Market Overview - UPSs: Keep On Running
When the power goes down, you?d better be ready or you?ll be dealing with the consequences of vast data loss. Roger Gann investigates uninterruptible power supplies
UPS has nothing to do with odd-looking brown vans delivering parcels on time ? it stands for uninterruptible power supply, a device intended to protect computers from the vagaries of the mains electrical supply. But just like the Yellow Pages, UPSs aren?t just for the bad things in life, like power cuts, they perform other tasks such as mains smoothing and conditioning, to take some of the rough edges of the supply.
UPSs come under the umbrella term of ?power protection?. This embraces a range of solutions, from a device as simple as a surge suppressor to tame electrical spikes to something as complex as a UPS to keep the power flowing even through a blackout.
Most network managers realise that uninterruptible power supplies are not a luxury but a necessity. If the power to servers is turned off without properly shutting down the server, data can be corrupted, so it?s prudent to protect all servers and any PCs that run critical applications from power cuts. The cost to the company in the time it takes to restore the data from backups, plus the interruption of service while the server is rebooted or recommissioned, can easily run to thousands of pounds an hour. Thus the cost of a UPS, which ranges from #150 to #3,000, is easily justified.
UPS may be de rigueur in the US, where voltage drops (or brownouts) are common and almost as bad as blackouts, but the quality of the mains supply here in the UK is somewhat better ? it is more reliable and consistent. In my office in Hackney I can recall only three total power cuts in the past 10 years, mainly caused by ?hole diggers?. Power supplies to rural locations are less dependable than those to city addresses, particularly during winter. But the fact that power cuts in the UK are rare doesn?t reduce the need for power protection ? anyone running a mission-critical system does not want any disruption to the power supply.
A growing problem is the increasing use of switching power supplies, which take AC current in short gulps that reflect back into the power lines as harmonic currents. These currents cause non-linear voltage drops in the power lines that can affect other loads. In Europe, this problem is being addressed by the harmonic current standard EN61000-3-2, originally known as IEC555-2.
The requirements of the UPS itself are changing. More specifically, there now appears to be a dual requirement: more of a critical demand to ensure data is saved on a power down, but also for equipment to stay up for longer in critical applications. In the past it was sufficient just to deliver power for 15 minutes, giving the user time to store data before shutdown. Then, when networking arrived, people didn?t want to be down even for a minute.
Today, with the influence of telecoms, the demand is for 24-hour backup and there is a minimum requirement of staying operational for one to eight hours. As a result, sales prospects for UPS remain healthy, particularly in the server market, and are boosted by burgeoning interest in the internet and Web servers. The UK market is dominated by two vendors, American Power Conversion and Emerson, although HP, Exide and Toshiba are also players.
In effect, all UPSs provide the same basic functions. Besides powering servers when the main electrical power is lost, they also protect against power spikes and voltage reductions. Some also provide surge protection for network or telephone connections as well.
The better UPSs constantly monitor and regulate the quality of the power coming from the mains socket (called line conditioning or power conditioning). And with the advanced software available with higher end units, network administrators can check the status of the UPS from a network node or even via the Web, and be alerted via email or pager if there?s a problem with the power in the server room. If power isn?t restored before the UPS battery is depleted, the software can initiate a graceful shutdown of the system, warning each user of the impending failure and even saving user files.
Most UPSs do what they should ? when the power fails they kick in, reliably and repeatedly. What differentiates them is how well they manage emergencies automatically and how elegantly they handle each power event, such as blackouts, voltage drops, surges or spikes.
The basic UPS is a fairly simple device. It consists of a battery, an inverter that converts the DC output from the battery into AC, and a charger that keeps the battery at full capacity. The capacity of the UPS is rated in volt-amps (VA). The devices come in a range of sizes, from #100 units that can keep a desktop PC running for a few minutes to #3,000 battery-packed giants that can keep a corporate data centre up for half an hour or more. The largest market segment comprises units in the 1kVA to 3kVA range, which are robust enough to protect and power two or three network servers, their monitors and modems.
A UPS system has three main components: the rectifier/ charger, the battery and the inverter. The rectifier/charger converts alternating current (AC) to direct current (DC) and supplies the input power to the inverter. The battery accepts direct current from the charger and uses it to keep the battery charged on float (standby) until it is needed. During a power failure, the battery supplies the input power to the inverter for a predefined period of time.
UPSs fall into two broad groups: standby/off-line systems or online interactive systems. Standby systems are the less expensive of the two: they consist mainly of a battery, fast switch and power sensor. They cut costs by using a smaller charger, one which is incapable of handling the load by itself and only able to charge the batteries slowly, which means a period of a few hours or overnight when the devices can deliver little or no backup protection after a failure. When the UPS senses that power has dropped below an acceptable level, usually just above 200V, it switches to the battery. The switch takes place in 2ms to 10ms, but PC chips can usually ride out a power loss of several hundred milliseconds. Beeps alert users that power has failed, if it isn?t already obvious.
The classic UPS is the online or double-conversion design, which uses batteries as a power buffer to continuously absorb spikes and surges. This kind of UPS first converts utility power to low-voltage DC for the battery, then converts that back to the AC needed by the PC. The batteries are always online, so power is never interrupted by switching even in the event of a power failure.
Some vendors offer hybrid systems that combine aspects of online and standby UPSs, keeping the inverter powered but not constantly draining and recharging the battery. This extends battery life but reduces the switchover time to zero, just as with an online UPS. Some UPSs can switch from one mode to another, going from standby to online as necessary.
A third type is the line-interactive UPS, which goes one step further by constantly monitoring the integrity of the mains power, increasing or decreasing voltage to a PC whenever line voltage rises too high or falls too low, and switching to battery power if the deviation becomes too great. This type is cheaper and claims to offer better long-term reliability.
The capacity of a UPS can be measured in two ways ? as power or energy. Power (measured in VA) determines how much equipment the UPS can safely operate. Energy (measured in Watt-hours) determines runtime ? how long the UPS can operate your equipment from its internal reserves. IT managers often mistakenly opt for a UPS with a higher VA rating to get a longer runtime, but a more cost-effective approach is to buy your extra runtime in the form of greater battery capacity. Many UPS units allow additional battery packs to be daisy-chained to deliver increased uptime.
Nearly all UPSs have a serial connector that allows UPS software on the protected computer to monitor the UPS?s status and initiate a shutdown once the power has been off for a specified time. Some UPSs allow other connections, such as an AS/400 jack or a 10Base-T network connection.
Some UPSs offer multiple serial ports so that several protected servers can monitor the status of the UPS, which eliminates the need for a separate UPS for each server. This can cut costs through better use of UPS capacity.
In time, many UPSs will also include a connector for the universal serial bus (USB), a new specification that is slowly being adopted by peripherals manufacturers. USB allows 126 devices to be connected to a PC, rather than the four devices using standard serial ports (which, in practice, is often only two devices, because of conflicts and a limitation in the number of physical ports available on most systems).
Another connectivity option on many UPSs is the SNMP module, an optional Ethernet or Token Ring module that can be installed in a special slot. This allows the UPS to be monitored from anywhere on the network, rather than only through the server?s serial port. An SNMP module can provide two useful functions ? it allows the monitoring of UPSs without having to run the UPS software on the server, and it allows the monitoring of network devices such as routers or switches where running UPS software is not an option.
All UPSs use some form of sealed lead-acid battery, similar to those used in cars. Its working lifetime is three to six years. As the battery ages, runtime gradually declines, then drops off dramatically. Better UPS designs monitor battery condition with periodic tests and warn when failure is imminent. Better software lets you program test times and logs the results.
A key development in lead-acid technology was the introduction of valve-regulated lead-acid (VRLA) batteries in the early 1970s. VRLA batteries are smaller and more powerful than the earlier, flooded lead-acid ones and they are maintenance-free. They have no free electrolyte, which means they can be used in close proximity to office personnel.
Nickel-cadmium (Nicad) batteries develop memories, which decrease their backup times and make them inappropriate for most UPS applications. Although Nicad batteries are powerful and unaffected by wide ranges in temperature, this characteristic is of value only when a UPS is used for ruggedised or outdoor applications. Lithium-ion and nickel metal hydride (NiMH) batteries have size and energy advantages over lead-acid ones, but UPS customers are not usually willing to pay the extra cost.
Most UPS suppliers make allowances for anticipated load growth. Generally, a UPS battery is oversized for current use and will be sufficient for some future expansion. Typically, load growth is the result of an add-on to the original system, such as a telephone system or new computer system. Batteries have a finite capacity: the more the load is increased, the faster the running time goes down, accelerating in a non-linear manner. A 20 per cent increase in load results in a 40 per cent reduction in runtime.
By itself, a UPS only delays power failures, giving you time to shut down the system before the batteries run out. The link between UPS and software can make the shutdown procedure automatic, so your PC or network powers down normally and resumes normal operation when utility power returns. Management software using SNMP has the advantage of being based on an open standard, allowing the monitoring and control of many UPSs from different manufacturers and a wide choice of software. The disadvantage is that monitoring and control functions are limited to functions that have been defined as part of SNMP.
Proprietary software allows much more extensive control and monitoring of the UPS, but each manufacturer has its own implementations and none is compatible with any other. UPS software uses the polled information to display the condition of its input and output power and battery reserves.
Most UPS software packages allow remote administration, so you can dial into your UPS and tell it what to do. The majority link to major network management packages such as HP Openview, to let you manage your UPS as a device on the network. They also provide links with applications such as Microsoft Office and Lotus Notes, automatically naming and saving open files before shutting down the system.