Print

Please login or register.
1 Hour 1 Day 1 Week 1 Month Forever
Login with username, password and session length

[mike343]

Hello all,

I just received my 343 and have flashed to the v1.02 firmware. So far, so good.  I'm impressed by the build quality of the unit; solid, black aluminum is nice.    The fans are a little loud when they spin up, though.  A bigger chassis with a 120mm fan would've been the way to go.  The web interface could use a little more polish too; clearer descriptions of what settings actually are/do would be nice, but I'll get used to it.

Two questions, please.

Does the 343 support expansion of a RAID5 array? I.E. I currently have 3, 500GB drives in a RAID5, can I add a fourth 500GB drive and expand the array without losing data?

Does the 343 support a hot spare?  If not, will it be added in a future firmware release?

Thanks.  Nice forum, DLink!  Great to see mods/engineers/dlink employees actively participating.

[hilaireg]

Hi mike343,

Does the 343 support expansion of a RAID5 array? I.E. I currently have 3, 500GB drives in a RAID5, can I add a fourth 500GB drive and expand the array without losing data?

You can add the extra HDD at a later time; unfortunately, it will only be available as a separate volume.  You will need to recreate your RAID5 set if you wish to include a newly installed HDD.

Does the 343 support a hot spare?  If not, will it be added in a future firmware release?

The DNS-343 does not provide 'Hot Spare' or 'Hot Swap' capabilities ... I would love to have had the 'Hot Spare' capabilities, although I suspect it may be a limitation of 'Software RAID'.


HTH,

[mike343]

Thanks for the detailed reply, Hilaireg.  With the 343 being a Linux-based, one-purpose appliance, I'd have bet that the RAID was hardware-based.  That's sort of disappointment...a $400 disappointment.  But as long as it works dependably, I will be happy.

I can understand hot-expansion not being supported, but hot spare seems to be an easy thing to implement/support.

Maybe in a future firmware update?    /crosses fingers

[mig]

I would love to have had the 'Hot Spare' capabilities, although I suspect it may be a limitation of 'Software RAID'.

Mdadm (the software RAID that the DNS-343 uses, http://neil.brown.name/blog/mdadm)
does have the "Hot Spare" capability for RAID5. [See http://linux-raid.osdl.org/index.php/HowTo 
and search for "spare-disk"]

The D-Link web GUI does not allow for this configuration.

[fordem]

Thanks for the detailed reply, Hilaireg.  With the 343 being a Linux-based, one-purpose appliance, I'd have bet that the RAID was hardware-based.  That's sort of disappointment...a $400 disappointment.  But as long as it works dependably, I will be happy.

I can understand hot-expansion not being supported, but hot spare seems to be an easy thing to implement/support.

Maybe in a future firmware update?    /crosses fingers

a $400 disappointment?

Have you looked at the price of hardware RAID controllers?  Maybe you should, and make sure the ones you're looking at are true hardware RAID and not "fakeRAID"

[mike343]

Actually, I currently own two hardware RAID controllers and have owned many back to the days of 9GB SCSI Drives that cost $1K each.

I bought the 343 so I wouldn't have to run a huge, dual-quad Xeon box just to have "always on" online backup/storage capability.

[fordem]

Well - there you have it - at $1000 for a dedicated controller plus additional costs for processors, memory, enclosure, etc. etc. - you knew there was no way you were getting hardware RAID for the price you were paying for the 343  - didn't you.

[mike343]

Honestly, no, I didn't know.  $400 is a fair price for a "closed, dedicated system" with only one purpose and no software licensing (GPL) payment issues to deal with.  I honestly thought it was a hardware solution.  It can only control up to 4 drives...it's not like it's a PCI-E 8x SAS controller.

Considering my 3Ware9650 8-port SATA controller with the battery backup was $600, I actually was expecting this $400, 4-port, unit to have a hardware RAID implementation of some type.

[azz710]

Folks,

Actually, the hardware vs. software issue isn't nearly as simple as you might imagine; for all RAID implementations involve programming, all RAID implementations are done with either software or firmware, and the distinction between software and firmware is artificial, to say the least.

The industry refers to RAID implementations which consist of a reentrant control program stored in ROM (the dictionary definition of "firmware") as "hardware RAID", and implementations using what amounts to an application program running under a more generally applicable operating system as "software RAID".

In this case, as both the Linux kernel and the RAID control program reside in persistent ROM, the complete package can legitimately be referred to as firmware, just as D-Link refers to it, which would make this a hardware RAID implementation, looked at from only one angle, at least.

But, as I say, the distinction is meaningless, for it would be quite possible to create a software RAID implementation which far exceeds the performance of any so-called hardware RAID implementation available, today.

So, then, my argument is that RAID implementations should be evaluated on the multiple bases of speed, feature set and reliability and nothing more.  Disappointment has been expressed here that the DNS-343 uses a sofware RAID implementation, but it's still quite fast and, assuming the long-awaited firmware level 1.03 solves the remaining problems (and I'll admit that's a heck of an assumption), I can see no reason for disappointment in D-Link's implementation or a purely semantic distinction.

Regards,
Jeff

[hilaireg]

Hi azz710,

All very good points. 

IMHO, the biggest distinction between Hardware RAID and Software RAID is that Hardware RAID is traditionally focused on doing a very specific function - managing a redundant array of inexpensive disks.   

Hardware RAID usually consists of a disk enclosure, backplane, and RAID controller and management software stored in ROM - this is not to say that they are bug free.  However, because they are typically self-contained at the hardware level and only visible to the Operating System as Volume(s) via a driver, they are less likely to be negatively affected by other components that make up or run on the Operating System.

Software RAID is traditionally implemented as part of an Operating System and, IMHO, is subject to problems that are usually introduced by that Operating System or application(s) that run on the Operating System.

If the DNS-3XX products were exclusively used for disk management, I would agree with you that there would be little difference.  However, the DNS products provide additional features such as iTunes, UPnP, DHCP, Time Service, User/Group Management, FTP, BitTorrent, Scheduled Backup, etc.  IMHO, each one of these features introduce additional overhead to the DNS thus introducing a potential for issues.

Cheers,

[fordem]

Folks,

Actually, the hardware vs. software issue isn't nearly as simple as you might imagine; for all RAID implementations involve programming, all RAID implementations are done with either software or firmware, and the distinction between software and firmware is artificial, to say the least.

The industry refers to RAID implementations which consist of a reentrant control program stored in ROM (the dictionary definition of "firmware") as "hardware RAID", and implementations using what amounts to an application program running under a more generally applicable operating system as "software RAID".

In this case, as both the Linux kernel and the RAID control program reside in persistent ROM, the complete package can legitimately be referred to as firmware, just as D-Link refers to it, which would make this a hardware RAID implementation, looked at from only one angle, at least.

But, as I say, the distinction is meaningless, for it would be quite possible to create a software RAID implementation which far exceeds the performance of any so-called hardware RAID implementation available, today.

So, then, my argument is that RAID implementations should be evaluated on the multiple bases of speed, feature set and reliability and nothing more.  Disappointment has been expressed here that the DNS-343 uses a sofware RAID implementation, but it's still quite fast and, assuming the long-awaited firmware level 1.03 solves the remaining problems (and I'll admit that's a heck of an assumption), I can see no reason for disappointment in D-Link's implementation or a purely semantic distinction.

Regards,
Jeff

Interesting - I was under the impression that hardware RAID was defined as such by the presence of a processor dedicated to whatever RAID calculations were required and in the absence of such a processor, when the system processor handled the tasks, it was considered software RAID.

By your definition - the recent crop of so called "FakeRAID" controllers would be considered hardware RAID - a definition that I am not sure I agree with and I believe I can safely say that I am not alone in that regard.

Prior to the advent of "FakeRAID" I used a very simple method to determine how RAID was being handled - use the operating systems disk utilities to see how many physical disks the OS was aware of - if I could see all the physical drives, it was software RAID, if I couldn't then it was hardware RAID.

With "FakeRAID", firmware either in the system board BIOS ROM for on board controllers or ROM on an add-in RAID controller, along with the the controller drivers completely mask the physical disk configuration from the operating system, exactly as would happen with hardware RAID, but all RAID calculations and operations are done by the system processor, exactly as would happen with software RAID.

In one instance I took a $20 two port SATA disk controller card and flashed the BIOS and turned it into it's $65 "bigger brother" two port SATA RAID card supporting RAID0 and RAID1.

I'd also really like to see a software RAID implementation capable of "far exceeding" the performing of for example a hardware based RAID6 implementation - I don't think it can be done - not whilst the system processor is also running the OS and a database engine.

[azz710]

Dear Fordem,

I believe I didn't make myself as clear as I would have liked.  For I wasn't referring to current, specific definitions and usage of terms but, rather, the basis for what I see as an immaterial distinction.

Regards,
Jeff

Interesting - I was under the impression that hardware RAID was defined as such by the presence of a processor dedicated to whatever RAID calculations were required and in the absence of such a processor, when the system processor handled the tasks, it was considered software RAID.