Distributed vs. A Central UPS  Modern office computer networks consist of a number of interconnected components such as PCs, file servers, work stations, routers, hubs and so on. When a UPS is being considered for a network, it is important to determine exactly which components require protection. Having determined this, it is then necessary to decide whether to use a central UPS or a number of smaller distributed UPSs. 

What Needs To Be Protected? 

The most important components to be protected are the file servers. This is because if they suddenly lose power, data on their hard disks may be corrupted. Since they hold a large number of files which are used by many users, any data loss can have a devastating effect on the organisation. In this situation, it is necessary to restore the hard disks from the last backup which means that all work done since that backup will be lost. 

The next most important components to be protected are the actual network components such as the routers, hubs, and so on. For small networks these will generally be located in close proximity to the file servers and will draw little power. It therefore makes sense to power them from the file server's UPS. 

The final items to consider are the PCs and work stations. If they lose power, any work currently being done on them will be lost. The question is whether the cost of re-creating this lost work justifies the cost of protection. The answer to this question will often vary from one work station to another. Once it has been decided to protect a work station, the next question is will it be powered from the central (file server's) UPS, or will it have its own local UPS. 

Power Distribution 
 

Figure 1 : Distributed UPS system

 

 
Figure 2: Central UPS system

 
The main problem associated with a central UPS is how to distribute the power from the UPS to the work stations. In the case of a small office network where all the components are in the same room, this can be easily achieved using power blocks and extension cords. If however the components are spread across several rooms, this approach cannot be used as it is unsightly, dangerous, and in most countries illegal. In these situations, the power distribution will take the form of special power outlets which will need to be installed by an electrician. This can be a costly exercise in older buildings, although if this wiring can be installed in a new building as it is being built, the cost is not too great. 

Another problem associated with these distributed power outlets is that it is possible for non computer friendly appliances such as refrigerators, coffee machines and vacuum cleaners to be accidentally connected to the UPS. These devices often produce large amounts of electrical noise which can upset some of the network components. This problem can be overcome by using special connectors for the UPS outlets. 

Obviously with distributed UPSs, the problems associated with power distribution do not exist. 

What Happens When The Network Grows? 

As the network grows, it may become necessary to provide protection to more sites. In the case of a central UPS, this will require the installation of additional power outlets, and may also require an increase in the capacity of the UPS. If the UPS is modular and has spare slots, than all that is required is to plug in extra modules. If however, the UPS is non modular or has no spare slots, then the cost of this upgrade can be very expensive. With distributed UPSs, all that is required is to plug in another UPS at the new site. 
 
Reliability 
 
Having decided to protect some or all of the computer network, the question arises as to which components need to be protected by a fault tolerant UPS. If a central UPS is used, and it is modular and fault tolerant, then only 1 additional module is required for every site to be protected by a fault tolerant UPS. 

In the case of distributed UPSs, if every site needed to be fault tolerant, then their UPSs would need to have at least 2 modules. Given that 1 module would normally be sufficient to power an average PC or work station, this effectively doubles the cost of each local UPS. A more practical approach would be to have 1 complete spare UPS and simply swap it for a faulty unit when the need arose. Obviously this approach would result in a small risk to each site if the mains were to fail before the UPS was replaced. 

Conclusions 

There are no hard and fast rules as to which approach is "best". Each installation needs to be looked at in detail, and the appropriate decisions made. The information contained in this paper is only meant to be a guide to this decision making process.
 
Back to Technical Bulletin