Take the 2-minute tour ×
Server Fault is a question and answer site for system and network administrators. It's 100% free, no registration required.

I wounder how important it is to have an UPS for a server. Everyone seems to use UPS for servers but not for client computers.

In example, if I use Windows Small Business Server on a small network for Active Directory and storing the users home directory for backup is it still useful with an UPS, why? If a power outage happens, shouldn't the filesystem handle that, so no files are corrupted? And in this case I think the problem is the same on the clients.

I do understand that it can be useful on servers when changing the server configuration, but that is done during a very limited time and it sounds like it's more worth to do good backups before the configuration.

And when thinking about database transactions, I think that the transaction logic in the database management system should handle it, that's what transactions are for.

And finally, if an UPS is used, doesn't the server applications need to have support for it? outherwise the power outage is only delayed? Or is it only the operating system that needs the UPS?

Why is an UPS needed for a server but not for clients?

share|improve this question
8  
I almost want to downvote this, because it is ridiculous to even think about NOT having a UPS on a server. –  DanBig Sep 28 '10 at 20:50
    
@Dan: I use an UPS, but mostly because everyone else use it and it seems to be a best practise. But now I wounder why? I rarely hear that UPS is used for clients and if I have a low-end server just for some file sharing I wounder if it still applies. In short the question is mostly for learning. –  Jonas Sep 28 '10 at 21:02
1  
You don't care about the clients because you don't [shouldn't] be storing important data on their end. You store important data on the server and grant clients access. If you lose power you want to be sure the server can shutdown gracefully as to avoid data loss/corruption. –  jscott Sep 28 '10 at 21:13

5 Answers 5

up vote 11 down vote accepted

The brief interruption to someone's work is more likely going to cost less than buying and maintaining a desktop UPS. That's one of many reasons why many professional solutions are server based and centralized, as the service level can be focused there.

If there is a business justification with Return On Investment (ROI) in your environment, I would suggest a cost analysis between providing generated or battery backed power to the entire building versus buying individual UPSes for every workstation.

A server does not necessarily need to "support" a UPS, it's just a battery that's inline with the power source. Any "support" is going to be automatic shutdown or taking a preferred action in case of power loss.

The idea is to protect the equipment by conditioning the power as well as allowing the equipment to be properly shutdown in case of an extended outage. Depending on your budget, it is also often to maintain availability in case of power loss but with that service level you often have power generation in addition to a UPS.

A UPS is in addition to backups and has no direct correlation to backups beyond being a best practice within properly run IT infrastructure.

Without a UPS you could lose data, have hardware failure, and have system availability interrupted. It is often easier to find the ROI here than with workstations.

share|improve this answer
    
Brilliantly stated. –  Cypher Sep 28 '10 at 23:29

Accidents do still happen. RAID cache batteries not working properly is a concrete example, but journaling filesystems and transactional integrity simply aren't guaranteed when the power is pulled.

Plus, any UPS worth a damn also does line conditioning to protect the hardware from current and voltage variations, extending the lifetime of your hardware.

Why servers and not desktops? Some places do, but it's expensive. How much is the data on any client worth? How much is the data on the server worth?

share|improve this answer
    
I could use built-in software mirroring in NTFS, and journaling filesystems are constructed to handle situations like this. I have used NTFS on my Desktop in ten years and never experienced problems with it or heard anyone who had problems with NTFS. I agree that an UPS can protect the hardware, but then again, I'm not sure if an UPS is worth it just for that reason. –  Jonas Sep 29 '10 at 15:32
    
Anyone with a decent amount of experience will have experienced "situations like this" where it did not go to plan. –  mfinni Sep 29 '10 at 16:42
    
Can you explain how? Do you think the journaling algorithm is wrong? –  Jonas Sep 29 '10 at 21:33
    
@Jonas - I know bupkis about how good the journaling algorithm is for extfs2, or NTFS, or JFS, or anything else. I'm certainly sure they don't use the same algorithm, so how could I claim it's "wrong"? You're missing my point. I've seen occasions where abrupt power-down has caused loss of data and corrupted disk. You're seriously arguing that UPS is a waste of money for datacenters? How many have you designed? How many have you worked in? –  mfinni Sep 30 '10 at 0:27
    
Journaling filesystems are not designed to prevent all data loss! They are designed to prevent a corrupt filesystem by making sure to write data to the filesystem in a safe way. They can still lose data that is not committed yet. At least the filesystem will be sane, and there's no need for a lenghty fs check on startup. –  Martijn Heemels Feb 25 '12 at 17:38

There are many, many reasons to use a UPS, but using your example of DB transactions, look at it this way. Even if a transaction is committed, you lose power, and your RAID write cache (you are using battery-backed write cache, correct?) batteries aren't functioning properly, you're out of luck.

Additionally, power outages are notoriously hard on equipment. Replacing a $500 workstation is one thing, but it's much more painful (both financially and otherwise) to replace a server that costs an order of magnitude more (data on that server notwithstanding). Most good UPSes also are line-interactive and will offer much better protection from power surges and brownouts than you'd get with a cheap power strip.

Overall, it's just a very well-accepted best practice.

I do understand that it can be useful on servers when changing the server configuration, but that is done during a very limited time and it sounds like it's more worth to do good backups before the configuration.

What does using a UPS have to do with changing a server's configuration?

share|improve this answer
1  
If your database says a transaction is committed before it makes it to disk, you need a new database! –  EvilRyry Sep 28 '10 at 20:31
    
While the change of a configuration occur, the system isn't in any consistent state I guess. All system changes isn't managed by a transaction manager, I guess. –  Jonas Sep 28 '10 at 20:37
    
@EvilRyry - I re-worded my answer. –  EEAA Sep 28 '10 at 20:39

And finally, if an UPS is used, doesn't the server applications need to have support for it? outherwise the power outage is only delayed? Or is it only the operating system that needs the UPS? Why is an UPS needed for a server but not for clients?

Who says that clients don't need a UPS? All systems in our environment are protected by a UPS - from every server to each desktop.

You are correct that a UPS only delays the power outage. However when a power loss occurs the UPS communicates to the OS through a USB cable (most common nowadays). This allows the UPS to notify the OS when the battery is about be empty and the OS can begin gracefull system shutdown. This is something that is not possible in a normal power outage without a UPS. The power company will not call you and tell you that a tree is going to fall in 10 minutes and kill your power.

Your assumptions about how a computer handles a loss of power are not realistic. Most of the time there might not be any harm, but there is no need to gamble. One problem on a production server and you will rue the day that you decieded you did not need a UPS.

share|improve this answer
    
Exactly, a friend's company spent 48 long hours rebuilding their Active Directory because their servers didn't fail gracefully during an outage. It will often go just fine, but when it fails, it can be a disaster. –  Martijn Heemels Feb 25 '12 at 17:42

Jonas,

It's a wonderful thing to THINK the OS should handle an outage - and it SHOULD. BUT, reality differs. Reads and writes take place very quickly but if it loses power at the wrong moment, you lose the information.

Do you want your server rebooting because the lights flicker? Some servers can take 10-15 minutes to start up and then have to go through a disk check when they aren't protected. Users accessing the server via laptop and the internet won't necessarily experience a problem, but you could end up WASTING your employee's time or worse - if an employee is in the middle of a proposal and hasn't saved the file, you could lose business. UPSs are not very expensive in the grand scheme of IT costs and should be used in ANY place that is subject to even occasional outages. If you think the $100-300 isn't worth it for the server and $50 isn't worth it for the PC, that's your prerogative. But most people understand TIME is MONEY and any disruption in work flow will cost you money.

share|improve this answer

Your Answer

 
discard

By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.