RAID for automation control application

There are several types of raid. Some use two drives and if you lose one, you replace it and the system sees the new drive and rebuilds the new one into a copy of the good one. Others use 3-5 drives and write the data over a number or drives in stripes and you select the diversity needed. Some raid arrays of 5 drives can tolerate the loss of any two drives. Obviously they are only storing 3 drives worth of data on the 5 drives. There are a few places to look this up.

here is one

http://en.wikipedia.org/wiki/Redundant_array_of_independent_disks

First, welcome and congratulations on you first post. The answer is, Yes. It is not surprising that a person might be far from up to speed on RAID: the descriptions out there being so unnecessarily confusing, and frequently either poorly written or incorrect. NO offense but, the wording of your post makes it confusing as to whether you are presently using Raid, or not? When you say increase availability, what are you understanding that to mean? or what deficiency of availability would you hope to correct, if any? You have asked how RAID might be configured by describing a basic RAID configuration. Finally, why exactly do you need RAID mirroring for your setup? Might it be because your input devise or channel is very unreliable? Is the data so sensitive that immediate recovery (or delayed "repair") is essential? So that using the extra HDD for periodic back-up (timed according to you experience of "failures" would not be just as effective, without incurring the need of greatly increased storage media which is necessitated by RAID? What do you mean, ...regular workstation is not so reliable? And what leads you to believe that adding complexity to its (or your network's) storage architecture will make it more reliable? If you could clear up issues like these, it might make it easier for responders to get a handle on your question in order to provide more targetted responses. You also should find this helpful since your query could easily become overwhelmed by a dizzying array of responses describing many possible permutations of RAID configuration, of which only a few, or only one actually addresses your situation.

Consider RAID 1 which is disk mirroring. This we have found is thwe best way to handle dirve C redundancy. It is possible to use RAID 5, I prefer afrter year of experience to keep drive C seperate from all other data drives. Raid 1 can be done in hardware or in the OS your choice.

Basically the others have answered you, but maybe there needs to be a little clarification.

The RAID 'numbering' does seem designed to confuse, but basically 1 is 'mirroring' - two drives being identical. Writing is done to both, reading from either. (The system takes care of this, you don't have to do it.) If you are using it for your primary drive I would recommend using 'hardware' RAID. That is a physical disk controller is doing the job rather than the operating system. O/S RAID doesn't always like it when the 'real' C drive dies. This is the more expensive solution - and still has a single point of failure in the controller and PC mother board.

You don't say what level of reliability you need. That should influence your decision. Ignore computerese like "mission critical" (a phrase designed to make you feel important and therefore pay more for stuff you don't need) and look at the reality of your situation:

• Does your process stop regularly for maintenance?
• Can you or do you go off-line regularly for reasons other than the PC's involved?
• What does it cost to have the process stopped due to machine failures?
• Do lives depend on it operating 100% of the time?

If your process stops for maintenance or just so people can go home, you really only need disk mirroring - preferrably but not necessarily by hardware - and you don't need 'hot swappable' drives because a dead drive can be replaced during the maintenance period. Same thing if other parts of the whole process fail regularly. Do the computer maintenance when the other stuff is down.

The cost of stopping the process will also let you know how much it is logical to spend on having the computer part stay up. From doing nothing and relying on well made hardware & software, through RAID and other relatively inexpensive improvements, through to doubling (or tripling) everything involved so failures just don't occur.

As to 'server' vs 'workstation' these terms are generally misused when applied to physical machines. Both can do the job of the other with the real differences being in how some of the functions - generally I/O - are configured. Operating as a servers generally involves more I/O and less actual processing so machines iwth faster or more disk and networking controllers are needed.

Reliability has nothing to do with the workstation - server choce. If the workstation is failing, don't buy a server from the same company.

Congratulations, you just discovered the Achilles heal of an automation system.

One important issue to consider on hard drives is that they rarely fail in an instant. Bad drives will begin to record corrupt data until the point where the will no longer function. With a RAID-1 configuration the backup drive will be a perfectly mirrored corrupt copy. We learned this the hard way.

In our opinion the RAID is only addressing the symptom of a fragile drive and frequent failures. Modern solid-state flash drive technology is far superior in reliability and has lowered in price to the point it is now economically feasible to specify. For applications where large storage capacity is required, you can put the operating system on the flash drive and backup non-critical process data onto a rotating or removable drive.

Regards, Delmar Schmidt
Melfi Technologies Toronto
www.melfitechnologies.com

Usually, RAID 1 (mirroring) does a write to one drive then an independent write to the other. If one drive records a corrupt sector (not all that common with the error checking and read-after-write verifications done by most things today) the other should not. If your RAID system is set to write on one drive then copy from that drive to the second, the real error is in the system chosen.

Disk drive manufacturers are going to hate you for the second alternative, though I would not count on a single one of those, either, if it were in, for example, a life support system. We use memory disks that plug right into IDE ports for networking applications where weather may be unfriendly to drives and maintenance, while possible, can be a major pain.

Commercial airliners typically use a three-copy system (for example three completely separate autopilots). This gives the advantage of a majority opinion since failures most likely occur one-at-a-time.

Again, it comes down to the application and how important is 100% uptime.

If your gathered data from the PLC is not as important as getting back up and running after a drive failure, forget RAID and just duplicate your C drive immediately after installation and start-up, while it is good and keep the duplicate as a spare. When the C drive goes down, grab the spare, install it and and off you go.

If the data's the thing, consider RAID plus periodic backups.