What is the difference between an online spare and a hot spare?

Report this post to a moderator

As the lack of replies reveals, there is not an easy answer for this one. I'll give you my interpretation.

Online spare would imply some sort of redundancy (ie. fault-tolerant). Industry never really came to a de facto standard for fault tolerance. Clustering is used for many critical apps., nowadays, to provide fault tolerance and load balancing. You see a lot of fault tolerance in telco and the process industries.

Hot spare is a replacement unit standing by with power applied. Not connected but ready to be placed into service. A hot spare is used when a short amount of downtime is tolerable. Not to be confused with hot-swapable which allows a unit, like a power supply or disk drive, to be replaced under power.

I'd welcome anyone else's slant.

__________________
it's just a jump to the left...
Tcat's Security+ PDF
Tcat's Network+ PDF
A+ 4Real - the missing link

Last edited by Supertech on 01-04-02 at 01:08 AM

Report this post to a moderator

This is indeed one of those awkward questions and I cannot disagree with Supertech on his comments.

What I may add to his hot spare comment is my way of remebering what it is:

A hot spare is one which is working under the same conditions as the main unit, except that it is not used. It is ready to take on the place of the main unit as and when failure occurs.

Hopefully we have both enlightened you!!

__________________
One Exam leads to another! Where will it ever end?

Report this post to a moderator

A hot spare has to do with hardware level raid. Lets say you have a server that holds six hard drives internally (just so as to not deal with external enclosures, etc) You have 2 drives set up as a RAID1 (mirror) and 3 drives set up as a RAID5 (stripe set with parity). That totals 5 drives. The sixth drive is usually on standby, or basically just sitting there. It is a hot spare. If one of the other drives fails, usually the RAID card (either PCI or integrated on the motherboard) does an automatic rebuilding of data to the hot spare drive. If one of the RAID 1 drives fails, then the RAID card will use the hot spare to re-create the mirror without any user intervention (usually). I am not sure about an online spare. Sometimes, the extra disk on the 6th disk would be distributed across the the stripe set. That is also called an enhanced raid level, but that is a little confusing. If you want more info, check out www.redbooks.ibm.com and do a search on the redbook entitled "Netfinity Server Disk Subsystems" I hope this helps.

Robert

Report this post to a moderator

Oh, also, the redbook site lets you view the Redbooks in PDF of HTML format for free, and there is a ton of good info, but pertaining to the different RAID levels, there are some good explanations and pictures of hot spares, etc.....

Report this post to a moderator

Since he didn't mention RAID, I have to assume that the bible dude is speaking in general hardware terms in which case the term is subject to Vendorspeak.

Is hardware specified in the RAID specification? or are you refering to IBMs interpretation of the spec.?

I don't wish to appear argumentative, I have just seen the subject handled differently from different vendors and different gear.

__________________
it's just a jump to the left...
Tcat's Security+ PDF
Tcat's Network+ PDF
A+ 4Real - the missing link

Report this post to a moderator

I have never heard of anything else being referred to as a hot spare. I could be completely wrong, but I think that is what is being referred to. If you go to your search engine, and type in RAID Tutorial there will be more info on types of hot spares, etc. Also, IBM's RAID cards (and their onboard scsi controllers) are all made by Adaptec, and they are pretty standard... I hope this helps....

Robert

Report this post to a moderator

From a "hardware" perspective (i.e. desktop/servers) the only case of something I can think of that might be termed "online spare" might be one of the configurations we have here on an older server set. We have a couple of Compaq Proliant 5000's running Netware. They are both operational, both turned on at all times, both attached to the same array and there is software running that recognizes when the primary server goes down, and automatically switches over to the back-up or failover server. As to hot spare, we've always referred to systems we have configured and "waiting on standby" (usually desktop level) that are configured and ready to be deployed to replace a failed machine for a period of time while the failed machine is being repaired. On the RAID level, Robert is absolutely correct!

Report this post to a moderator

I remember hearing about the use of mirrored servers for fault tolerance, redundancy, and load balancing. The "Mirrored" or "online" server is used to help ease the load on say a webserver. Both servers are kept synchronized. It's used in this way, except for when the main server crashes or there are hardware problems. In that case the online server takes over instantly to guarantee that the resources will still be available.

Please let me know if this is off base


Chris

Report this post to a moderator

No, you are not off base at all. The mirroring of servers is usually called clustering. Microsoft has an elective exam dedicated to this. There are a couple of kinds... Load balancing clustering, or high performance clustering, and high availability clustering. Load balancing would be when 2 servers share resources usually with an external storage unit of some type, either scsi or fiber channel, and high availability is almost the same, but the servers are there to mirror one another. I hate to be IBM specific because that is what I know, but they have a couple of redbooks related to clustering, and a free online course about clustering... go to www.pc.ibm.com/training, and get a customer log in (just fill out the form and its free) and you can have access to 15-20 online courses that get pretty in depth and a specific course on clustering. Another cool thing about clustering is that the university of new mexico clustered something like 72 intel based servers to make a supercomputer that compares with mainframes.... sorry that we seem to be getting away from the online spare subject, but I love this stuff!

Robert

Report this post to a moderator