Patent Application
20080056142
Disclosed are Test methods for message paths in communication networks and network elements. Also disclosed are redundant network arrangements.
Highly reliable communications systems often use redundant message paths to ensure that a fault affecting an individual message path does not lead to restrictions in communication. At the same time the redundancy of the message paths, i.e. for each message path there exists at least one alternate message path to which communication can be switched in the event of a fault, must be supported by the service platforms or hosts as well as by the communications system itself, i.e. by its elements, e.g. switches and routers, and its structure.
Moreover, for communications systems with real-time requirements, for example in the case of voice communication, very fast switchover times from a faulty message path to an alternate message path are also very important in order to limit to a minimum the negative effects on operation in the event of failure of a message path.
Faults to be taken into account include total failures and/or partial failures in individual elements of the communications system, e.g. service platform, switches, routers, and failures of the connections between the individual elements.
A communications system very often encountered in practice includes one or more hosts or service platforms that are connected to an IP network (IP=Internet Protocol) via a redundant local network LAN (LAN=Local Area Network) and two gateways.
The following means of checking message paths for freedom from faults are typically used:
IP Networks (Layer 3 Switching):
For the logical protocol level of the IP networks there exist standardized routing protocols such as e.g. Open Shortest Path First OSPF, Routing Information Protocol RIP, Border Gateway Protocol BGP, by means of which failures of a path can be detected and reported to other network elements in order to initiate a switchover to alternate routes. In this case the topology of the IP networks plays an insignificant role. The interruption of a message path which is connected directly to a network element is usually detected very quickly, e.g. inside 60 ms, and the switchover is typically completed after a few seconds, e.g. within 1.4 s.
The interruption of a message path which is not connected directly to the network element can only be communicated and detected by means of a routing protocol. In this case the switchover times are usually much greater and lie, for example, in the range of 30 s. to 250 s.
Local Area Networks LAN (Layer 2 Switching):
For the logical protocol level of the LANs there is no standardized procedure for detecting faulty message paths especially with redundant configurations with the structure referred to. In order to monitor host—LAN—gateway connections, the Spanning Tree Protocol SPT can be used, for example.
The SPT protocol is very slow-acting, however, i.e. a considerable period of time, for example about 30 s, is typically required in order to define a suitable alternate path. For this reason efforts are being made to introduce a faster form of SPT, called the Rapid Spanning Tree Protocol RSPT, which is described in IEEE Standard 802.1w. However, the monitoring times for RSPT are still in the range of several seconds (default value for bridge hello time=2 s).
For LANs with a ring topology, solutions are known, e.g. Ethernet Automatic Protection Switching EAPS or Resilient Packet Ring RPR, by means of which very short switchover times, e.g. less than 1 s, are to be achieved. However, all these methods use a LAN with ring topology, which is not the case in all application scenarios.
Considering the known methods for checking message paths described in the foregoing, the following problems result:
One embodiment of the present invention specifies a test method for message paths in communications networks as well as an improved network element, by means of which the disadvantages of the prior art are avoided.
One aspect of the present invention is a test method for message paths which can advantageously be used if two devices exchange messages of a first protocol layer, for example IP packets, via a communications network of a lower protocol layer, for example a LAN, the messages exchanged between the devices via the communications network being transmitted transparently, i.e. unmodified, through the communications network. According to the invention, a device initiating the test method sends test messages of the first protocol layer, e.g. special. IP packets, at short time intervals, the address of the first protocol layer, e.g. the IP address, of the initiating device being selected for such test messages both as the send address and as the receive address. It is also possible that the test method is executed by both devices, with the result that both (terminal) devices of a communications relationship know the status of the message paths.
A major advantage of the invention is that the test messages sent by the first to the second device are processed not by the switching processor of the second device, but already by the interface unit of the second device. In this way the test messages, which are sent frequently, for example every 100 ms, in order to detect faults on message paths as swiftly as possible, are prevented from generating processor load in the second device.
In a preferred embodiment, in which message paths of a LAN between a host and a gateway are tested, there is therefore an important advantage in the fact that the link test according to the invention does not lead to an overload situation at the gateway. In conventional implementations, PING or route-update messages and RIP messages are used at time intervals of 30 s to 300 s, as a result of which fast detection of faulty message paths, which is typically preferred for voice communication for example, is not possible. The use of the known ICMP PING or RIP messages would lead to overload if these messages were to be, sent at the high frequency mentioned, i.e. several times per second for each message path, when many hosts are connected.
By means of a timer it can advantageously be monitored whether the test messages were received correctly and within an expected time interval that is in line with the expected message transit time in the communications network via the message paths via which the test messages were sent. If test messages are not received or are received after the timer has elapsed, there is probably a fault on the corresponding message path. So that the loss of individual test messages does not lead to the false assumption that there is a general failure of the respective message path, the loss of multiple test messages can be used as a criterion for a fault on the message path.
The information concerning the faults on individual message paths can advantageously be used to select the optimal remaining message path in each case. Here, the optimal message path can be selected according to the chosen topology of the participating networks and taking into account factors such as costs associated with individual message paths and number of redundant interfaces or devices present.
The invention requires no modifications to be made to components of the communications network and can therefore be implemented easily and cheaply. Its realization is therefore simple and concerns only the device initiating the test.
Also provided according to the invention is a network element comprising means for executing this test method.
The present invention is also directed to a redundant network arrangement which advantageously allows for swift detection of faulty message paths and fast switchover to fault-free message paths.
The present invention is further directed to a redundant network arrangement which can be used with physically very remote network elements. At the same time the network arrangement incorporating long-distance or wide-area connections is intended to allow swift detection of faulty message paths and fast switchover to fault-free message paths.
According to the present invention, a network arrangement for a communications network N1 which connects a first device Host and a second device G0, is provided,
A major advantage of the invention is to be seen in the fact that when multiple devices Host are connected to the second device G0 by means of the network arrangement N according to the invention, each device Host has two redundant message paths to the second device G0 via two interfaces IF0, IF1. In this arrangement, one of the message paths runs via the crosslink Q1 between the two redundant subnetworks, while the other runs within a subnetwork.
In a preferred embodiment, in which the message paths are formed by a network N between a host and a gateway G0, second gateway G1 can advantageously be used for reasons of reliability. This avoids the failure of the default gateway G0 leading to isolation of the entire network N.
In combination with the second gateway G1, multiple message paths advantageously result, said message paths enabling communication between hosts and at least one of the gateways G0, G1 even in the event of problems on individual message paths due to faulty connections or faulty switching elements.
A further advantage is that multiple hosts can communicate with one another by means of the crosslink(s) Q1 between the subnetworks N0 and N1 independently of the gateways, and furthermore can also do so when different interfaces of the hosts are active. For example, a first host with first active interface, connected to the first subnetwork N0, can exchange messages with a second host with second active interface, connected to the second subnetwork N1, via the crosslink(s). This would not be possible without the crosslink according to the invention.
Compared to the solutions in which only local area networks LAN are used in order to connect the first device Host and the further devices G0, G1, the use of wide area networks (WAN) according to one aspect of the invention allows much greater physical distances between the devices mentioned. This is of advantage, for example, when one of the redundant gateway devices G0, G1 is set up at a remote location, e.g. in order to reduce costs and to increase security and/or availability.
It is further of advantage that the network arrangement according to the invention considerably simplifies the administration of the overall network, since many hosts distributed over great areas can be reached from the centrally located gateway devices G0, G1 via only a single IP subnetwork. This minimizes the probability of an administration error and increases reliability.
In order to check the message paths, an advantageous test method for-message paths in communications networks can be used without modifications, since the long-distance (WAN) segments of the communications network forward the frames or packets of the networks N0, N1 or N01, N02, N11, N12 that are to be transported, transparently and so the end-to-end test of the paths between host and gateway(s) G0, G1 is not affected.
The invention is explained in greater detail below as an exemplary embodiment with reference to three figures.
The invention will be better understood by reference to the Detailed Description of the Invention when taken together with the attached drawings, wherein:
With reference to
The host is connected via a communications network N to a second device G0. This second device may, for example, be one of the gateways referred to in the introductory remarks. However, the second device can likewise be any communications device having L3 communications capabilities. For simplicity, the name Gateway will be used below to designate the second device.
In the preferred exemplary embodiment, the communications network N is a local area network LAN which operates e.g. according to the Ethernet standard. Other networks and/or protocols can be used for the transparent message transport between host and gateway.
Without special knowledge of the communications network N or its topology, the invention is already suitable for testing the message path or message paths via the communications network. However, the topology presented below is particularly suitable for use with the invention, particularly with regard to the possible alternate message paths in the event of a fault.
The communications network N is subdivided into two independent subnetworks N0, N1. In the simplest case this subdivision is implemented at logical level, but is also advantageously carried out physically in order to provide the greatest possible fault tolerance. In this scenario, N0 includes a number of switching components or switches S00, S01, S02. Three switching components are shown, although this number is purely exemplary and arbitrary from the point of view of this invention, in the same way as the structure of the subnetwork N0 is arbitrary, being represented as linear only as an example.
The switches S00, S01 are connected by means of a link L01, this link standing as representative of a logical, bidirectional connection between the switches; it can be formed physically, for example, by multiple links. In the same way the switches S01, S02 are connected by means of a link L02.
Subnetwork N1 includes a number of switching components or switches S10, S11, S12. Three switching components are shown, although this number is simply an example and arbitrary from the viewpoint of this invention, in the same way as the structure of the subnetwork N0 is arbitrary, being represented as linear only by way of example. The switches S10, S11 are connected by means of a link L11, this link standing as representative of a logical, bidirectional connection between the switches and can be formed physically, for example, by multiple links. In the same way the switches S11, S12 are connected by means of a link L12.
N0 is connected to the host via a link L00. N1 is connected to the host via a link L10. Here, the host has two separate interfaces IF0, IF1, a first interface IF0 serving the connection to subnetwork N0 and a second interface IF1 serving the connection to N1.
A link L03 serves to connect subnetwork N0 to the gateway G0. Depending on the type of redundancy topology, subnetwork N1 likewise possesses a connection to gateway G0—not shown—and/or, via at least one crosslink Q1, to subnetwork N0. Advantageously, this crosslink is implemented as closely as possible to the transition point from N0 to the gateway G0, i.e. for example between S02 and S12 as shown in
In an alternative embodiment, a standby gateway G1—represented by dashes—is provided in addition to the gateway G0, for example in case of the failure of the gateway G0. Here, the gateways G0, G1 can likewise be connected by means of a crosslink Q2. A link L13 connects N1 and gateway G1. Depending on the type of redundancy topology, N0 likewise possesses a connection to gateway G1—not shown.
The gateways G0, G1 can be prioritized by suitable administration of the routing tables. For example, the connection of gateway G0 into the further IP network IP can be set up as a lower-cost route, and the connection of gateway G1 into the further IP network IP can be set up as a higher-cost route. Prioritization is a means of ensuring, in the event of a fault on the crosslink Q1, that the host always uses the network (in this case: N0) connected to the default gateway G0 for communication.
However, such a prioritization is not required in all cases, for example if the crosslink Q1 physically includes multiple links—not shown. In this case the prioritization is not necessary, since at least one further link is available in the event of the failure of one of these links.
Based on the network topology presented, the following message paths, for example, result; only network-internal paths are considered here:
If the mentioned prioritization is provided for the gateways G0, G1, and if the interfaces IF0, IF1 are also prioritized in addition, IF0, for example, having the higher priority, the following prioritization of the paths mentioned results, provided the gateway prioritization is to take precedence over the interface prioritization:
Further message paths are produced in similar fashion if the cited crossover connections from N0 to G1 and N1 to G0 are present and/or if further crosslinks or also crossover connections exist inside the communications network N between subnetworks N0 and N1.
The message paths are now tested, in that the host sends special test IP datagrams via each interface IF0, IF1 to each gateway G0, G1 at very short time intervals, e.g. every 100 ms. The IP address of the respective dedicated interface IF0 or IF1 is entered as both source IP address and as destination IP address. Thus, the test packet is mirrored back to the sending interface IF0, IF1 of the host by the gateway.
The following table shows the IP and MAC addresses to be chosen for testing the message paths Path1 . . . Path4:
Basically, therefore, the layer 2 messages are addressed correctly using the respective MAC (MAC=Media Access Control) addresses, whereas the addressing of the higher layer 3 messages is modified such that the layer 3 messages are routed back to the sending entity. This principle is based on the fact that as a rule layer n messages are not modified during transport through a layer n-1 network and that layer n address information is not interpreted by the layer n-1 network.
For IP test messages, an important advantage is that only the “IP forwarding” function, which is implemented on the very powerful interface cards of the gateways, is required for mirroring or sending back the test messages to the sending entity. Thus, an overload situation in the gateway due to the method according to the invention cannot occur, since the switching processor of the gateways is not involved in any way in the processing of the test messages.
If the test message mirrored at the respective destination is not received again by the host within a specific period of time, e.g. 100 ms, there is probably a fault on the corresponding message path. This is recorded in a storage buffer for example. In a development of the invention, the fault on the message path is only recorded as a permanent fault if the following test message associated with this message path is also not received again at the host. In a further development, the number of consecutive messages that may be lost per message path before this is interpreted as a fault can be adapted to the particular requirements.
Alternatively, it is also possible to identify the transmitted test messages by means of consecutive numbers or sequence numbers. These are entered in the payload of the test messages. The loss of a configurable number of not necessarily sequential test messages can also be used as a criterion for failure detection, i.e. the message paths are monitored by numbering of the test messages. In this case the counter for lost test messages can be designed such that a lost test message increments the counter by 1 and a configurable number of test messages received without loss, e.g. 1000, decrements the counter by 1. Alternatively, the counter can be decremented upon expiration of a time interval during which no test message loss has occurred. If the counter reaches a limit value, the message path is deemed faulty.
If the message paths are checked at sufficiently short time intervals with the aid of the method according to the invention, every 100 ms in the exemplary embodiment described, and if a failed test is repeated precisely once before the corresponding path is deemed faulty, the message path will be recognized as faulty after a very short delay, in this case 200 ms, if the repeated test fails.
With reference to the actual application scenario, it is a straightforward matter for the person skilled in the art to adapt the described parameters of the test method according to the invention to the particular application.
After a fault has been detected and recorded, the user data traffic of the faulty message path is redirect to a fault-free message path. The methods for doing this are well-known. However, advantageous strategies for selecting the alternate message path are presented below with reference to FIGS. 3 to 6, where FIGS. 3 to 6 contain examples of faults on message paths.
If Path1 also becomes faulty as a result of a further failure without the fault on paths 2 and 3 being rectified, a failover is then made directly to the lowest prioritized path 4. As the fault information is always current because of the tests continuing to be run every 100 ms even for faulty paths, this failover can be effected without delay, without a failover to paths 2 or 3 being attempted first.
The failover strategy described with reference to FIGS. 3 to 6 is illustrated in the following table. The meaning of the various symbols is as follows:
Here, a gateway failover means that the host uses a different gateway for sending IP packets in the direction of the IP network, whereas interface failover means that the host uses a different interface for sending and receiving messages. For “internal” communication, i.e. communication between multiple hosts connected to the communications network N—not shown, it is preferred that all hosts always have a connection to the same default gateway G0 or G1. In this way, host-to-host communication is ensured even in the event of partial failures, for example failures of the crosslink path Q1. A failover to the standby gateway G1 is effected only if the default gateway G0 cannot be reached either via IF0 or via IF1, which is also reflected in the prioritization of the paths.
Although the exemplary embodiment of the invention is described with reference to an IP/LAN environment, the invention is not limited to this protocol environment. Connection-oriented protocols can, for example, be used for monitoring the host-gateway connection if these support a connection setup “to itself”, i.e. source address=destination address. If an interruption to the connection is detected by the protocol, a failover to a redundant transmission path can be initiated. Examples of such protocols are the Real Time Protocol RTP or Stream Control Transmission Protocol SCTP.
In certain networks it may be necessary for both the first device Host and also the second and third devices G0, G1 to know the status of all message paths. In order to achieve this, the method according to the invention can be implemented for all devices that need to know the status of the message paths. Alternatively, the status can be transmitted by means of status messages from one device executing the test method to all other devices. The advantage of the present invention is that the test messages initiated by different devices, e.g. multiple hosts, do not mutually influence one another.
An exemplary network element Host, for which the method described in the foregoing is implemented, comprises, in addition to send-receive devices or interfaces IF0, IF1 to the communications network N, for example control logic which converts the described method. Control logic of this type also has a device for providing test messages having destination addresses and source addresses, e.g. source IP address and destination IP address, which correspond to the address of the network element and/or its interfaces.
The control logic further comprises devices for monitoring the individual message paths. In this case the message paths can be predetermined by operator intervention or determined automatically by suitable processes.
The control logic establishes on the basis of the criteria already explained in detail whether a message path is faulty and initiates the selection and failover to an alternative message path according to the failover strategy. For this purpose, the control logic has suitable switchover elements, as well as storage elements in which the prioritization of individual message paths is stored.
Although multiple crosslinks can be provided between the subnetworks N0, N1, it is advantageous to provide only one crosslink Q1 at the switches located nearest to the gateways G0, G1. In this way Layer 2 loops and hence the use of a Spanning Tree Protocol SPT can be avoided.
However, prioritization is not necessary in all cases, for example if the crosslink Q1 physically includes multiple links—not shown. In this case the prioritization is not required, since at least one further connection is available if one of these connections fails.
The links L01, L02 and L11, L12 between the switching elements S00, S01, S02 and S10, S11, S12 shown in
This is shown schematically in
In
The exemplary embodiments of the connection of a host device to gateway device(s) represented schematically in
Taking the schematic view from
The message paths represented schematically in
Here, the redundant ring structure permits the configuration of alternate paths. For example, if the ring segment between M1 and M4 fails, this section of Path1 can be alternately switched as follows:
In similar fashion, internal alternate paths with regard to the WAN can be specified for other failures; methods in this respect are sufficiently known.
Taking the schematic view from
In contrast to the arrangement represented in
The connection of the host device to the (optional) gateway G1 is implemented by means of a local area network (e.g. LAN) N1 and the RPR. Link L12 connects the subnetwork N1 to the RPR.
The message paths represented schematically in
How the communications paths run in the RPR in this case depends on the current state of the ring itself and is not important for the method described here, since the redundant ring structure and the ring protocol ensure the automatic configuration of alternate paths. For example, if the ring segment between E1 and E4 fails, this section is alternately switched by the ring protocol as follows: E1<->E2<->E3<->E4.
In similar fashion, internal alternate paths with regard to the WAN can be specified for other failures; methods in this respect are sufficiently known.
The network arrangement according to the invention can advantageously be combined with the method for testing the message paths described above.
After a fault has been detected and recorded, the user data traffic of the faulty message path is redirected to another, fault-free, message path. The methods for doing this are well-known. For example, the host sends a “gratuitous ARP”, i.e. an ARP request in respect of its own IP address. The host uses the interface from which the request originates as the source MAC address, and its own IP address as the sought IP address. As a result of the ARP broadcast, the ARP caches of all connected hosts and gateways are updated with the MAC/IP address relation. The switchover is effected, for example, to the mentioned alternate message paths, which are selected according to their prioritization.
With SONET and Resilient Packet Ring, the present invention has been described for two typical redundant WAN methods. Other WAN methods can, of course, also be applied to the present invention, particularly in connection with the theory outlined in
The above description is presented to enable a person skilled in the art to make and use the invention, and is provided in the context of a particular application and its requirements. Various modifications to the preferred embodiments will be readily apparent to those skilled in the art, and the generic principles defined herein may be applied to other embodiments and applications without departing from the spirit and scope of the invention. Thus, this invention is not intended to be limited to the embodiments shown, but is to be accorded the widest scope consistent with the principles and features disclosed herein.
Other embodiments and uses of the invention will be apparent to those skilled in the art from consideration of the specification and practice of the invention disclosed herein. All references cited herein, including all written publications, all U.S. and foreign patents and patent applications, and all published statutes and standards, are specifically and entirely incorporated by reference. It is intended that the specification and examples be considered exemplary only with the true scope and spirit of the invention indicated by the following claims.
| Number | Date | Country | Kind |
|---|---|---|---|
| 02019298.5 | Aug 2002 | EP | regional |
This application claims priority to European Application No. 02019298.5, filed Aug. 28, 2002, U.S. Provisional Application No. 60/406,309, filed Aug. 28, 2002, and U.S. Provisional Application No. 60/429,313, filed Nov. 27, 2002, the contents of which, are hereby incorporated by reference.
| Number | Date | Country | |
|---|---|---|---|
| 60406309 | Aug 2002 | US | |
| 60429313 | Nov 2002 | US |
| Number | Date | Country | |
|---|---|---|---|
| Parent | 10648832 | Aug 2003 | US |
| Child | 11847178 | Aug 2007 | US |
