At the moment, I'm leaning towards the Areca 1212, don't think I'll need more than the 4 internal drives for the time being, really I just want Raid for my 'important' data. I'm planning on using a separate Raptor drive (scavenged from my pc) for operating systems and applications. Could you tell me, with Raid 5, when a disk failed the array rebuilt itself on the fly and you could still use the OS (the array was vulnerable, but still usuable), is this the case with Raid 10?
Generally, I'd recommend getting an additional 4 ports for future expansion. It's cheaper and easier (saves time) to just add drives than to have to swap them out each time you outgrow the capacity.
There's already a couple of users who've run into this. Bought a 4 port card, and realized they've already grown in capacity to need the extra ports.
As per the rebuild question, Yes. If a drive drops out, the array will continue to work in degraded mode. But it's best to get a new drive in as quickly as possible (assuming it's mechanical in nature, otherwise it could be cabling, or firmware). The reason is, there's always the possibility that another drive can fail during degraded mode. You even have a risk of a second drive failure during a rebuild. Both usually occur if you have multiple drives from the same batch, and there are problem/s consistent throughout the entire batch. It gets ugly.
The ARC-1680ix12, and ARC-1231ML are a little bit out of my price range. The money saved from not needing a third party cable could go towards the backup battery instead.
The 1680 series isn't inexpensive by most people's definition, but you get a lot for the $$$. The ARC-1231ML is also a good card, but one I don't typically mention, as it will NOT BOOT in EFI. You can run OS X via driver support, but that's it. It's a good card for Windows or Linux (boots these) though (even in a MP).
RAID cards don't handle recovery the same way as an OS does, so there's differences in the recovery timings set in the drive's firmware. Check the card makers site for a Hardware Compatibility List (not all offer one, but Areca does), to see what drives will function with the card (it's not going to cover every drive made). But it will save you headaches if you get a drive on the list when starting from scratch. Drives matter, even the firmware revision on them. SAS cards btw (ARC-1212 = SAS card), are picky with drives, so don't assume a drive will work. Check the list first before buying drives (you will even want to keep a spare on hand). If possible, don't buy all the drives from the same place, as you don't want sequential serial numbers (drives from the same batch). If there's problems with a batch, all drives are affected.
Test out the drives first (SMART and bad sectors), and check the firmware revisions (just in case you have to flash them to a version that will run with the card; more likely if the drives just released).
And again, I'd recommend getting an 8 port card. Say the ARC-1222. You can always upgrade the card later, but it's not convenient to do so IMO.
I have an Arc-1210 in my MP4,1 and after some initial disappointments (can't do bootcamp) I have come to like the card.
You had a bit of a tough time with it to begin with, when attempting BC on it.
Glad you got it sorted.
I believe the 1212 is made for external storage, not internal.
It's an internal 4 port SAS model. The ARC-1222 is the 8 port version of the family, and they're similar to the 1680 series.
One important feature on the MP4,1 is the elimination of harnesses to the internal HDD SATA heads. They sit directly on the logic board now. It means you have to find a way to use the traces in the PCB to connect between your RAID card and the internal ports. That may be a PITA. For me it did not matter because I use the card to run a bunch of SSDs from the 2nd optical bay and I simply connect by SATA cables. When you have zeroed in on a card have nanofrog advise you how you may be able to connect. Its a bit over my head.
The ICH10R (4 port grouping) has to be dissabled as well to make it work. It seems that's a firmware setting that can't be accessed by the user or 3rd party cards. So the adapter from Maxupgrades would have to be used in the '09's in order to use 3rd party cards.
Cabling isn't that bad. Pics really help in understanding what the differences between MiniSAS and MultiLane are, as is the internal vs. external versions are. Then pay attention to the specifications to both the card, existing cables (if any), and enclosures used.
Actually when nanofrog made his recommendation he did stress that RAID-5 was unsafe with the write hold issue and the best way was to avoid it. If I really wanted NVRAM, I would need a form of NVRAM to store it.
NVRAM = cache + battery in it's simplest form.
The cache is usually ECC, but may not be in some entry level cards, so check the specs (not sure if any still do this). But the addition of Partition Tables in firmware is an extension per se. It can allow for the recovery of an array that is otherwise lost if that feature is missing. It's not absolute, but it can help. So not all cards are equal, even if they employ an NVRAM solution.
Batteries are typically an option, but advisable to have. So is a decent UPS.
Also most of the RAM uses in the RAID controllers these days utilize ECC, so writing erroneous parity code is less likely.
Most have this. At least the cards mentioned. It helps, but it's only part of the equation.
Sorry if you feel no one is giving you credit, but the problem is you're not explaining your position. Yes we now know of the write hole issue, but why is that important? A lot of folks will want to know why. A RAID card is not cheap by any means and I'm sure everyone would appreciate a lot of feedback.
I've been hoping he'd give the details in the posts. RAID 5 does have the danger of the write hole, and should never be under-estimated. That's why I really don't like Fake RAID cards with support to create a level 5 array. They don't have an NVRAM solution of any kind, and don't bother to explain the risk to users. Not on the site, or in the manuals. On those cards,
sidewinder is absolutely correct.
NVRAM helps extensively, but nothing's absolute. It even gets into the details of whether or not a battery is used, and the type of UPS. In the end, each system is different, and the solution is a compromise. Just like any other form of RAID.
Overall, you test the system out first (computer, card, drives, and even cabling). It will let you determine if the specifics are stable. If it has a write hole problem, it will surface if you test properly. Typically in power outage scenarios (i.e. pull the plug off the UPS, and the system during a write). If you use files larger than the cache, you get to see what a corrupted file does, as this is the major limit of an NVRAM solution. In those cases, you get to re-perform the work. Woo hoo.
This card is for internal use. There are no external ports except for a Ethernet jack that you can hook up to your network, so you can configure the card on a web browser. To use a on a Mac Pro Early 2009 model, you need the sledges that can be found on MaxUpgrades.com
Actually, it's meant for remote management (and connecting to a Network Time Protocol server for accurate time).
The Web Util that pops up in your browser will work if you never connect the Ethernet port, as it's built into the firmware.
