Hot spare and hot swap/Raid 5 configurations 6

[psasu]

I have started the mcse course.What is the difference between the terms hot spare and hot swap and raid 5 configuraions.
I would like to build a win2k server and incorporate these features.
What best quality parts do I need to buy to accomplish this project and how do I configure the disks to be raid 5.
I built the win 2k professional desktop myself to help me pass that exams so I believe if I can do that of the server, it will help me in my studies

 

[lesgray]

In simple terms:

Hot Spare: Drive in the case, that is hooked up to IDE or SCSI chain, ready to be used but not actually part of the raid setup, that the raid controller knows is available to be used to replace a malfunctioning drive on the fly. (bit more comples than that , but basically thats it.

Hot Swap: removable hard drive bay that is electrically capable of being removed while the machine is running and a different drive inserted on the fly, so that a defective/malfunctioning drive can be replaced without impacting the server functions.

Raid 5: one of several raid configurations. Raid 5 requires a minimum of 3 drives to work and can be made up of as many drives as you have either SCSI or IDE connections on your raid controller(s). If software based raid, as many drives as can be accessed by your setup. Generally speaking except for JBOD Raid (Just a bunch of disks) all drives should be the same size and speed. For forms sake and consistency, you usually want them of the same make, but that is not stricly necessary. If they are not of the same size, then the raid size will be determined by the size of the smallest drive in the raid.
Just for reference:
Raid 0 (minimum 2 drives) is what is termed a stripe. All drives in a Raid 0 are formatted as a single volume with no consideration for redundancy (backup information). Volume size if n x GB, where n is the number of drives and GB is the capacity of the smallest drive in the mix. Very fast as the raid controller can write information to any drive, so your data is spread out over the entire set of drives. If a single drive malfunctions at any time all the information is gone. Useful where data is not paramount, but speed is.
Raid 1 (2 drives)is termed a mirror. Each volume is composed of 2 identical drives (if the drives are not the same size, then the volume size will be at the capacity of the smaller drive). The system will only show one set. Anything written to the visible set is also written to the mirror set. If something should happen to either set, the raid controller will warn you that there has been a failure and you will have the opportunity to replace the failed set without losing data. Of course if both sets of drives fail at the same time, all data is lost, but that would be extremely unusual. Useful where redundancy is necessary but speed is not, as speed is impacted by the raid controller having to write the same information to both drives at the same time.
Raid 0+1 is a combination of the two. Two identical stripes are mirrored. Bit faster than Raid 1 due to the nature of Raid 0.
Raid 5: Probably the most common raid setup for file servers. In this setup (minimum 3 drives) you have elements of both Raid 0 and Raid 1. All the drives are striped together, but a portion of each drives capacity is reserved for redundancy information. However no redundant information for a particular drive in the raid is kept on the same drive as the original information. In this way a drive may fail, but the raid keeps on serving or writing information as all the information that was on the failed drive is still available the raid controller. Raid 5 is very fast, basically the same as Raid 0, but it gives you the peace of mind of Raid 1 - all your information is safe in the case of a drive failure. Raid 5 volume size is (n-1) x GB, where n is the number of drive and GB is the capacity of the smallest drive in the raid. For example is you had 4 x 120GB drives in a Raid 5 configuration, then your volume size would be (4-1) x 120GB or 360GB. The missing 120GB is for the redundant information, so the raid controller may rebuild the original configuration if a drive malfunctions and is replaced and also for the raid controller to keep the volume alive and working until such time as the drive is replaced. Raid 5, at least in my experience works more efficiently with more drives. The minimum number is three, but I never setup one with less than 4, and I like to use at least 8 if possible. That last bit is my opinion and I'm know there is way more variations on that out there and someone will differ greatly with me, but thats what works best for me.

I'll continue in the next post.



Les Gray

http://www.creartec.com

 

[lesgray]

You can build a Win2k server with a Raid 5, completely in software, which is the cheapest way to go. Not the most efficient or fastest, as a hardware solution is better, but certainly cheaper and will do the job. All you need is a boot drive for the OS and three similar drives for the raid. After loading the OS on your boot drive, use the computer management console to access the Disk Management console and you can create a Raid 5 when formatting your three or more raid drives. You'll have to make them dynamic partitions. Depending on the size of your drives the OS will crunch away for awhile and you will the be presented with a raid 5 volume equal to the size of two of the drives. You can then load some data on the drives and if you like you can fool around with the raid by shutting down the system and disconnecting power from one of the drives and see the system respnse in the Drive Management Console. If you have another drive the same as the three (or more) you have used to setup the raid 5, you could even hook it up and rebuild the raid thru the Management Console and see how thats done. Messing around like this will give you experience in how the thing works.



Les Gray

http://www.creartec.com

 

What best quality parts do I need to buy to accomplish this project and how do I configure the disks to be raid 5."

To construct a better raid 5 system with hot swap and hot spare capability you would be better off with a hardware solution, i.e. a separate PCI Raid Controlle with Raid 5 support. For IDE Raid, Promise makes a couple of good, not too expensive controllers with the capacity to handle four ide drives. They of course make controllers that can handle more drives, but they tend to be very expensive. Promise also makes sATA Raid cards which offer better throughput than standard ATA drives, but these tend to run a little more money.

The FastTrak SX4000 (FASTTRAK-S150-SX4 is the sATA equivalent) is a nice lower priced Raid 5 solution that has the features you are looking for. It supports hot spare and hot swap. The limitation of the card is that it only supports a maximum of 4 hard drives. Having said that, with the fairly large IDE drives available today, you could put four 300GB drives in the raid without a hot spare and have a raid 5 volume of about 900GB. Should be enough for most of us

There are other manufacturer's cards that are similar in cost and performance. I only mention the Promise ones, because they tend to be widely available.

Hot swap capabilty is built-in to the better removable drive bays. Look at the documentation before buying to make sure it will do what you want. Since most of the controllers with Raid 5 support support hot swap bays as long as you label your bays properly, you will know which drive, if any, has failed and which bay you need to pull to replace the drive. The hot swap feature will ensure that the raid continues to run, while the drive is being replaced and the raid controller (most of them anyway) will automatically start rebuilding the replaced drive as soon as the new drive is in place.

As far as hard drives go, these days use only the drives with 8MB cache. The cost difference is minimal and performance is greatly enhanced. I prefer to use Seagates and Western Digitals. I don't use Maxtors, because lately I have had a lot of trouble with failed drives. This is strictly my opinion and I know many people who only use Maxtors and are very happy with them. If you want the best in ATA drives and money is no object and super big size is not a consideration, you will probably want to look at the Western Digital Raptor sATA drives. I think their largest one is 74GB or so, but they have a have a spindle speed of 10k rpm, which I believe is the fastest around.

I have't discussed SCSI drives, not because the ATA/sATA solutions are better, but because they are far more expensive and probably not what you are looking for.

Any questions, please ask...

 

[lesgray]

BTW, in my first rambling post, I mentioned JBOD Raid. Termed Just a Bunch of Disks (Drives), it is a striping solution similar to raid 0, that can utilize many different sized drives without restricting the maximum size of the volume to n x the capacit oif the smallest drive in the raid. In other words if you constructed a striped raid of 1 x 40GB, 1 x 80GB and 1 x 120GB you would get a volume of 120GB with raid 0 (3 x max size 40GB- smallest drive), whereas with JBOD you'd get a volume of size 240GB (40+80+120GB)

Les Gray

http://www.creartec.com

 

[Pagat]

lesgray, well done. voile a star for you
tia

 

[Echoe]

Wow, very useful information in both of your posts. Well done! I had read smaller descriptions before, but none as good as this....