Thunderbolt Storage (RAID) for mini server

[Zjef]

I've been looking for a decent Thunderbolt based solution for my Mac mini server. I prefer a rack mountable solution. Organised from least to the most expensive, I've got the following candidates found:

1. 4 bay 1U Thunderbolt Enc w/RAID - PC-PitStop
2. Thunderbolt | USB3.0 | eSATA 1U rackmount 5x Drives HW Raid with LCD and DirectAir
3. OWC Mercury Rack Pro + La Cie Thunderbolt adapter
4. Netstor Rackmount 16-Bay

My questions are:

1. When using server grade HDD's are any of the above decent solutions? It is hard to find (independent) information on the net.
2. Anybody have experience (good or bad) with the vendors? Other than the OWC, I don't know any of them.
3. Are there alternatives I've not find?

Remark
Since it the above is to host shares managed with the OS X server, it is not an option to use a NAS solution.

 

[jasesname]

Thunderbolt RAID

I've been looking for a decent Thunderbolt based solution for my Mac mini server. I prefer a rack mountable solution. Organised from least to the most expensive, I've got the following candidates found:

1. 4 bay 1U Thunderbolt Enc w/RAID - PC-PitStop
2. Thunderbolt | USB3.0 | eSATA 1U rackmount 5x Drives HW Raid with LCD and DirectAir
3. OWC Mercury Rack Pro + La Cie Thunderbolt adapter
4. Netstor Rackmount 16-Bay

My questions are:

1. When using server grade HDD's are any of the above decent solutions? It is hard to find (independent) information on the net.
2. Anybody have experience (good or bad) with the vendors? Other than the OWC, I don't know any of them.
3. Are there alternatives I've not find?

Remark
Since it the above is to host shares managed with the OS X server, it is not an option to use a NAS solution.

If you PM me I can give you some in depth information regarding these products and vendors

 

[calaverasgrande]

any reason you are avoiding Qnap, Synology and Drobo?

 

[freejazz-man]

Because NAS isn't DAS???

I have 4 different Promise thunderbolt RAIDs. Two r6s and two r4s. They are great!!!

If you want to deal with more hands-on stuff and less 'just works', go with the mercury or other vendors. Otherwise, PROMISE is king!

 

[Silencio]

That DAT Optic unit looks intriguing. I don't have a lot of personal experience with that brand, except for an older 4-bay Firewire unit that was a little painful to configure with outdated, Java-based Oxford applets and a semi-functional LCD display, but otherwise works just fine.

If I were buying a Thunderbolt RAID today, I'd get the 8 bay Areca. It's a tower unit, so might not suit your needs exactly.

http://www.areca.com.tw/products/thunderbolt.htm
http://barefeats.com/hard167.html

 

[FireWire2]

Thunderbolt 2 RAID for MAC/Windows

As today i would look at this:

http://www.datoptic.com/ec/jbod-raid-data-storage-solutions/thunderbolt.html

12x tray-less for range 1400's dollar, and higher capacity compare to Areca and Promise

 

[ssmed]

If you want to deal with more hands-on stuff and less 'just works', go with the mercury or other vendors. Otherwise, PROMISE is king!

Promise don't use server grade HDs in the R4 and R6. Over 2-3 years we have had 3 HDs fail and I have just upgraded one R6 with six new drives as a result. Using RAID 5 however we have not lost any data and the rebuild is easy. The speed is great and you even feel that speed over the network via the mini-server.

You can buy the new R4 II without disks which means you could put more expensive disks with a longer mean time to failure.

 

[monokakata]

It looks to me as though the Promise Pegasus2 R4 diskless is only available through Apple (in the US anyway).

Am I just not looking hard enough? I have nothing against buying direct from Apple, but this surprises me. Plenty of places have the ones with disks, but not even B&H has the diskless one.

Pegasus offers a list of wholesale vendors, including Avnet -- I liked seeing Avnet there because back in the day when they were Hamilton-Avnet I bought my first PDP-11 from them. I had a business then, but now I'm just an ordinary consumer.

The Pegasus site has an amusing typo:

"Multiply the Mac Pro’s ash storage capacity by daisy-chaining multiple Pegasus2 systems to achieve massive amounts of additional storage space."

So even Pegasus is making nMP jokes? Who knew?

 

[FireWire2]

Promise don't use server grade HDs in the R4 and R6. Over 2-3 years we have had 3 HDs fail and I have just upgraded one R6 with six new drives as a result. Using RAID 5 however we have not lost any data and the rebuild is easy. The speed is great and you even feel that speed over the network via the mini-server.

You can buy the new R4 II without disks which means you could put more expensive disks with a longer mean time to failure.

This is my personal experience with RAID in 16 yrs.
You can make desktop RAID5/RAID6 reliable as it should be.
Everyone knows HDD will failed eventually and during service life cycle HDD will develop bad sector(s)
Most of RAID's owner never bother to check RAID volume until it's too late.. and some not even utilize the check volume function of the enterprise RAID, until corrupted sector (un-readable) showed up and the RAID start misbehaves.

I have used a 8TB RAID5 populated with 5x 2TB desktop HDD for 5 five years, ONE drive gave out in the 4th year, but no service interrupted

You do not need expensive HDD to have a reliable storage.

Here is what I did.
Use: http://www.amazon.com/USB3-0-eSATA-tray-less-hardware-enclosure/dp/B002QYECOI/
5x 2TB SATA Hitachi drive

Every 3~4 month i force it to rebuilt, by open HDD's door for 10~15 sec.
After the door close, the RAID start to rebuild, it takes about 6~7hrs
It's already into the 5th yr of service!.
Why do i do that? Simple.
HDD will have bad sector(s), whether we like it or not and if the BAD sector(s) is not mapped out properly then you will have data corruption.

Therefore the force rebuild process will exam every single sector in the entire RAID volume and bad sectors will be mapped, in turn insure the data integrity

Matter of fact, the force rebuild has a similar functionality of check volume in enterprise RAID. Mapping bad sectors for not to use

So you can can use low cost drives to create a reliable RAID. Just force RAID rebuild from time to time, then your RAID volume will last very long

One thing you may aware REBUILD speed are diff from manufactures/raid engine

Mine for 8TB for 6-7hrs, which is not too bad

 

[jljue]

Because NAS isn't DAS???

I agree with this. I currently use a Synology DS1512+ with iSCSI as an "attached" drive, and while it is faster than the FW800 array that I was using, it still isn't as fast as an eSATA array using a Lacie Thunderbolt-eSATA adapter or a Thunderbolt array. Since I already use a Thunderbolt Ethernet adapter to use my NAS on its own network for file transfers, I could gain a little more speed using Link Aggregation with both Ethernet connections. My main reason for using a Synology NAS in this configuration was to still simulate an attached "DAS" or extenal drive for some applications but relocate the drives away from my Mac Mini server in my office to another location in the house to reduce noise and heat in the office and not worry about code issues with cables in the wall, like I can with my Ethernet drops.

If a person is already looking into Thunderbolt, I'm sure it is safe to assume that the drives will be with the server and will need to be speedy, which will eliminate a NAS, unless there are some code issues with wiring in the walls for drives and server in separate rooms, as it is in my case.

 

[disconap]

I'm sure this was solved months ago, but for the record I'm using a FirmTek Thundertek with a compatible PCIe eSata card, and a 12 drive eSata bay (2x5 drive on port multipliers, 2xstraight connect). Everything, including hot-swapping, is working just fine; I do dislike their drive icons but I can't be bothered to change them since the server is headless.

Pretty happy with the set up overall, I have three different software raids running smoothly (other than some permissions issues, but those only started when I installed Server)...

 

[unplugme71]

This is my personal experience with RAID in 16 yrs.
You can make desktop RAID5/RAID6 reliable as it should be.
Everyone knows HDD will failed eventually and during service life cycle HDD will develop bad sector(s)
Most of RAID's owner never bother to check RAID volume until it's too late.. and some not even utilize the check volume function of the enterprise RAID, until corrupted sector (un-readable) showed up and the RAID start misbehaves.

I have used a 8TB RAID5 populated with 5x 2TB desktop HDD for 5 five years, ONE drive gave out in the 4th year, but no service interrupted

You do not need expensive HDD to have a reliable storage.

Here is what I did.
Use: http://www.amazon.com/USB3-0-eSATA-tray-less-hardware-enclosure/dp/B002QYECOI/
5x 2TB SATA Hitachi drive

Every 3~4 month i force it to rebuilt, by open HDD's door for 10~15 sec.
After the door close, the RAID start to rebuild, it takes about 6~7hrs
It's already into the 5th yr of service!.
Why do i do that? Simple.
HDD will have bad sector(s), whether we like it or not and if the BAD sector(s) is not mapped out properly then you will have data corruption.

Therefore the force rebuild process will exam every single sector in the entire RAID volume and bad sectors will be mapped, in turn insure the data integrity

Matter of fact, the force rebuild has a similar functionality of check volume in enterprise RAID. Mapping bad sectors for not to use

So you can can use low cost drives to create a reliable RAID. Just force RAID rebuild from time to time, then your RAID volume will last very long

One thing you may aware REBUILD speed are diff from manufactures/raid engine

Mine for 8TB for 6-7hrs, which is not too bad

Its funny because a lot of the big names in online backups and other huge companies with datacenter use consumer class WD Blacks. They are cheaper and very reliable. And when you need to buy thousands of drives, even if you have to replace 1% of them each year, its still cheaper in the long run.

 

[burne]

Its funny because a lot of the big names in online backups and other huge companies with datacenter use consumer class WD Blacks. They are cheaper and very reliable. And when you need to buy thousands of drives, even if you have to replace 1% of them each year, its still cheaper in the long run.

I have several petabytes of storage in my care. An annual failure rate of 1% would be reasons enough for a party. The best drives I know of (seagate barracuda 750G in the 1.5-platter variant) is around 3.5%. Most drives average about 7%, some peak at 10% (VelociRaptor) and anything above 10% gets the boot. Like Samsung Spinpoint with a annual failure rate of 24%. That was in the 1Tbyte-days, but our experience was so bad we're not likely to try that one again any time soon.

 

[brokendownjoe]

Highpoint (Netstor) NA333TB w/mac mini, ATTO NT12

I have a 2012 macmini server sitting in a Sonnet Rackmac enclosure, connected to a Highpoint Netstor NA333TB w/Highpoint 2740 RAID controller and (8) 2Tb WD Black SATA drives. Tests for read/write are really impressive, and right on specification for the Netstor.

The only (giant) issue I've had is with the 10Gbe NIC installed inside the Netstor. The ATTO NT12 PCIe 10Gbe card will not connect on system restart (soft restart). And if you do a restart you have to go into /Library and remove files, and jump to the 2nd Thunderbolt port, power off, and then both ports work again. Just don't ever restart: only shutdown & complete poweroff works (which is really tiresome). This was true in Mavericks 10.9.4, and is true in current Yosemite server (as of yesterday).

If I ever figure out the 10Gbe this will be an excellent fileshare platform when connected to a 10Gbs switch. Unfortunately, neither NetStor, Highpoint, or ATTO has been able to help. Maybe I'll figure it out this week.

 

[FireWire2]

Update:

DATOptic Offers 15% off on their Thunderbolt 2 and Thunderbolt 3 raid/jbod - remember the code: SAV15
Do not know when it's end - most likely the end of this month