Software, Onboard, Hardware RAID... what is best in this situation?

Raid 1 (and 0 for that matter) from motherbd and software raid work just fine, The problems gets worse once you move to the higher raid levels which need parity calculations performed on each byte.

So overall you should be fine with motherboard or software raid1. The only issue you have is protecting your students data from theft or intentional damage (students can be vicious LoL) I would suggest you make them copy it to flash drives also.

RAID1 is not a good backup system. It was not designed for that. Granted, it keeps a mirror copy of the first HDD in case one of them fails. But it does NOTHING for other types of data loss: malware which corrupts files, bad data written because of power glitches, file deletions by mistake, fire and flood, etc. All of those things will be mirrored onto the second HDD unit so BOTH are damaged.

However, if you want to use RAID1 as a simple but imperfect backup system, go ahead. Listen to popatim: make them do their own backups TOO, onto flash drives or whatever, and keep them safe elsewhere. Besides being an improved backup system, you will teach them to DO BACKUPS routinely.

Now, differences you asked about.
Software vs Hardware RAID
Software RAID is programmed procedures as code in a form of ROM. It is executed by the main system, using system resources like CPU time and RAM space. There are two sub-types.

"Built-in" RAID on a mobo means the RAID code is included in the BIOS chip on the mobo. As such it comes cheap, may be limited to only some RAID forms (but almost always includes RAID0 and RAID1 at least), and also uses mobo HDD ports and controllers. Its consumption of system resources is small, but not zero, so it slows other processes slightly.

Cheap add-on RAID controller cards often also are really using Software RAID. The main difference is that their code is in chips on the board, but that code still is executed by main mobo resources. These cards, however, contain their own HDD controllers and ports so they can make it easier to add several HDDs to a system for a large RAID array.

True (and more expensive) hardware RAID cards have their own code chips and HDD controllers and ports, but they also have their own processor and RAM, so virtually all the work in running the RAID system is done on the added board, and not using main mobo resources. Hence, these systems have a really minimal effect on system performance for other tasks.

As popatim has said, more advanced RAID systems like RAID5, RAID6 and RAID 10 require more processing work. Hence such systems impose a larger impact on system performance if you're using a software RAID. That is why so many of these, especially at a professionally-administered level, are done as true hardware RAID systems.

One major factor to consider for ALL forms of RAID is that there is NO "Standard" way to do it. Every RAID system does it the way its maker decided. This can become a problem when a RAID system fails - not the drives involved, but the management system. The data are encoded on the drives in a particular way, and can only be read back in exactly the same way. So what happens if your RAID management system - and this could be on-board RAID in BIOS, a low-cost software-type RAID card, or a more advanced hardware RAID card - fails? The simplest answer is: you need to find an identical RAID management system and use it to replace what failed. These days, however, the major makers of RAID systems at least offer some assurance that this will work. Many basically use the same RAID coding system for all of their products over a long time. For example, years ago nVidia assured people that all its future supplied for use on mobos as "built-in" RAID systems would use the same code. Thus if your mobo with an nVidia RAID system failed, you could replace it with an new mobo made by just about any company as long as it also used an nVidia RAID system, and it would work with the old drives. I actually had to take advantage of this feature once, and it worked perfectly.

If you're using an add-on RAID card, either low-priced software type or high-priced hardware type, what you need to do with the FIRST system you build (planning for future failure recovery) is choose a card maker with a long and stable history who are likely to still be around in 5 to 10 years. If failure happens, you consult them about what system failed, and what new system they can offer to work with the old drives.

If none of that seem possible with a failed RAID system, there are software tools that claim to be able to read almost any drives written by any RAID system. Some of these certainly work, but I have no direct experience to tell you how reliable this approach is on a wide scale.

Having said all that, of all the RAID systems, RAID1 is usually the easiest to handle in cases of a damaged RAID management system. This is because almost all RAID 1 systems write identical data to both HDD units and do so in the SAME way that a normal HDD is written. The only real difference from a "normal" one-drive system is in some of the data structure details at the beginning of each HDD unit. So there are several ways that a RAID1 drive full of data can be made readable by a "normal" system. In fact, almost all RAID 1 systems have a simple way of "breaking" a RAID array of 2 HDD units back to individual "normal" drives that are fully readable, and just happen to have identical data on each unit. This feature is really useful if the failure is in one of the HDD's in the system. The management system can "break" the RAID array, identify the failed HDD so you can replace it, then rebuild the array by copying data back from the surviving unit to the new replacement unit.

Be warned: RAID0 is very different, and one result is that BOTH HDD units MUST be working, or you can lose ALL the data the RIAD0 system contains. This means the probability of total data loss in a RAID0 system is twice as high as a "normal" single-disk system.