Intel Haifa current situation, equipment and needs

Intel Haifa(74) is a very advanced high-tech industry campus, which contains several thousands of network nodes, enhanced with very cunning equipment and emitting a very large amount of quality technology production. The way Intel Haifa production process goes it requires a near zero LAN downtime, therefore many efforts are invested in order to make it a stable environment. New technology solutions and up-to-date equipment are being hunted down and tested on purpose to make the LAN even more solid.

At current stage Intel Haifa LAN roughly consists of following components:

Each workstation and mobile that are connected to the LAN have a name and an IP number in the network, and of course have their unique identifire - MAC address of the Network adapter. Each such station supports network protocols to verify it being connected to the LAN.

Each TO is uniquely labelled, so that we are able to trace its location within the building easily. The TOs are one-to-one connected to Patch Panels. The maps of the locations of TOs in cubicles and their respective tables in Excel were being prepared by Bynat Ltd. lately and would present an example of an external database used.

A name is given to each workstation, many workstations have two or more names, while all workstations are chained to their owners by names through an another external database, an ASCII table maintained by Intel Computer Centre, and often the owner name can be derived from the workstation name.

An IP is given to any connected workstation. Any NT/Mobile/MS Windows/Windows 95 workstation receives an IP through a permanently running DHCP service, which represents about 40% of the workstations. The rest receive the IPs manually, by Network Administration. The IP's as related to MAC addresses of their workstations are all stored in cashes of the campus routers. The permanent nameservice is always able to connect between an IP and a name or names of the node.

Each router in campus belongs to advanced Network equipment. They're a production of Cisco and all support SNMP protocol. They maintain large cashes about the nodes in the network, of which important to us is the relation between a MAP address and the IP. The cashes are often refreshed, usually automatically, but sometimes manually. The relation interesting us is obtained by an SNMP request to the router.

Most of the HUBs also speak SNMP, they are made by Bay Networks, although there are few old LanNet HUBs, that do not support SNMP and therefore are being replaced, step by step. The new HUBs support SNMP and answer many different informational requests through snmp queries. The one that interested us yielded a relation between a MAC, experiencing a connection to a specific slot in the HUB and this slot.

The PatchPanels inside the campus are provided by RIT and have many features an usual patch panels don't have. Nevertheless, until recently those features weren't used. Now, it was decided to supply panel agents, which also support SNMP (the panel itself is just a piece of plastic with wires and leds) and may report which wires or switches are being used on the panel. The agents are supervised from a MS Windows oriented software, also supplied by RIT and called "PatchView"(TM) and it yields datain MS Access db format about wires and switches plugged in. During the time project was being worked on, another project was started with RIT which was in order to synchronize information received by PatchView with other sources of information on LAN and to transmit or dump the data from PatchView into the central pool of data.

The Cubicles are numerous, and usually populated by more than one person. Unfortunately no database connecting between the cubicles, which are labelled by their locations in buildings and their inhabitants existed, so it was a whitespace for the project. Luckily, cubicle represents an end of a data chain in that relation, so it didn't affect the demostrativity of the prototype.


romm@empire.tau.ac.il
Last modified: Thu Jun 5 03:33:14 1997