![]() |
|||
![]()
|
![]() |
![]() Click Here! |
![]() |
Recommended Implementations Exhibits 1-6-20 and 1-6-21 illustrate the best practical approaches for meeting the full range of power and grounding requirements of a distributed system. A Departmental LAN Exhibit 1-6-20 shows a LAN or LAN segment that is supplied entirely by a single low-voltage power distribution system within one building. In an ideal installation, small power conditioning systems (that use isolation through dedicated transformers) provide clean power and a solid reference ground for each device on the network. The otherwise pristine installation shown in Exhibit 1-6-21 can be corrupted.
Improper installation of data communications cables (i.e., running data cables in parallel with power circuits or near other sources of electromagnetic interference such as fluorescent light fixtures, or large motors) can let electrical transients into the network through the data cables. If surge suppression devices, which divert surge currents to ground, are placed out on the branch circuits, the integrity of the system reference grounds can be compromised. Illegal neutral-to-ground connections at the panel (X) or somewhere within a branch circuit would divert to ground some portion of the normal return currents on the neutral conductor. Sometimes these illegal connections are made intentionally to solve computer performance problems caused by common mode noise By mistake or otherwise these connections can induce a repeating low-frequency waveform on unbalanced communications cables, potentially corrupting the data stream. It is not uncommon to find illegal neutral-ground connections in newer as well as old buildings. Extended LAN Exhibit 1-6-21 shows a LAN that overlaps separate power distribution systems or extends beyond a single building whether the situation is two sources in one building or two separate buildings, the electrical grounds can be at different electrical potentials. The best solution to the possible ground skew and ground offset problem in these environments is to use a fiber-optic communications backbone between LAN segments. Fiber links do not make an electrical connection between different power distribution and grounding systems. The circuits supporting each local segment should have a common ground source if the building has been wired according to the National Electrical Code. If a fiber connection cannot be used and LAN segments in different buildings are being connection, then supplemental data line transient protection is recommended.
The advantage of the fiber-optic backbone is that each LAN segment can be treated as a departmental environment. Power conditioning systems should be installed on internetworking devices as well as on the individual network components in each LAN segment. For mission-critical networks, backup power is recommended not only for file and communications servers but also for critical path bridges, routers, and concentrators. SUMMARY There are many issues to consider when trying to design a trusted, fault-tolerant network. First, the value of the work that depends on the network environment has to be determined. A model that can be used to place a dollar value on network reliability has been provided. If a high value is placed on systems uptime, then it pays to do things right. Many network owners look to redundancy to provide added system fault tolerance. Redundancy is expensiveand for some systems it does make sense. But for all systems, attention to providing the correct electrical environment will provide a solid foundation on which to build reliable network systems. As networks become increasingly complex and critical to an enterprises business processes, they require managing. Installing a network management system is an expensive and complex undertaking. Some networks warrant this level of investment in reliability. But for all systems, a network that operates reliably is inherently more easily managed than one that does not. Again, it pays to install a solid power foundation to improve network reliability and simplify network management. This means assuring correct building electric wiring, selecting appropriate communications topologies, and careful placement of appropriate power conditioning devices throughout the network. The solution most like that used in data centers calls for the local placement of small, dedicated power conditioning systems that use low-impedance transformers as a key component. This provides a clean reference ground for each network device. It also protects the device and the network from the effects of transient electrical energy. Critical network devices such as servers, hubs, and routers should in most cases also be supported by battery backup when continued uptime or controlled shutdown is required. In those networks that do warrant a comprehensive network management system, a UPS used in that network should be included in the management scheme. A remotely manageable UPS has the capability to send alerts over the network when its batteries need replacing or when other conditions arise that require preventative or immediate maintenance. These advanced UPS management agents can also be used to restart locked devices or to gather information about power failure incidence in that segment of the network. In the end, the decision of how much time and money should be spent on preparing the LAN operating environment should be based on the costs of downtime for a particular network or network segment, the estimated mean time to fault occurrence, and site specific risks factors related to power quality that can adversely impact overall system fault incidence.
|
![]() |
|
Use of this site is subject certain Terms & Conditions. Copyright (c) 1996-1999 EarthWeb, Inc.. All rights reserved. Reproduction in whole or in part in any form or medium without express written permission of EarthWeb is prohibited. Please read our privacy policy for details. |