Suche Bilder Maps Play YouTube News Gmail Drive Mehr »
Erweiterte Patentsuche | Abbildungen der Seite | Webprotokoll | Anmelden

Patente

  
[merged small][table][merged small][merged small][merged small][merged small][merged small][merged small][merged small][merged small][merged small][merged small][merged small][merged small][merged small][merged small][merged small][merged small][merged small][table]
[merged small][merged small][merged small][merged small][merged small][merged small][merged small][merged small][table][merged small][merged small][merged small][merged small][merged small][merged small][merged small][merged small][merged small][graphic]

limits the number of Mini-Networks served to 16. The resultant cell header format is shown in FIG. 1

SUMMARY OF THE INVENTION

1 2

TELECOMMUNICATIONS NETWORK (PVCs) for inter-site links). This means that the Source

Mini-Network address is carried in the source Mini-Network

BACKGROUND OF THE INVENTION where it is redundant, being implicit, and similarly the

„.,„., „ ... ,T „„„„ Destination Mini-Network address is carried in the destina

In United Kingdom Patent Application No. GB 2255257 tion j^N^t nG. 2 shows how, by eliminating this

is described an earlier concept for an Asynchronous Transfer redundancy, in accordance with the present invention, the

Mode (ATM) self-routing network which may be used for address capabmty may be expanded.

Private Networks, Virtual Private Networks and distributed ^ ±e flgures shaded ^ show where ch haye been

Local Area Switches in the Public Network. The service is made from the earlier format

connectionless using a, special format of the address octets TwQ gU h maaafanasbi ^ used for the Source and

i n the cell header; cells are directed to their respective Destination Mini-Networks respectively and, since a Mini

destinations by an Interpreter held in each elemental switch , T . , . ... , . . ,. . „ ...

or at each switch in ut ort de endin u on im lementa Network is relatively unstructured internally, it is necessary

or a eac swic inpu po , epen ng upon imp emen a- to infonn the mteipreters which format is being used and this

tion. No overhead is required in the ceU Mor^tion Field [& done ^ me ^ ^ ^ ^ most

which therefore corresponds to the payload m the ATM layer si ^can/octet 06f the ceU heJa as a Source/Destination

or a composite network; hence the use of standard adaption- ^ fomat ^ ^.^^ ranction

layer formats is not compromised. This application is bom ... and ... Mini-Network identities must

included herein by reference. be ^ sjnce {he ^ Mormation required t0 be held

The network is divided into Mini-Networks typically with exceeds the capacity of the standard cell header format, the

up to 256 user Network Ports. The size of the address field 2Q function must be internal to a switch element.

The rules for Interpretation are as follows:
If the S/D bit is zero AND the D-Minet 1,2 field indicates

It will be seen that such a network is limited to a own Mini-Network then route on D-Port 1,2.

maximum of 4k (4,096) User Network Ports, although thee ff the S/D bit h zero ^ thg 12 fidd mdicates

use of multiplexers can substantially increase this number. 25 thatthedestination is NOT own Mini-Network then route on

D-Minet 1.2.

If the S/D bit is one then route on D-Port 1,2.

According to the present invention there is provided a xhe theoretical maximum addressing capability using the

telecommunications switching network comprising a plural- ceu header formats of the present invention is 128 Mini

ity of sub-networks, each sub-network having a plurality of 30 Networks with 256 User Network Ports (UNPs) each. The

User Network Ports wherein the Destination and Source resultant maximum address range is 32k UNPs which is a

addresses in a cell-header carried by a cell switched by the verv i^ge number for a private broadband network,

network are each divided into a Port Address part and a ^ order t0 improve fault tolerance and to simplify con

Sub-network Address part and wherein the Source Sub- gestion avoidance haif of ^ capacity is sacrificed by

network part is implicit whilst the cell remains in the Source 35 pairing Mini-Networks so that two adjacent addresses serve

Sub-network and similarly the Destination Sub-network part the same of ... Ls bit of me Miner field is

is implicit from when the cell arrives in the Destination used t0 sepaiate the two, this bit is referred to as the 'A' (for

Sub-network, wherein the Source and the Destination Sub- Alternative route) bit

networks are linked via a cross-connect switching device The A.Wt h &m in (he int tation of me

wherein the address formats are changed. 40 ^ ^ ^ fc ^ ^ ... of me port

BRIEF DESCRIPTION OF THE DRAWINGS ^e^' ^ means alternate paths are provided both for

intra-Mini-Network and to cross-connect switches and pub

The present invention will now be described by way of lie network ports,

example, with reference to the accompanying drawings, in 45 The rules for interpretation are therefore modified as

which: follows:

FIG. 1 shows an example of a prior art cell header format; If the S/D bit is Zero AND the D-Minet 1,2 field indicates

FIG. 2 shows a variation of the format shown in FIG. 1 own Mini-Network then route on D-Port t,2 and the

with the cross-connect function of the present invention; A-bit.

FIG. 3 shows a further variation of the format shown in 50 If the S/D bit is zero AND the D-Minet 12 field indicates

FIG. 1 having a modified cross-connect function; that the destination is NOT own Mini-Network then

FIG. 4 shows diagrammatically the paths between users route on G-Mi^t1-2 ^ ^ A"bit

on two mini-networks ^ ^e ^ *s one tnen route 011 D-Port 1,2 and the A-bit

_ . .. „ ^_ , It will be seen that the A-bit is included in the interpre

FIGS 5 and 6 show diagrammatically the transmit and 5J ^ ... ^ ^

receive functions for a public network link; T~ = , ,. u.t . „„m11l7 „„+=0=,'0 „„„„„„

r The value or the A-bit is normally not the user s concern,

FIG. 7 shows diagrammatically the signalling system being set tne Exchange Terminator (ET) from a table for between peripheral nodes for the present invention; the 16 possible Virtual Connections (VCs) and maintained

FIG. 8 shows diagrammatically the use of a Public by the Resource Manager. However, for certain functions Network Port address with addresses allocated in site net- 60 (for example Multicast codes), the two addresses may be works. used individually. The S/D bit, normally used to identify the

header format, is redundant in the upstream UNP link; and is used instead to inform the ET that the A-bit value given is real. Note that this feature is implemented by user choice, The earlier network, as described above, uses a common 65 in a network where the address range is not a pressing address format throughout (apart from changes at the Public problem users may prefer to leave the choice up to the Network Port used in mapping to Public Virtual Circuits manager.

DETAILED DESCRIPTION OF THE
PREFERRED EMBODIMENTS

5,627,824

If the connection is to another Mini-Network the A-bit (in fact the Destination A-bit (DA-bit)) would be used to select one of two separate cross-connect switches. In the chosen cross-connect switch the DA-bit is ignored. The crossconnect function will replace the D-Miner field with the 5 S-Minet field and set the S/D-bit to '1'. The Source A-bit (SA-bit) will be set from the interpreter to a value chosen by the Resource Manager resulting in two routes to the destination UNP from each of the two cross-connect switches. It will be seen that, by this means, the choice of preferred path is localized in each Mini-Network. FIG. 3 shows the modified connect function.

The DA-bit may also be used to choose one of two Public Network Ports; however this bit is ignored in the transmit function it being assumed that the public network provides its own means of avoiding congestion and faulty equipment. 15 At the Public Network Port receive function the SA-bit is added in thee same way as for a cross-connect.

FIG. 4 illustrates the alternative paths available from a user on UNP "X" on Mini-Network "P" to a user on UNP "Y" on Mini-Network "Q" and also to a user on his own 20 Mini-Network and to the Public Network ports.

The theoretical maximum addressing capability using the cell header format of the present invention is 128 MiniNetworks with 256 User Network Ports each; however, in a Virtual Private Network (VPN) using PVCs on ATM cross- 25 connects in the Public Network to interconnect sites, the maximum is unlikely to be achievable. The reason for this is that the Public Network is performing the cross-connect function described earlier but, being a large and distributed network, it cannot offer the increased address space avail- 30 able internal to a cross-connect switch. This problem is reduced, but not eliminated, by the alternative-routing feature (A-bit) which halves the effective number of MiniNetworks served over public network links and is further reduced by using the A-bit real address feature, where two 35 Public Network Ports are fitted which may, in this case, be accessed under user control and have different translators.

The User-Network Interface (UNI) format for links with the public network is too restrictive with the VPI field limited to 8 bits; however, it is possible to negotiate an 40 Network-Network Interface (NNI) access format with the public network operator. The Generic Flow Control (GFC) field is lost but public network generic flow control adds little to the inbuilt congestion control of the VPN management tools, which monitor queue lengths and provide a 45 service-friendly mechanism for restricting bandwidth on specific links. All that is necessary in addition is a communication channel between the private network manager and the public network manager, a channel provided in the site networks by the network signalling. 50

Even using the NNI format on the UNI link the address capacity is still not enough for a single access link to address the complete network. In the scenario described below User Network Ports (UNPs) are of two types:

Those with access to/from the Site network only (type 55 "S).

Those with Universal access (type "U").

Unlike the private network described earlier where the Virtual Path Identifier (VPI) and Virtual Channel identifier (VCT) cell header fields carry absolute addresses for UNPs 60 in the private network, these fields in the public network provide a unique path/channel identity only on a single link. Excluding the LS 4 bits of the VPI field there are 24 bits available and the value carried in these bits must uniquely identify, at the source end: 65

The identity of the (type U) destination UNP in the total. external network.

The identity of the (type U) source UNP on the source site. Rule 1

The product of the total number of User Network Ports of type U served by the link in the external network, and the number of User Network Ports of type U served by the link in the source network, must not exceed 224.

ATM switches used in a public connection-mode network have to provide translation of the address field values. For ATM transport network cross-connect switches it is likely that translation is limited to the 12-bit VPI field thus: Rule 2

The number of destination sites served by the link must not exceed 212.

The VCI field content is carried unchanged by the public network.

A complete network address is 15-bits long. Full source plus destination addresses would require 30 bits, but the ATM cell header (less the 4-bits used for the MUX field) provides only 24 bits; this available address space must be used as efficiently as possible. The alternative routing feature reduces the Network Port address to 14-bits so that the overlap between Source plus Destination addresses and the available address space on a public network PVC from 6-bits to 4-bits.

The space occupied by the source address depends upon the number of type U ports to be served; similarly the space required for the destination address depends upon the number of type U ports in the rest of the network. The boundary between these two number domains is variable, depending upon the configuration of the network; the maximum size of the source and the destination domains is 14 bits (assuming alternative routing is implemented) which implies a boundary region of 4 bits. Separate translators for source and destination addresses are required; each translator gives a 14-bit output word and the two words are merged with a 4-bit overlap. In the overlap region the two words are added; the effect of this is that the boundaries between the two domains need not be on a power-of-two boundary although in most cases, for ease of management, it will. The resultant interface function is shown in FIG. 5.

The corresponding function at the receive end of the link is shown in FIG. 6.

Note that the transmit function is closely associated with a multiplexer for inputs from several site Mini-Networks; similarly the receive function is associated with a demultiplexer for outputs to several site Mini-Networks.

The value of broadband ports which only have access to the on-site network (non-U) must be questioned, but this style of access is all that is needed for a user who works via an X.435 based broadband store-and-forward network with at least one on-site Node; the only theoretical limitation to such a mode of working is when live communication is necessary (e.g. a video link); however, store-and-forward has the very significant advantages of not interrupting current work and of avoiding problems with time differentials for widely separated sites (e.g. for US and European collaboration).

In order to enable management of the network a comprehensive signalling system must be provided. The signalling system is as described in GB 2255257 and is described briefly below.

There are five kinds of equipment node in a network as described, that is Switches, Public Network Ports Exchange Terminators (ET at the upstream end of an access link), Multiplexers and Network Terminators (NT at the downstream end of an access link). For Switches and Public 5

Network Ports a unique signalling address is allocated to each node but this is not possible for peripheral nodes because the result would be a halving of the number of Network Ports which may be served. The mechanism used for peripheral nodes, that is ETs, Multiplexers and NTs, is described below, with reference to FIG. 7.

Switches route on the destination address only. Signalling to peripheral nodes uses the normal destination address of the Network Port but one of three reserved addresses is carried in the VCI field, normally used to carry the source address. The three addresses relate to the ET, the Multiplexer and the NT respectively. The displaced source address is carried in the Information Field of the cell but since signalling is only used intra-network, external equipment never sees this infringement of the "free-of-overhead" rule. All peripheral nodes are required to test all cells, on upstream and downstream intra-network ports, for the presence of the relevant signalling address.

As well as access to the network manager, the signalling provides for NT-to-NT (i.e. User-to-User) signalling and ET-to-ET signalling for test purposes and fault location.

Public Network Ports are allocated an address for signalling purposes. FIG. 8 shows how this address may be used in a reciprocal arrangement with addresses allocated in both Source and Destination site networks.

A principal use of this mechanism is the establishment and change of translation tables in the two ends or the link, from an instance of the resource manager running in either network.

There is one area where the lack of full connectivity between sites using the present invention can be an embarrassment and that is for the ubiquitous voice service; however, it its expected that the excellent service provided by modern PABXs will continue to be used. One of the several options possible to avoid the inefficiency and cost of

6

separate inter-site Synchronous Transfer Mode (STM) links for the PABX network is to adapt primary-rate multiplexes to ATM channels which may be carried by the network. STM channels carried may use constant-bit-rate (CBR) or

5 variable-bit-rate (VBR) coding ith silence suppression and compression (e.g. using G.764/G.727 modified for ATM). Access into the STM public network will, however, not be possible by this means unless/until the public network operator provides a compatible service.

10 I claim:

1. A telecommunications switching network comprising a plurality of sub-networks, each sub-network having a plurality of User Network Ports wherein the Destination and Source addresses in a cell-header carried by a cell switched

15 by the network are each divided into a Port Address part and a Sub-network Address part and characterized in that the Source Sub-network part is implicit whilst the cell remains in the Source Sub-network and similarly the Destination Sub-network part is implicit from when the cell arrives in the

20 Destination Sub-network, wherein the Source and the Destination Sub-networks are linked via a cross-connect switching device wherein the address formats are changed.

2. A switching network as claimed in claim 1, wherein the cell-header includes a format indicator to indicate the use of

25 a Source or Destination format.

3. A switching network as claimed in claim 1, wherein the cell-header includes an alternative route bit to provide alternative paths for a cell to a group of User Network Ports (UNPs).

30 4. A switching network as claimed in claim 1, wherein the network includes a public network, the User Network Interface format including the space occupied by the Genetic Flow Control Field.

« ZurückWeiter »