JP2012209660A - Path setting method considering effect of data compression in network - Google Patents

Abstract

PROBLEM TO BE SOLVED: To make optimum path settings in a network possible by appropriately expressing the network by taking into account the effects of data compression by network equipment included therein which has data compression/decompression functions.SOLUTION: Nodes 14, 15 which include network equipment having data compression/decompression functions calculate the sum of transmission delays in links 21, 23 connected to the respective local nodes and processing delays due to data compression processing as delays in the links 21, 23 and make the delay a metric of the links 21, 23. Nodes 11-13 which include routers calculate transmission delays in links 21-25 connected to the respective local nodes as delays in the links and make the delays a metric of the links 21-25. Nodes 11-15 share the metric of each of the links 21-25 and set an optimum path based on the metric.

Description

  The present invention relates to a path setting method in consideration of the effect of data compression in a network, and in particular, even when a network device having a data compression / decompression function such as a WAN acceleration device is included in the network, the data compression in the network device is performed. The present invention relates to a path setting method capable of managing a network in consideration of the effects of the above and setting an optimum path.

  In a conventional network, an optimum path is set based on bandwidth information (available bandwidth) of each link constituting the network. For this purpose, each node (router) in the network obtains and stores the bandwidth information of each link connected to the node, advertises the bandwidth information to the network, and advertises the bandwidth advertised by other nodes. Store information. Each node in the network can share network bandwidth information and set an optimum path based on the bandwidth information. Non-Patent Documents 1 and 2 describe the standardization of such network management technology.

  In recent years, in addition to conventional routers, network devices (hereinafter referred to as WAN acceleration devices) having a data compression / decompression function such as WAN (wide area network) acceleration devices are partially arranged in the network. A case is coming out. The WAN acceleration device is described in Non-Patent Document 3.

  FIG. 4 is a block diagram schematically showing a network in which WAN acceleration devices are partially arranged. Here, a network 10 configured by connecting routers 11 to 13 and WAN acceleration devices 14 and 15 via links 21 to 25 is shown. The WAN acceleration devices 14 and 15 are connected to the links 21 and 23. Users 31 and 33 use data compression, but users 32 and 34 do not use data compression.

  When a WAN acceleration device is included in the network, data transferred through the WAN acceleration device is compressed / expanded there. The WAN acceleration device is arranged at an appropriate user side end in the network, and the data compression rate provided by each WAN acceleration device varies. This prevents the network from being properly represented by link bandwidth information as in the prior art, and therefore prevents the network from being properly managed. The reason will be described later.

  As described above, in the conventional technology, the effect of data compression in the WAN acceleration device cannot be appropriately expressed as a distance in the network on data transmission, that is, “metric”, and as a result, the WAN acceleration. The network must be managed ignoring the effect of data compression on the device. “Metric” is a technical term representing “distance in the network on data transmission”, and the smaller the metric value, the shorter the distance in data transmission in the network and the lower the communication cost. . That is, the value of “metric” represents the communication cost.

RFC3630 (Traffic Engineering (TE) Extensions to OSPF Version 2) RFC53058 (IS-IS Extensions for Traffic Engineering) WAN acceleration (http://www.cisco.com/web/JP/news/cisco_news _letter / tech / waas2 / index.html) M. Shimamura, H. Koga, T. Ikenaga, and M. Tsuru, "Compressing packets adaptively inside networks," IEICE Transactions on Communications, vol. E93-B, no. 3, pp. 501-515, March 2010.

  Data compression in the WAN acceleration device has the effect of effectively increasing the bandwidth of the link. Therefore, as in the prior art, it is conceivable to manage the network by expressing link bandwidth information as a metric. Specifically, conventionally, since an OSPF (open shortest path fast) protocol such as RFC3630 is used, a method of applying OSPF to a WAN acceleration device can be considered.

  However, in the method of managing the network based on the bandwidth information of the link, for example, it is not meaningful to advertise only the link 21 and the link 23 in FIG. In other words, WAN acceleration has a feature of functioning end-to-end, in which the data compression side and the data decompression side function for the first time as a pair. Therefore, the WAN acceleration on the data compression side and the data decompression side It is necessary to set the bandwidth amount of all the links constituting the path through which the data to be processed take into consideration the data compression in the WAN acceleration device.

  For example, in FIG. 4, it is necessary to change the bandwidth of the link advertised by the routers 11 to 13 for the link 22 or the links 24 and 25 in accordance with the data compression in the WAN acceleration device 14 (or the WAN acceleration device 15). Comes out.

  In an actual network, there are often routers between the paired WAN acceleration devices, and the routers operate independently of the WAN acceleration devices and respond to data compression in the WAN acceleration devices. The data is not compressed. Therefore, according to the data compression in the WAN acceleration device, even if the change in the bandwidth of only the link connected to it is advertised, the bandwidth of the other links is not changed. It will not be expressed and will not be properly managed.

  Considering the case where several pairs of WAN acceleration devices are arranged and the traffic of users (users 32 and 34 in FIG. 4) that are not subjected to data compression in the WAN acceleration device, it is assumed that WAN acceleration will be performed. Even if OSPF is installed in the device, it is difficult to solve the problem. In other words, there are many difficulties in trying to manage a network by expressing the effect of data compression in the WAN acceleration device as link bandwidth information.

  The object of the present invention is to solve the above-mentioned problems, and even if a network device having a data compression / decompression function such as a WAN acceleration device is included in the network, the network is considered in consideration of the effect of data compression in the network device. It is an object of the present invention to provide a path setting method that can be appropriately expressed and managed and can set an optimum path.

  In order to solve the above problems, the present invention is a path setting method that takes into account the effect of data compression in a network including a link in which network devices having data compression / decompression functions are arranged, and has a data compression / decompression function. The node including the network device calculates the sum of the transmission delay of the link connected to the own node and the processing delay due to the data compression processing at the own node as the delay for the link, and calculates the delay as the metric of the link. A node including the router calculates a transmission delay of the link connected to the node as a delay for the link, calculates the delay as a metric of the link, and includes a node including the network device and a node including the router. , Share and manage the metrics of each link, and the optimal path based on the metrics It is basically characterized in that setting.

  In the present invention, since the metric about the link between each node is expressed as a delay, even if a network device having a data compression / decompression function such as a WAN acceleration device is included in the network, data in the network device is included. It is possible to appropriately express and manage network metrics including the effect of compression, and to set an optimal path based on the metrics.

It is explanatory drawing which shows notionally the path | pass setting method in consideration of the data compression in the WAN acceleration device arrange | positioned in the network by this invention. It is a flowchart which shows an example of the metric calculation operation | movement in each node. It is a flowchart which shows the other example of the metric calculation operation | movement in each node. 1 is a block diagram illustrating an example of a network in which WAN acceleration devices are partially arranged. FIG.

  For applications and services, the effect of data compression in the WAN acceleration device is conscious of the aspect that greatly varies the transfer delay. Therefore, in the present invention, the delay (processing delay) caused by the data compression processing in the WAN acceleration device is considered. The above problem is solved by grasping the variation amount as a metric.

  FIG. 1 is an explanatory diagram conceptually showing a path setting method considering data compression in a WAN acceleration device arranged in a network according to the present invention. In FIG. 1, the same or equivalent parts as in FIG.

  FIG. 1 shows a network 10 similar to FIG. The network 1O is configured by connecting routers 11 to 13 and WAN acceleration devices 14 and 15 via links 21 to 25. The WAN acceleration devices 14 and 15 are connected to the links 21 and 23. Users 31 and 33 use data compression, but users 32 and 34 do not use data compression.

  Since the routers 11 to 13 and the WAN acceleration devices 14 and 15 each constitute a node, they are hereinafter referred to as nodes 11 to 15. The nodes 11 to 13 are nodes including a router, and the nodes 14 and 15 are nodes including a WAN acceleration device.

  The node 11 measures the transmission delay of the links 21, 22, and 24 connected to the own node, stores the transmission delay as the metric of the links 21, 22, and 24, and advertises the metric to the network 10. Further, the metrics of links advertised from other nodes 12 to 15 are stored.

  The node 12 measures the transmission delay of the links 22, 23, 25 connected to the node, stores the transmission delay as a metric of the link 22, 23, 25 and advertises the metric to the network 10. Further, the metrics of links advertised from other nodes 11 and 13 to 15 are stored. Here, the metric for the link 22 calculated by the node 13 overlaps with the one advertised by the node 11, but since both are the same value, no problem arises even if they are stored.

  Similarly, the node 13 stores the metrics of the links 24 and 25, advertises the metrics to the network 10, and stores the metrics of the links advertised by the other nodes 11, 12, 14, and 15. Again, the metrics for the links 24, 25 calculated by the node 13 overlap with those advertised by the nodes 11, 12, but no problem arises.

  On the other hand, the node 14 measures the transmission delay of the link 21 connected to the own node, and the delay due to the data compression processing in the WAN acceleration device of the own node (processing delay), and the transmission delay and the transmission of the link 21 The sum of delays is calculated, and this sum is calculated and stored as the metric of the link 21. The metric is advertised to the network 10, and the metric of the link advertised from the other nodes 11 to 13 and 15 is stored. Here, the metric of the link 21 calculated by the node 14 is different from that advertised by the node 11. That is, the metric calculated by the node 14 includes a processing delay due to the data compression processing in the WAN acceleration device, whereas the metric calculated by the node 11 does not include the processing delay. , Both are wrinkled. For this defect, if a different metric is advertised for the same link, the smaller metric may be adopted.

  In addition, the node 15 measures the transmission delay of the link 23 connected to the own node, and the delay (processing delay) due to the data compression processing in the WAN acceleration device of the own node, and the transmission delay and the transmission of the link 23 The delay sum is calculated and stored as the metric of the link 23, the metric is advertised to the network 10, and the metric of the link advertised from the other nodes 11 to 14 is stored. In this case as well, there is a discrepancy between the metric of the link 23 calculated by the node 15 and the metric of the link 23 advertised from the node 12, but it may be dealt with in the same manner as described above.

  The transmission delay of each link, the processing delay due to the data compression processing in the WAN acceleration device, and the metric calculation method will be specifically described later.

  In FIG. 1, the node 11 calculates and advertises the metrics “0.03”, “0.31”, and “0.25” of the links 21, 22, and 24, and the node 12 calculates the metrics “0.31” and “0.06” of the links 22, 23, and 25. ”,“ 0.11 ”is calculated and advertised, node 13 calculates and advertises metrics“ 0.25 ”and“ 0.11 ”of links 24 and 25, and node 14 calculates metric“ 0.02 ”of link 21 and advertises In this example, the node 15 calculates and advertises the metric “0.05” of the link 23.

  Here, “0.02” is advertised from the node 31 as the metric of the link 21, while “0.03” is advertised from the node 11, and the smaller “0.02” is the metric of the link 21. Similarly, the metric of the link 21 is “0.05”.

  As described above, the metrics of the links 21 to 25 calculated by the nodes 11 to 15 are advertised to the entire network 10 by a protocol such as OSPF. As a result, the nodes 11 to 15 are connected to the links 21 to 25 of the network 10. Can share metrics. In FIG. 1, the metrics of the links 21 to 25 stored by the node 15 are shown here, but the other nodes 11 to 14 also store the same metric.

  Based on the metrics stored by each node 11-15, the path of the shortest path (the path with the minimum delay, that is, the path with the minimum sum of metrics (the path with the lowest cost)) can be calculated by the Dijkstra method, for example. . In the example of FIG. 1, the shortest path from the node 14 (user 31) to the node 15 (user 33) is a path via the link 21-link 24-link 25-link 23 in which the metric sum is 0.43. An optimum path can be set for this route.

  Next, a delay and metric calculation method for each link will be described. Below, the bandwidth of the output link of the WAN acceleration device is B [b / s], the data packet size is S [b], the time required for data compression is C [s], and the data compression / decompression ratio is R And

  According to Non-Patent Document 4, the maximum value PR of the ratio of compressible packets (for example, if PR = 0.9, the maximum 90 packets out of 100 can be compressed) is x when the output queue is congested. If one packet is compressed, “(x−1) * S / B + R * S / B = C” is satisfied. From this, “x = B * C / S + (1-R)” is derived. Since the ratio PR of the number of compressible packets is the reciprocal of x, it can be expressed as “PR = 1 / (B * C / S + (1-R))”. WAN acceleration devices typically try to compress packets at a PR rate.

  By using the above, it is possible to estimate the processing delay due to the data compression processing in the WAN acceleration device. In other words, if the time required for packet transfer when data compression is not performed in the WAN acceleration device is W, data compression is applied to the ratio of PR in W, so one WAN acceleration device The processing delay due to the data compression processing can be estimated as “R * PR * W + (1−PR) * W”. As is clear from the above formula, there is a relation that the processing delay increases as the data compression / decompression ratio R in the WAN acceleration device increases.

  By using the above formula to express a metric with processing delay corresponding to the degree of data compression in the WAN acceleration device, it is possible to manage the network including the effect of data compression in the WAN acceleration device Become.

  Traditionally, OSPF has treated the inverse of the bandwidth of a link as a metric, but instead treats the delay for each link as a metric with the same value as it is, using the same algorithm as OSPF, with low end-to-end delay. The route can be calculated. In other words, the metric is expressed by the processing delay due to the data compression processing in the WAN acceleration device and the transmission delay of the link, and the network is appropriately added by adding it as, for example, RFC 3630 Sub TLV (time length value). It becomes possible to manage.

  FIG. 2 is a flowchart illustrating an example of a metric calculation operation in each node. Here, a basic algorithm is shown in which each node calculates a delay for a link connected to the node, and the delay is a metric for the link. The delay calculated here can be handled as a metric for each link by a protocol such as OSPF. The configuration for calculating the metric can be realized by hardware or software.

First, Ping measurement is started (S21). The transmission delay (D _L ) of the link is acquired by Ping measurement (S22). The link transmission delay (D _L ) is preferably an average value of measurement values obtained by periodic measurement. The transmission delay (D _L ) in the link does not include a processing delay due to data compression processing in the WAN acceleration device.

  Next, whether or not the link is a link connected to a node including the WAN acceleration device and the node for calculating the metric is a node including the WAN acceleration device (in FIG. 2, “link for data compression”). ("?") Is determined (S23). In the example of FIG. 1, S23 is Yes in the case of metric calculation of the links 21 and 23 at the nodes 14 and 15, and S23 is No in the case of metric calculation at the nodes 11 to 13.

When S23 is Yes, since the link is a link connected to a node including the WAN acceleration device and the metric calculation is performed at the node including the WAN acceleration device, first, data compression in the WAN acceleration device is performed. A processing delay (D _W ) due to the processing is calculated (S24). The processing delay (D _W ) in data compression of the WAN acceleration device uses the compression / decompression ratio (R) set in the WAN acceleration device, and the above formula “R * PR * W + (1−PR ) * W ". Note that the bandwidth B [b / s] of the output link of the WAN acceleration device, the packet size S [b] of the data, the time C [s] required for data compression, and the packet transfer when data compression processing is not performed The time W required for the above is known, and PR and W of the above formula can be obtained from these values. Next, as the delay for the link, the sum of the link transmission delay (D _L ) and the processing delay (D _W ) due to the data compression processing in the WAN acceleration device is calculated, and the sum is stored as the metric of the link (S25).

On the other hand, when S23 is No, since the metric is calculated at the node including the router, the link transmission delay (D _L ) is calculated and stored as the link metric (S26). If different metrics are calculated and advertised for the same link by the above flow, the smaller metric may be adopted as described above.

  As described above, the delay, which is the effect of data compression in the WAN acceleration device, is handled automatically as a metric using a protocol such as OSPF, so that the network including the WAN acceleration device is optimally managed and the optimal path is set. It becomes possible to do.

FIG. 3 is a flowchart showing another example of the metric calculation operation in each node. In FIG. 2, the transfer delay at the router is not taken into account when calculating the metric, but in FIG. 3, the metric is calculated taking into account the transfer delay (D _R ) at the router.

  S31 to S35 in FIG. 3 are the same as S21 to S25 in FIG. In the example of FIG. 1, the nodes 14 and 15 including the WAN acceleration device calculate the metrics of the links 21 and 23.

When S33 is No, the metric is calculated at the node including the router. Therefore, the sum of the transmission delay (D _L ) of the link and the transfer delay (D _R ) at the router is calculated as the delay for the link. Is calculated and stored as the metric of the link (S36). Here, since there are two links on the input side and the output side in packet input / output at the router, one half of the transfer delay (D _R ) is allocated to the two links on the input side and the output side. 1/2 of the transfer delay (D _R ) at the routers on both sides of the link is added, and as a result, the sum of the link transmission delay (D _L ) and the transfer delay (D _R ) at the router is the delay for the link. It is said. As the transfer delay (D _R ) in the router, for example, a preset value such as an average of transfer delays in a plurality of routers constituting the network is used.

Note that when a different metric is advertised for the same link, the smaller metric is used, and as a result, the router speed is used for the link that takes into account the processing delay (D _W ) in the data compression of the WAN acceleration device. Although the delay (D _R ) is not taken into consideration, the transfer delay (D _R ) at the router is sufficiently smaller than the processing delay (D _W ) due to the data compression processing in the WAN acceleration device. So this doesn't cause any problems.

Although the embodiment has been described above, the present invention is not limited to the above embodiment. For example, in FIG. 3, the transfer delay in the router is not taken into account for the link to which the processing delay due to the data compression processing in the WAN acceleration device is added, but the processing delay due to the data compression processing in the WAN acceleration device is not included. Even for the added link, 1/2 of the transfer delay (D _R ) at the router can be added. In addition, if a mechanism for notifying the transfer delay in the local router from the router to the adjacent router or adjacent WAN acceleration device is provided, the delay and metric for each link can be calculated using the notified transfer delay. It becomes possible.

10 ... Network, 11-13 ... Node including router, 14,15 ... Node including WAN acceleration device, 21-25 ... Link, 31-34 ... User

Claims (4)

A path setting method considering the effect of data compression in a network including a link where a network device having a data compression / decompression function is arranged,
A node including a network device having a data compression / decompression function calculates a sum of a transmission delay of a link connected to the own node and a processing delay due to a data compression process at the own node as a delay for the link, and calculates the delay. Calculated as the metric for that link,
A node including a router calculates a transmission delay of a link connected to the node as a delay for the link, calculates the delay as a metric of the link,
A path in consideration of the effect of data compression in a network, wherein a node including the network device and a node including a router share and manage a metric of each link and set an optimal path based on the metric. Setting method.   The bandwidth of the output side link of the network device is B [b / s], the data packet size is S [b], the time required for data compression is C [s], the data compression / decompression ratio is R, the network device When the time required for packet transfer when data compression is not performed in W is W, the maximum value PR of the ratio of compressible packets calculated by PR = 1 / (B * C / S + (1-R)) The processing delay due to the data compression processing in the network device is calculated by R * PR * W + (1−PR) * W using the above, and the data compression effect in the network according to claim 1 is obtained. Path setting method in consideration.   The path setting considering the effect of data compression in the network according to claim 1 or 2, wherein a node including the router calculates a delay for each link, further taking into account a transfer delay in the router. Method.   5. The data compression effect in the network according to any one of claims 1 to 4, wherein a metric having a smaller value is adopted when metrics having different values are calculated for the same link. Path setting method.

Images