken for the traffic to flow to the new path after the backup path is set up normally.

In general, failures that greatly affect the communication network occur at the physical layer, so the failure detection time is very short. In addition, the failure detection time is almost constant regardless of the recovery method, so the time required for automatic recovery is greatly affected by the transmission delay time. However, reducing the transmission delay time depends on the speed of propagation and the performance of the vendor of each node. Therefore, a technique for segmenting and managing networks is needed to reduce the delay time itself.
Fault detection Time to detect a fault Fault Notification Cisco Router Control SONET Alarms Various physical alarms Within 50ms (Physical) Interface down notification RSVP Hellos (MPLS Signaling) RSVP Hello Extension: node failure detectoin 20ms or less (RSVP Hello interval time = 10ms) RSVP-TE notify message
Table 1 shows an example of failure detection time in a Cisco router.

delete

delete

The propagation delay is a part of the time duration needed to locate and analyze faults in a real network.

Assuming that the general propagation speed is 2 * 10 ⁸ m / sec and the time to process data at each node is 2ms, the propagation delay time is shown in Table 2 below. If the end-to-end distance exceeds 1500 km, in the worst case, it takes 110 ms to send fault detection only. If this time is taken into account, the time taken to analyze and recover from the fault is considered a high priority. The branch cannot guarantee the quality of service of a user who is provided with a real-time multimedia service.

Therefore, the present invention proposes a quality-safety failure recovery criterion in a system for segmenting and managing a network so that a propagation delay time is about 25 ms. In fact, the US major ISP, MCI, continues to measure and publish packet delivery time every five minutes, with round-trip times in the US within 45ms, 30ms in Europe and between the US and London. The probability of delivery within 90ms is 99% per year.

Table 1 is inferred about the failure detection time in the Cisco router. In SONET, the failure recovery time is set to 50ms or less, so it takes less than 50ms to detect a failure in SONET. You can get the message "Physical Interface down" from the trap message of the Cisco router.

In addition, the node failure can be determined within 20 ms by adjusting the interval time of the RSVP Hello message. The default interval time of the RSVP Hello message in the Cisco router is 50 ms, but it can be set within 50 ms by adjusting the interval time. To get a faster fault detection time, you can reduce the interval time of RSVP messages further, but you need an appropriate interval time to reduce the load on the network.

T _detection is detected by a neighboring node or link of a failed network element, and T _notification is generated from the network and analyzed through a fault indication signal (FIS) to determine where the fault occurred. In this case, consideration should be given to propagation delay.

Table 2 below shows an example of propagation delay according to distance and number of nodes. In fact, the larger the link distance, the larger the propagation delay time, and as the propagation delay time increases, the total recovery time becomes larger.
Street 500 km 1000 km 150 km Node Transit node Transit node Transit node Less than 10 Don't consider 15 or less Don't consider 20 or less Don't consider Propagation delay 25 ms 5 ms 40 ms 10 ms 55 ms 15 ms
Therefore, it is necessary to design the size of the network to be applied to the appropriate propagation delay in consideration of propagation delay in network management for fault recovery. Message transmission delays that may occur from a time point of failure in a general global network to a time point of automatic recovery are shown in FIG. 2.

delete

delete

delete

Referring to FIG. 2, the global network includes a working node of W1 to W9 and a backup node of B1 to B7. In such a global network, the delay time decreases as the location of the failure nears the receiving end. However, the delay time gradually increases when a failure occurs near the receiving end. Therefore, by partially modifying FIG. 2 and dividing it as shown in FIG. 3, it is possible to reduce the transmission delay that occurs in the case of a failure occurring at the same location.
3 is an example of a propagation delay of a failure indication signal (FIS) in a segment-based network according to the present invention. Referring to FIG. 3, it can be seen that the global network of FIG. 2 is divided into two segments, and the W4 and W5 nodes are overlapped in the two segments. That is, the first segment is composed of nodes W1, W2, W3, W4, and W5 and nodes B1, B2, B3, and B4, and the second segment is composed of nodes W4, W5, W6, W7, W8, W9, and B5, B6, It consists of B7 nodes.
In FIG. 3, the reason for overlapping two nodes (W4, W5) between two segments is to consider the failure of the node in addition to the failure of the link generally considered. If a problem occurs, it cannot be processed.

delete

In addition, if automatic recovery is performed in units of segments, distributed processing for multiple failures is possible, thereby reducing the time for automatic recovery. As shown in FIG. 2 and FIG. 3, as the scope of failure recovery is reduced, the time that occurs in failure recovery can be reduced.

In fact, the smaller the size of the segment, the shorter the message transmission delay time, so it can be clearly seen that the automatic recovery time is reduced. However, dividing the size of the segment too small will result in more nodes being used as alternative paths, which leads to large losses in terms of resource utilization, and further complicates the mechanism of automatic recovery. Therefore, transmission delay time and resource utilization rate are inversely related to each other, and thus trade-off is required according to the characteristics of the communication network.

delete

Considering the propagation delay that has the greatest impact on the failure recovery time in order to present a quality-based differentiated failure recovery technique, when applying the segment-based failure management system, the number of nodes is within 4 km and 250 km. The network is divided and managed to distribute and manage the network within 25ms.
In the present invention, a method for setting quality guaranteed differentiation failover is as follows.
In order to set up the quality-differentiated failure recovery scheme, it is necessary to first determine the failure recovery option based on the service level agreement (SLA).
One. First, a method of configuring 1 + 1 path protection will be described.
Since the 1 + 1 path protection technique is not provided in the MPLS router, the same data is transmitted through the ingress node (W1). In FIG. 4, Data (1) and Data (2) are the same data, and Data (1) and Data (2) set the router such that only a working path exists without a backup path.
4 is an example of a 1 + 1 path protection configuration according to the present invention. In FIG. 4, W5 should function as a PML, but since W5 does not have a PML function, it should be processed in software. Even if a failure occurs in the working path of Data (1), data continues to be transmitted through the working path of Data (2), so there is no packet loss even if a single link failure occurs.
2. How to set 1: 1 path protection is as follows.
The 1: 1 path protection method includes a stand-by mode, a path-option mode, and a fast rerouote method. For all three options, backup should be set to disjoint for working. If the working path is Tunnel 100 and the backup path is Tunnel 200, the standby mode ( stand-by mode) is set when Tunnel100 is set as backup interface Tunnel 200 . Stand-by mode uses a CLI command called tunnel mpls traffic-eng path-option . For example, if you set "tunnel mpls traffic-eng path-option 100 explicit name working 100", the working path is a path option of 100 and the explicit path of work100 is 100. This is the path to the tunnel with (path-option) 100.
3. The method of setting 1: N and M: N path protection is as follows.
1: N and M: N are used by a plurality of working paths sharing a backup path. 1: N and M: N protection can be set using CLI commands in stand-by mode and path-option mode. Stand-by mode designates a common backup tunnel for the backup interface, and path-option mode sets the same explicit path name. You can set 1: N or M: N.
4. Subnet-based segment recovery function setting method is as follows.
FIG. 5 is an example of a subnet-based failure recovery scheme in which a global network is segmented into four or two subnets according to the present invention. (A) of FIG. 5 is an example of a global network consisting of 16 nodes, (b) is a segmentation of the global network into four subnets in consideration of the number of nodes and links, and (c) Each recovery option is applied by segmenting the global network into two subnets, taking into account the number of nodes and links.
6 is a class diagram of a failure module management object according to the present invention, and FIG. 7 is a sequence diagram illustrating a failure recovery procedure according to the present invention.
In the present invention, as shown in FIG. 6, the detailed fault management module includes a helper man server socket (DoumiMan ServerSocket), a helper man (DoumiMan), an MPLS network (MPLS Network), a fault handler (Faultandler), and an MPLS LSP list (MPLS_LIST). It consists of MPLSLER class.
The fault handler management object (FaultHandler MO) receives fault-related information and information about performance degradation in the hierarchical network, records fault information, and returns to the network state before the fault after the test of the fault recovery and the recovery of the fault point. It is supposed to be. In addition, the fault handler management object (FaultHandler MO) locates the fault through a callback function and stores a fault log in a database.
The MPLS Network MO manages the Management Objects (MOs) for nodes and links in the MPLS area. The MPLS Network MO manages the LSP and sets the backup path. Setting recovery options according to traffic characteristics is set through the createLSP () method of MPLSNetwork MO.
In the 1: N protection switching method, back up the LSP that has not been processed by error after reading the LSP group ID information applied at the time of LSP setup using the LSP object reference. Check if LSP (backup LSP) is being used in LSP Group, and in MPLS Network, access specific working LSP MO and process according to the failover mechanism of the management object (MO). Will be done.
Even in M: N protection switching, if the LSP group does not use all M backup LSPs, it accesses a specific working LSP MO and accesses the corresponding working LSP MO. (MO) will be handled according to the failure recovery mechanism.
Differentiation fault recovery techniques of the present invention are shown in Table 3 below.
Option to select recovery time of SLA (example) Recovery option (example) Requirement option 10 msec 1 + 1 Optical Protection Link / Path Level Protection MPLS Fast Reroute (cisco) Local Protection (if L = 150, Pd <5ms) 30 msec 1: 1 protection Link / node protection, segment network (Sn = 3, L = 150, Pd <15ms) MPLS Fast Reroute Local Protection (L = 200, Pd = 10ms) 50 msec 1: 1 protection 1: 1 SONET / SDH Local Protection (Pd = 25ms) 100 msec 1: 1,1: N, M: N Protection LocalProtection (Pd = 50ms) 200 msec 1: 1,1: N, M: N Protection Path protection / Global protection segment network (Sn = 3, L = 300, Pd = 9ms) 500 msec 1: 1,1: N, M: N Protection path protection (Pd <100ms) 1sec MPLS TE LSP Rerouting (Dynamic) -

delete

delete

delete

delete

delete

delete

delete

delete

delete

delete

delete

delete

delete

delete

delete

delete

delete

delete

delete

Table 3 shows a failure recovery scheme according to recovery time. Recovering from a failure within 10msec is not an easy task in a real network. Therefore, the resource efficiency is the lowest to recover from a failure within 10msec, but the data can be transmitted for a single link failure because the same data is transmitted by different paths even if a failure occurs by using 1 + 1 optical protection. No loss
However, in identifying the fault and analyzing the cause, the administrator is notified of the fault in consideration of the propagation delay and the fault detection time. Cisco's fast reroute failover scheme guarantees failover performance within 50ms in a real network. As a result of performance measurement in the test bed without considering the actual link, when the 1000byte packet is transmitted with guaranteed bandwidth of 1Mbos, the failure recovery performance is within 5ms.
Therefore, if the propagation delay is guaranteed only within 5ms, local protection is possible within 10ms. The 1: 1 path protection recovery option can recover from failures within 30ms by segmenting the network to minimize propagation delay, and use the MPLS Fast Reroute technique. If the performance of propagation delay is less than 15ms, failover is possible within 30ms.
A failure recovery option having a failure recovery capability of 200 msec to 1 sec can be applied by managing a network by segmenting path protection rather than local protection. Since all the traffic in the network is not data that is sensitive to failures, the efficiency of network resources is improved by proposing differentiation recovery methods according to traffic characteristics.
Class Type Recovery option importance purpose of use EF 1 + 1 Very high Company business, e-commerce, business meetings 1: 1 height Company general business M: N middle For regular phone calls 1: N Normal For regular phone calls (cheap toll) Dynamic routing - Private Private Calls with Cheap Calls
Class Type Recovery option importance purpose of use AF3 1: 1 height electronic commerce M: N - High speed internet 1: N - Slow internet Dynamic routing - Personal web surfing, P2P
Table 4 is a Case 1 -differentiated fault restoration for VoIP, and Table 5 is a Case 2 -differentiated fault restoration for Web application.
With respect to the failure recovery options that can be applied according to the recovery time through Table 3, the service provider may apply a failure recovery scheme according to the purpose as shown in Table 4 and Table 5 for each service type.
Table 3 above presents differentiated failure recovery schemes when using VoIP services. VoIP is sensitive to delay and jitter so that it can be carried in EF or AF4 class-type traffic, and the failover option is set differently according to the purpose and importance of VoIP.

delete

delete

delete

delete

delete

delete

delete

delete

delete

delete

As described above, according to the present invention, the differentiation / differentiation in the failure management function may provide differentiation in the recovery ability in the event of a failure in consideration of the importance of the service quality provision level and the user's cost burden.

Claims (4)

In the quality assurance network management method in a multiprotocol label switching (DiffServ-over-MPLS) global network environment supporting multiple services, Setting a failback option by a service level agreement (SLA); Segmenting the global network into a predetermined number of sub-networks such that the number of nodes is 4 to 5, the distance is within 250 km, and the propagation delay time is within 25 ms in consideration of the number of nodes and the number of links; And 1 + 1 path protection scheme for transmitting the same data through different paths using the subnetwork according to the failure recovery option by the service level agreement (SLA), 1: 1 path protection method, or 1: N, M Differentiating failover with any of the: N path protection methods Failure recovery method according to traffic characteristics, characterized in that the propagation delay is minimized during failure detection. The method of claim 1, wherein the subnetwork Failover method according to the traffic characteristics, characterized in that it is possible to cope with node failure by overlapping a few nodes. delete The method of claim 1, wherein the 1: 1 path protection method is A failure recovery method according to traffic characteristics, characterized in that the standby mode (stand-by mode), the path-option mode (path-option mode) and fast rerouote (Fast rerouote).

Images