9.1 LOCAL AREA NETWORK9.1.1 Introduction to Local Area NetworkingToday's businesses require a computer system capable of multiple input and processing. These capabilities have been available in minicomputer systems for many years. In recent years the microcomputer has been replacing the minicomputer in office environments. This has forced micro systems to have the same multi input and processing features as minicomputer systems. Microcomputer systems are now being used in multi-terminal configurations to form local area networks (LAN). LAN's are distributed processing systems that incorporate multiple user stations, mass data storage, and background printing facilities. there are basically three types of networks; centralised, decentralised and distributed. A centralised network has a "master" system controller controlling network operation. A decentralised network is basically a group of interconnected centralised networks. A distributed network has no main controller and each node on the system shares in the network control process. The microcomputer operating system along with the network hardware allows the user multiple station input (multi-user) and concurrent processing (multi-tasking). The Sirius LAN will be initially offered in a multi-user, single tasking configuration. The network software for this configuration will be MS-DOS version 2.0 (servers) and MS-DOS version 1.25 for network stations. The network will (at a later date) be offered in a multi-user, multi-tasking configuration. The multi-tasking operating system will be based on the BELL LABS UNIX III operating system. Multi-tasking will allow concurrent processing of data, thus increasing system throughput. The International Standards Organisation (ISO) and the American National Standards Institute have developed a seven layer hierarchical network model. The Sirius LAN board will implement the four lower levels. 9.1.2 ISO Seven Layer Network Model
The Sirius network software and applications software will implement the upper three layers. Network communication protocol will be the packet. A packet may contain data, node addresses, error information and control information. Each station on the network will need some method of accessing the network communications channel. The technique used will be carrier sense multiple access (CSMA). CSMA is a technique where a node requiring access to the channel will sense the channel for the presence of a carrier (activity), if no carrier is sensed for a predetermined amount of time, the node accesses the channel. If activity is sensed, the node will calculate a "waiting" period before trying to access the channel again. The "wait for retry" period is calculated only when no carrier is sensed to prevent nodes from queueing up to the channel. The Sirius LAN system uses Positive message acknowledgement to ensure proper message reception. Positive message acknowledgement is a system in which all packet(s) transmitted MUST be acknowledged by the receiving station. If the packet is not acknowledged the host retransmits the packet until it is acknowledged or the packet retry limit is reached. CSMA along with positive message acknowledgement should effectively eliminate collisions. Mass data storage is an important feature of a LAN. The Sirius LAN mass storage units will be hard disk units with expanded memory (256K min.). These units will be called network servers. The network servers will also handle background printing of user files (spooling). 9.2 Sirius 1 Local Area Network OverviewThe Sirius LAN is based on the Corvus OMNI-NET network system. The network supports 64 users (10 servers, 54 stations). The communications channel is a shielded, twisted pair cable with a maximum end to end length of 1.2km. The network data transfer rate approaches 1 mega-bit per second. Each node on the network will contain a transponder board (network card). The network card is directly connected to the network communications channel and the host (node) computer's data bus. The network card will handle all network functions thus freeing the host processor from the duty of controlling the network. The network board communicates with the host system data bus. The network board contains a DMA (direct memory access) controller that allows the network card direct access to memory without host CPU intervention. The network card will perform functions such as packet transmission and reception, packet formatting, error detection, and DMA transfers.
The network board default setting is at E810h (hexadecimal segment address).
Example: SW. 1-6 "ON" selects address segment E810h Switches 7 and 8 are for selecting interrupt priority levels.
The eight position DIP switch at location 3M is used to select the user node address (0-9 servers, 10-63 stations). The switches are used to set a binary value from 0 to 63. A switch in the "ON" position represents a binary 0.
Example: switches 1,3,5 "ON" all others "OFF" selects node address 38. The network card can be functionally broken down into four parts:
The DMA controller chip is a custom gate array chip designed to control the interface between the Sirius computer bus and the Network card. Each DMA cycle is explicitly invoked by the 6801 microprocessor allowing the network software complete control of the DMA transfer between the Sirius computer and the network card.>/P> The MC6854 advanced data link controller (ADLC) controls many of the functions specified in the ISO data link and network layers. The ADLC performs such functions as data serialisation, error code detection (CRC) and generation, packet framing, bit protocol implementation (NRZI non return to zero inverted) and zero insertion. The 6801 microprocessor oversees the operation of the DMA controller and the ADLC. The 6801 controls the transfer of data and control information between the Sirius computer and the Network card. The SN75174/75175 chips form the RS-422 transceiver. The 75174 is a differential line driver, and the 75175 is a differential line receiver. Because the driver and receiver use differential circuits, they offer high noise immunity. The twisted wire pair was chosen as the communications medium because of its ease of installation and RFI (radio frequency interference) immunity level. 9.3 Network Software OverviewThe Server Network product supports a local network of Sirius workstations with one or more network servers providing mass storage and printer access for network users. The network servers are dedicated to providing services for network users and cannot be used as work stations. The network can include discless network stations that boot from the network and have all their mass-storage on the network server(s). Each network server supports a ten megabyte hard disc and double-sided floppy disc drive, but does not need a screen or keyboard. The network servers run network software under the MS-DOS 2.0 operating system while the network stations run a network interface in conjunction with MS-DOS 1.25. The appearance of the file system on a network server is transparent to programs running on the work station, with the network server's hierarchical file system being used to provide each user with private directories. Except for the private directories, all directories on a network server are treated as common storage and can be shared by one or more network users.
An INSTALL program adds users to the network by taking an eleven-character user name, optional password, and assigning drive designators to link the user to disc volumes on the network servers. The INSTALL program takes the available drive designators (those not used for local work station storage) and sequentially assigns them to network server volumes with private and common directories being assigned in a ratio of one to two. The assignments are displayed, and can be changed if desired. Choosing the default assignments ensures that all users have consistent links to all network server volumes. Network installation is simplified by the use of standard AUTOSET files to configure network server hard disc volumes. Common directory assignments can be made to a subset of network users allowing groups that work on common data to share files that, by installation conventions, are unavailable to other users of the network. Login to the network can be performed automatically if the network station is not shared among several people, or a user can be required to login by giving his or her name and password (if required). Utilities can be run to check the status of the network, list the network users, print files on the network server, or reserve files for exclusive access. The network server supports three protection schemes for basic file sharing in common directories.
To facilitate sharing of up to three printers connected to a network server the network station interface can redirect list output at the network station to a specified file on a network server for later printing. Network server volume organisation is determined by choosing an AUTOSET file or using HDSETUP with the server machine configured as a local work station (keyboard and screen connected). AUTOSET configuration can be performed by creating a special configuration disc, keyed to the server number, that automatically formats the server when booted. This alleviates the need to have a screen and keyboard connected to the server. Each server's hard disc can be configured as a number of logical disc volumes, but must have at least two. These are assigned to A: and C: (B: is used for the floppy). The hard disc is optimised for the smallest possible allocation unit size without regard for memory usage (this is why 256K RAM is required in network servers). Small allocation unit sizes have the effect of increasing the effective size of the disc and allow more files to be stored for the network users. A network server appears as a set of remote volumes to programs running on the network stations, with the server's hierarchical directories providing private directories for each user. A users private directory is a sub-directory on a network server volume named with the user's eleven character login name. Each volume's root directory contains the common files for that volume. No change directory command is provided, so that network users do not have to understand hierarchical directories, and, they cannot access any sub-directories that have not been assigned to them by INSTALL. This ensures that other user's sub-directories are not accessible, and remain private to their owners. The network station accesses remote volumes on the network by using standard MS-DOS drive designators (A: to O:). If a network station has local disc drives, the remote volumes should use different drive designators than the local drives, although the INSTALL program does not prohibit this. The file system is composed of all network servers on the network. Each network server has its hard disc volumes partitioned into multiple directories. The private directories of each user are only accessible by that user. MS-DOS 2.0 will enforce "read/only" or "read/write" protection on all files on both regular and private directories. These file attributes are setable by anyone able to access that file (only the user for private directories, and anyone with a drive assignment for common directories). A network server may have attached printers, or other output devices which may be in demand by multiple network users. The devices may be connected to the server through the parallel interface or through Serial A or Serial B RS232 ports. In order to share these devices output is spooled to the network server's disc. The names of the spool files and the user names are placed into a queue maintained by the network server for each printer attached to the server. |