happy sailor Loc: Ontario, Canada

Good Morning all,

I have decided to get a Synology 5 drive NAS system and at present have 4 3 tb drives to go in it. Not sure if I should go with a full mirror (raid 1) an array (raid 5) or a combination mirroring with striping (raid 10).

What are the thoughts of those that have NAS systems.

Thanks for sharing

nicksr1125 Loc: Mesa, AZ

RAID 10 can be a little trickier to setup but gives you a major advantage. By combining striping & mirroring, you have more storage area and a 2nd backup. It's been 10 years since I setup an array so, things may have gotten easier.

Dngallagher Loc: Wilmington De.

Definately if you have the number of drives required, go for RAID 10.... good performance, drive aggregation and mirroring....

happy sailor Loc: Ontario, Canada

Thank you both for your replies, the new synology systems and software make it quite easy to set up any of the raid configurations they support you just have to choose what is best for you.

oldnomad

RAID 10, best protection and best performance. You will not have as much available space, but that is the trade off.

From a system admin with 30+ years of experience.

JoeBiker Loc: homebase: Houston, TX

Hmmm, I don't know specifically which Synology system you are getting, and how you plan to connect it.

But, assuming a single GE connection (No link aggregation), I think that your performance will probably be limited by the GE link, so I'm not sure you will see increased performance by using RAID 10.

oldnomad

Raid 10 even over 1G should have a performance advantage over RAID 5 in writes. RAID 10 should also have a performance advantage over RAID 1 or 5 if the array is degraded due to a drive failure.

One could always setup the array under each RAID level and run a throughput utility such as iometer to see the real world difference. IOMETER can be configured to represent anticipated usage (for example 70% read 30% write) to make the test more valid. You would also see the real space available for each RAID level. Most vendors list their array storage capacity as RAW space which few people actually use.

JoeBiker Loc: homebase: Houston, TX

After thinking about the disk and network performance, I think there are too many unknowns.

(assuming it's not too difficult to reconfigure the system) When you get this system, before you load your personal data onto it, it might be worth your time to configure the system in each of the modes(at least RAID5 and RAID 10), and run Crystal Disk Mark (or something similar) against it. That will tell you if RAID 10 is really buying you anything.

HDDs start storing data on the outer ring, which is the fastest, so it won't be a worst case test, but it will give you the best quick answer to your question.

JoeBiker Loc: homebase: Houston, TX

Hi OldNomad, it looks like our posts crossed in the mail :)

oldnomad

Actually although the outer ring is moving faster, the data has larger gaps between data to handle the latency (in the olden days, we had to set the interleave factor manually to compensate for this latency to get optimal performance), so data is read and written at the same speed regardless of where on the drive the ring is.

oldnomad

Hi, JoeBiker. Yes they did, what the heck are we doing up this late talking computer storage in a photography forum?

JoeBiker Loc: homebase: Houston, TX

I guess we just can't help ourselves (and just trying to help out fellow member, but I'm not sure happy sailor is even reading any more).

BTW, Manually setting the interleave factor is not really before my time, but I wasn't personally involved with it then.

However, since then, the HDD manufacturers have set up zones to take advantage of the larger circumference of the outer tracks. You can look at the screen shot at http://www.hdtune.com , and run hdtune on your own drives to see the impact. There used to be more prominent stair steps as it switches from zone to zone, but perhaps they now have a huge number of zones.

Castor Loc: Lancashire

Hi, I'm not too well up on raid systems but I do use a synology DS1513+ with 5 x 2tb drives. I just us the inbuilt synology hybrid raid which gives 1 drive fault tolerance.
i believe that raid is not a valid backup system and that it is far safer to backup on a variety of drives. I know I go over the top with this and have my photos etc in about six places including off site.
The Nas is great for accessing everything and I even have films and music on it which I can watch on iPad and could access from the Internet if my broadband speed was half decent.

happy sailor Loc: Ontario, Canada

[quote=JoeBiker]I guess we just can't help ourselves (and just trying to help out fellow member, but I'm not sure happy sailor is even reading any more).

Yes Joe, Happy Sailor is still reading and thankful for all the replies. I plan on buying the DS1513+ and then adding the expansion unit later which will bring it up to 10 drive capability. More than the speed issues I was concerned with the talk about how recovery from a raid 5 taxes the drives and could create the circumstance for a second drive to fail while rebuilding when one has already failed.

I used to manage a Tandem Mainframe system where everything was mirrored and we brought down one half of the mirrored drive and replaced it with another set of platters each day, and then a weekly backup done the same way each Saturday. System performance was hurt during the backup period while the mirror drives were re-imaged but it was done during a relatively quiet period for the system. That system had to stay up 24/7 at all times so it was a bit of a challenge.

I was leaning to a raid 5 but as the space requirements grows I thought raid 10 would be easier to add to.

again thanks for your input.

maybe i should just go with the 1813+ now and that should last for a few years before I need to upgrade.

JoeBiker Loc: homebase: Houston, TX

In response to Castor & Happy Sailor's latest:

RAID doesn't eliminate the need for (preferably off-site) back-ups.

And yes, RAID reliability calculation are based on the assumption that the 1st failed drive will be replaced before a 2nd drive fails. There are several ways that assumption can be compromised.

If all of the drives in the system are the same model and the same lot, you run the risk that they will all fail close to the same time (especially if they turn out to all come from the same contaminated lot). Been there, done that, and got the scars to show for it. So, as much as the system supports it, the system should be build with drives that are as dissimilar as possible: Different Models, if possible; and definitely, different lots. The disadvantage of that is that the overall performance is determined by the slowest (or best case, 2nd slowest) drive in the system, but that is the price paid for reliability.

Also, S.M.A.R.T. is so dumbed down that it can't be counted on to identify failing drives. Somewhere in the system there has to be intelligence to proactively identify failing drives. Not sure if Synology provides anything to assist with that.

And, yes, reconstruction causes all the remaining drives to be read, and can uncover a 2nd failing drive; a side effect that is not usually factored into the reliability calculations....