Design and Hardware Implementation of a Connection Identification System for ATM Networks

ABSTRACT 

This paper presents a connection identification system and its hardware implementation architecture for use in ATM (Asynchronous Transfer Mode) Networks. The system is responsible for connection recognition and identification. This is applied to suitable connection identity fields. These fields are VPI/VCI (virtual path/virtual channels identifiers). These offered functions are useful for operation of terminals, switches, or interworking units. The system has been implemented experimentally as part of ATM network integrated circuits. 

INTRODUCTION 

Operation algorithm, design, and implementation architecture of an ATM (Asynchronous Transfer Mode) Connection Identification System (CIS) are presented in this paper.  This system is mainly used to support recognition and identification functions of virtual paths/virtual channels  (VP/VC) in ATM Networks. The ATM method that has been standardized for use in IBCNs (Integrated Broadband Communication Networks) use these channels whose value is the connection identity. The flexibility in allocating bandwidth to connections (VC) or groups of connections (VP) and the possibility to increase the utilization of network resources through statistical multiplexing of bursty traffic streams, are among the distinct advantages of the new information transfer method. The system firstly extracts the connection identity fields, and then, it searches to find if these fields belong to a used connection, or not. In this way if the extracted VPI/VCI (Virtual Path/Virtual Channel Identifiers) fields match with the stored ones then, the connection is used. This system which can be used in terminals, in switches or, in interworking units implements basic functions of the lower Layers of the ATM PRM (Protocol Reference Model) [1], [2], [3], [4]. 

CONNECTION IDENTIFICATION & OPERATIONAL ENVIRONMENT 

CIS is responsible for checking the VPI/VCI fields of cell headers. The role of it is to inform other systems about the identity of cells. The received cells are in parallel form (bytes) and transferred with a byte clock rate of 19.44 MHz (~50 ns). Firstly, a unit extracts the VPI/VCI fields, provided that header error control has done and the checked cell isn’t erroneous or has been corrected. The fields are stored temporarily into a register until the begin of a new time slot period. Then, the recognition/identification procedure begins [5], [6]. According to this the extracted fields compared to other values of a list. The list includes other used/unused VPI/VCI fields. If a matched field found which corresponds to a used connection, then the current cell is accepted and the operation control transferred to an other system. Otherwise, the current cell doesn’t belong to a used connection.

 Figure 1(a). CIS in an ATM Switch

Figure 1(b). CIS in a Terminal Environment

 CIS can be used in a switch (Fig. 1(a)). It can support basic functions of the switch as header translation, cell routing and others are. In order to do these, connection identification must be done firstly. A microprocessor is responsible for the control of the system. CIS can be used in a terminal as well (Fig. 1(b)). It offers VPI/VCI fields recognition/identification when the cells are in microprocessor adapter of the terminal. In this way deactivated cells or idle cells can be discarded without being transferred to the terminal and so let it save cell transfer cycles. CIS can be used for recognition/identification of other types of cells as resource management, maintenance cells etc.

VPI/VCI FIELDS RECOGNITION/IDENTIFICATION 

Since header error control operation has finished then, VPI/VCI fields are extracted and stored with the rest of the header in a FIFO (First In First Out) system. Extraction is done via suitable registers (24 bits) and appropriate logic gates, which controlled by a Counter outputs. Since, VPI/VCI fields have been extracted then, connection recognition/identification is done by a suitable system. It can manipulate N different connections simultaneously. The accepted VPI/VCI values (used connections or open connections) have already been stored into a table (VPI/VCI Table), [7], [8]. These values consist of 24 bits according to CCITT specifications [2].  Other 24-bit values (masking bits) correspond to VPI/VCI values and enable or disable the completely or partly use of VPI/VCI fields. Masking bits are stored into the VPI/VCI Table as well (Fig. 2). The use of these depends on the position of the Connection Identification system in the network. The N VPI/VCI fields can be classified in s groups. Each group can be assigned a 24-bit masking field and s may vary between 1 and N. The system examines the values of VPI/VCI fields. It accepts cells that have VPI/VCI with same value as stored VPI/VCI. Differently, the cell is rejected because doesn’t belong to an active (open) connection. Every time slot when there is a non-erroneous (or corrected) cell (Header_OK) with a new VPI/VCI identifier (VPI/VCI_avail=1) (and the “bypass connection identity” procedure is disabled – “NOTBypass_CI” logic is true) then, CIS examines the relation

(Header_OK) AND (VPI/VCI_avail) AND (NOTBypass_CI)=1           (1)

 If it is true then, table searching begins. A VPI/VCI Table Counter is used for producing the table addresses. The procedure has as follows (Fig. 3): When there is a VPI/VCI field in the table equal to the current cell VPI/VCI field (VPI/VCI_stored=VPI/VCI_extracted), then the signal “VPI/VCI OK” is activated (Comparison System). Then, the value of the masking bits is checked. If not all of them equal to “1” then searching stops and the current value of the table address (the contents of the VPI/VCI Table Counter) is provided  (Connection_ ID) and is available to any system that needs the results of the connection recognition (e.g. a microprocessor). In this way the current cell is identified and translated. The “Connection_Valid” signal is activated. In systems as terminal receivers where a few connections can be open simultaneously a 4-bit table address is required. If the current VPI/VCI value doesn’t match to a table value of a current position then the system examines if the extracted VPI (only) field matches to a stored VPI field (VPI_stored=VPI_extracted).  If yes, then the signal VPI_OK is activated. If not all of VPI Mask Bits equal to “1”, then the Connection_Valid signal is activated and the table searching is frozen. In all cases if all masking bits equal to “1” then searching is going on. Masking technique used as follows:

Every bit  (l=0...23) (VPI/VCI_Extracted) is compared with every bit  (VPI/VCI_Stored) in parallel. They are used 24 masking bits  (l=0...23) which value determines if comparison cares. If it is true, . If it isn’t true, . In this last case the extracted bit must be considered equal to the stored bit. If it wouldn’t be considered equal and the rest-extracted bits were equal to the rest stored bits, then the comparison resulted to an in-equation (although the masked bits were “don’t care” bits). The system uses the set of the logic equations:

x_l XOR x¢_l = c_l ("l=0…23)                 (2)

 c_l AND (NOTx¢¢_l) = c¢_l ("l=0…23)      (3)

 In the case that the compared VPI/VCI field is a “don’t care” pattern, then all masking bits equal to 1 and

x¢¢₀ AND x¢¢₁ AND … AND x¢¢₂₃ = 1    (4)

Figure 2. VPI/VCI Recognition/Identification Circuit

 Then, the  fed to a tree of OR gates. The output gives the result of comparison. If the searching reaches the end of the table (k=2ⁿ-1) without a valid connection found then, the searching stops and the flag   “Connection_Invalid” takes the value “1”. For handling special cases, some positions have been assigned special values. So, the first position (k=0) of the VPI/VCI table includes the Idle Cell VPI/VCI field. When the extracted cell is idle the matching with k=0 activates the flag Idle_Cell. Moreover, other types of cells can be used as maintenance, metasignalling, unassigned cells etc. It must be noticed that VPI/VCI table includes only used VPI/VCI fields. The remaining positions (unused) of the table include the “Idle Cell” VPI/VCI field. In addition, previously used connections continue to exist but they all have masking bits equal to “1”.

If CIS must not be used, then a “bypass” procedure is used (Bypass_CI=1), then the Connection_Valid flag takes the value of an external identification tag (Ext_Tag).

For updating the stored values 7 registers are used. The first six (VPI/VCI RegUpdate(0), VPI/VCI RegUpdate(1), VPI/VCI RegUpdate(5)) registers include the new value of VPI/VCI that will be written to the table (masking bits are included). VPI/VCI RegUpdate(6) includes the address value that shows the internal table position that will be updated.  A flag shows if the registers are available for use (VPI/VCIRegUpdate_avail). When writing ends (7 steps), then the CIS waits until the end of a current connection identification searching that may occurs. Finally, the registers contents transferred to the table. 

Figure 3. VPI/VCI Recognition/Identification Algorithm

The next figure (Figure 4) shows a snapshot of the implemented CIS functions. We can see the exact timing of the system incoming data (RXHEADATA), incoming cells VPI/VCI bytes (RXVPCI), mask bits (MASK), current cell validation signal (RXVPCIVALID), selected VPI/VCI address signal SELVPCIADDR), incoming cell clock signal (RXCS), current cell acceptance signal (ACCEPT) and other signals of the VPCI/VCI Recognition/Identification system. The circuit described in the previous, has been designed and implemented on an ASIC chip using ECPD10 CMOS technology of the European Silicon Structure (ES2). The VPI/VCI Table implemented for 16 connections and occupied silicon area of 0.63 mm². 

 Figure 4. Implemented functions of the VPI/VCI Recognition/Identification System

 USE OF CONTENT ADDRESSABLE MEMORIES-IMPLEMENTATION GUIDES

Content-Addressable Memories (CAM) offers the possibility of applying associative processing [9]. They are useful to designs that require list searches and data translation as embedded functions within systems. The combination of a CAM and a state machine creates an economical controller [9], [10]. CAMs simplify the controller by switching a lower-cost processor or a state machine. Associative processing manipulates data based on matching, or associating an input value with other values stored in an array. Associative processing hardware incorporates a limited amount of computational capability at each memory location and allows examining an entire memory array at once. In the previous system  (Fig. 2) the integration of a memory (VPI/VCI Table), a counter (VPI/VCI Counter), a comparator (Comparison System), some registers (VPI/VCI RegUpdate (j)), and glue logic constitute a CAM system.  The system has been implemented in an ATM Receiver system circuit as a part of an ASIC chip [8], [10]. The VPI/VCI Extraction part FIFO implemented with five registers in series for saving area reasons. The VPI/VCI Recognition/Identification part uses a table for storing VPI/VCI values. Using CMOS standard-cell technology (ES2 - ECPD10 Process) ASICs were implemented by using megacell procedure. For 16-connections VPI/VCI fields (48 bits) a buffer of 1.68 mm² silicon is used. The design is synchronous and uses a 50 ns period clock (Byte Clock). The implemented table operates using a 100 ns access time.

CONCLUDING REMARKS 

In this paper design and circuit architecture of an ATM Connection Identification System are presented.  This circuit, is mainly used to support connection identification functions for operation of terminals, switches or, interworking units. This is applied to suitable connection identity fields. It has been implemented experimentally as part of ATM network modules. The circuit can be very useful to everyone who develops ATM components and Networks

REFERENCES 

1. Kawarasaki, M. and Jabbari, B., “B-ISDN Architecture and Protocol,” IEEE JSAC, No 9, pp. 1405-1415, Dec. 1991.Vol. 9,
       

2. CCITT, “B-ISDN User Network Interface,” Draft Rec. I. 432.
      

3. Minzer, S.E., "Broadband and Asynchronous Transfer Mode (ATM)," IEEE Communications Magazine, p.p. 17-24, Sept. 1989
       

4. Black Uyless: “ATM Foundation for Broadband Networks,” Volume 1, Prentice Hall, Second Edition, 1999
       

5. Sato, Y. and Sato, K., “Virtual Path and Link Capacity Design for ATM Networks”, IEEE JSAC, Jan. 1991, Vol. 9, No 1 pp. 104-113.
       

6. Koukos, A.K., Papadopoulos, A.I., Doumenis, G.A., Koukoutos, H.E., “The HEC BYTE Generator-VCI Ectractor chip,” RACE Project 1022, NTUASWP4_1091PR_CC, Ver. 1.0, 6/14/92.
      

7. Koukos, A.K., Kavidopoulos, K.A., Papaspyrides, A.C., Kamaras, J.G., Mitrou, N.M., “The ATM Receiver Integrated Circuit,” Hellenic VLS.I Design and Prototyping Environment, HV/NTLC/0034, Ver. 1.2, 20/1/95.
       

8. Koukos, A., Perissakis, S., “Communication Systems Development with the use of ASICs,”  in Proc. Technology and Automation Conference, Piraeus, Greece, May, 9-10,1998
       

9. Weldson, T., “Content-addressable memories add processing power to embedded systems,” Electronic Design News, p.p 137-148, May 9, 1996.
        

10. Koukos, A., “(F)PGAs: A Technological Invitation-An Application: A PC Adapter for B-ISDN (ATM) Networks,”Panhellenic Association of Mechanical and Electrical Engineers Magazine, Nr. 273, p.p. 65-70, April 1995.

Return to This Issues Home Page