Could you please provide both write and read results?btw tim, all my tests were run using x-raid2
NETGEAR kicks loose Six-Bay SMB NASes
- Thread starter corndog
- Start date
Regardless, I would focus on the realistic RAID levels that folks use. RAID 0 is a train wreck waiting to happen and pure striping with no parity is not a good measure of the RAID engine. If anything, I would measure X-RAID2 which is what 80% of our users will end up using and RAID 6, which will be used by the rest. With larger drives causing potential UREs (unrecoverable read error), RAID 6 is becoming more and more important, in light of possible error even during simple disk replacement. There's a healthy thread on the ReadyNAS forum here about this.
Anyway, I understand your intention to provide apples-to-apples comparison with other NAS out there with your enormous performance database, but there needs to be a balance between that and what users will ultimately use the Pro for. It really makes no sense IMO in trying to compare some of the 1- or 2-bay NAS with the Pro, and especially using a client box that really cannot drive the Pro and one that users can no longer purchase.
using iometer? because the nfs-benchmark already has both results listed.
for iometer i would have to run the test again which should be done in a couple of hours. will post the results then.
with iometer the filesize i used was 6gb (my nas has 3gb ram). with bonnie++ i used a 9000megaqbyte testfile. bonnie++ is the only unixbased diskbenchmark im aware of so i used that when testing via nfs. if anyone has a better linux-diskbench let me know
Iozone is available for linux. For more information about iozone settings look here ... http://linux.die.net/man/1/iozone
00Roush
See http://forums.smallnetbuilder.com/showpost.php?p=4702&postcount=34thx, now i basically only need tims command line parameters and i can test the speed from both windows and linux and get results that can be compared to the charts
W
wpns
Guest
Speed
I get about 3X the performance of a ReadyNAS NV with a little Shuttle KPC box I built, so it can't be hard to exceed their old ones...
I get about 3X the performance of a ReadyNAS NV with a little Shuttle KPC box I built, so it can't be hard to exceed their old ones...
W
wpns
Guest
Functionality test?
With the box in RAID-1 (mirror) mode, Infrant told me I could pull a drive, stick another drive in, and keep theone I pulled as a fully functional offsite backup.
It's never worked. Even the latest (4.1.4) firmware either doesn't flash the lights when the drive is removed, or hangs when putting another drive in, and Infrant/Netgear keep insisting it works, but never actually make it work.
If you get a chance, can you try this with the new one? Hopefully they rewrote the firmware from the ground up (which it's needed badly for years now), but I suspect it's a port...
I've certainly bought my last one (three on the shelf as I write this), but it'd be nice to know if they finally solved the problem...
With the box in RAID-1 (mirror) mode, Infrant told me I could pull a drive, stick another drive in, and keep theone I pulled as a fully functional offsite backup.
It's never worked. Even the latest (4.1.4) firmware either doesn't flash the lights when the drive is removed, or hangs when putting another drive in, and Infrant/Netgear keep insisting it works, but never actually make it work.
If you get a chance, can you try this with the new one? Hopefully they rewrote the firmware from the ground up (which it's needed badly for years now), but I suspect it's a port...
I've certainly bought my last one (three on the shelf as I write this), but it'd be nice to know if they finally solved the problem...
I have some questions about the snapshot functionality:
How flexible is the schedule for the snapshots? How much space do they tend to take up? And how much space could one conceivably allocate for snapshots? What happens when the allocated snapshot space runs out? Does it bleed over to regular storage or not? Purging of oldest data automatic or just warnings and stop?
The reason that I'm asking this is that I currently use an enterprise level solution at work for this kind of thing, but would love to keep snapshots of data going back for long periods of time; say once a week for a couple of months, then once a month for a year or more. Our data changes around a lot, but we're still < 600GB of data for the company, and that's without me chasing down the abusers on a regular basis.
I'd have no problems trying to get the funding at work to get one of these and set it up in an alternate building for long-term snapshotting (say all files copy over to it once a week or something like that) if the snapshotting would work as I think it would.
I'm not hugely worried about the time it takes to retrieve a snapshot unless it's horrifically slow (e.g. 10 minutes for a < 10MB file), just that the snapshot process is reasonably efficient on drivespace and such.
-Biggly
How flexible is the schedule for the snapshots? How much space do they tend to take up? And how much space could one conceivably allocate for snapshots? What happens when the allocated snapshot space runs out? Does it bleed over to regular storage or not? Purging of oldest data automatic or just warnings and stop?
The reason that I'm asking this is that I currently use an enterprise level solution at work for this kind of thing, but would love to keep snapshots of data going back for long periods of time; say once a week for a couple of months, then once a month for a year or more. Our data changes around a lot, but we're still < 600GB of data for the company, and that's without me chasing down the abusers on a regular basis.
I'd have no problems trying to get the funding at work to get one of these and set it up in an alternate building for long-term snapshotting (say all files copy over to it once a week or something like that) if the snapshotting would work as I think it would.
I'm not hugely worried about the time it takes to retrieve a snapshot unless it's horrifically slow (e.g. 10 minutes for a < 10MB file), just that the snapshot process is reasonably efficient on drivespace and such.
-Biggly
A snapshot practically takes no space unless you start making changes to your volume. Then it "backs up" the original data in the snapshot reserved space -- sort of opposite of what you would typically think. It's pretty space efficient in that this "copy-on-write" only copies the blocks -- not the files. When you have filled the snapshot space, your snapshot becomes invalidated. You do need to make sure that you've allocated enough snapshot space to ensure you do not run out.
We currently support one outstanding snapshot -- this might change in the future. Nevertheless, you may want to consider what others are doing -- having a 2nd ReadyNAS offsite that you schedule your snapshot backups to. Having a bunch of snapshots is nice, but that doesn't prevent data loss from an equipment disaster.
wpns - using RAID-1 hot-pull as a method of backup is insane. I've posted previously about this. There are a few reasons why you should never ever do this. First - RAID is designed to be a "first level" method of data protection. That means it is supposed to protect the live data on your system. You still need to do backups, but RAID is supposed to protect against most problems, so you only need to go to your backups in the case of a real catastrophe. RAID protects the live running data - there is no guarantee about disks that are pulled. By design, pulled disks are crap. RAID makes no effort to ensure that the data on those disks is meaningful in any way.
Second, when you pull a disk, there are multiple levels of commit that you have just broken. What guarantee do you have that the NAS wasn't right in the middle of saving a file right when you pulled the disk? Just because you made sure all your clients were not writing anything doesn't mean the NAS wasn't clearing a cache or something - you simply do not know. When you do not know, you should EXPECT corruption.
Third, your file system is not cleanly dismounted. Think about it. Say for instance you want to keep a safe copy of your Windows data somewhere, and you are planning to pull the disk that is in your Windows machine. Wouldn't you make sure you did a clean shutdown and the drive is dismounted properly before pulling out the disk? You wouldn't just pull the power plug and kill the machine, then pull the disk right? That's exactly what you are doing to the contents of the disk when you hot-pull a RAID1 drive.
Fourth, think of what RAID1 is all about - protection against drive failure. Now, think about how hard your drives are working during a typical day. Failure is most likely when the drive is most busy. If each day you hot-pull that disk, not only have you just walked away from redundancy until your new replacement disk is all mirrored again, you've also just put your drives to work - HEAVY work - remirroring. Guaranteed your disks are working waaaay harder during the re-mirror process than at any other time during the day. So just when they are most likely to fail (when they are working the hardest) you have also intentionally removed the RAID protection at this same time too - that's just plain CRAZY!
Fifth, when you hot-pull a drive, it is powering off right in the middle of running - it has not been safetied, and it is MOVING (i.e. sliding out of your NAS). No matter how carefully and how smoothly you pull that drive, compared to the fine tolerances of space between the heads and the platter, you are giving it a massively rocky ride. Are you sure you aren't causing the heads to crash onto the platter while you are doing this? Where have you ever read that it is safe for the disk to hot-pull a good drive? Hot-pull is meant for drives that are already dead. Plain and simple.
Bottom line - when RAID was designed, it was never intended as a method of hot-pull-backup. For the reasons outlined above, it should never be done. Think about it this way: Backups are important. They protect your precious data. Backup solutions (such as backup software or synchronizing tools like rsync) are designed specifically for doing exactly this job - backing up your important stuff. If your data is really so important, why would you rely on a method that was never intended to even work, to protect it, when there are other, much better methods that were actually designed to do this job properly? For examply, why would you try to put new tires on your car while driving down the road, when there are perfectly good garages you could pull into, and they can do it right?
I regularly see people on many different forums complaining about trying to use RAID1-hot-pull as a backup method and they always seem to have problems - makes me shake my head in disbelief every time. They mustn't really care about their data, or they would use a proper backup method.
Tim,
For your ReadyNAS Pro retest yesterday with the new 4.2.3 firmware (and on the new testbed), do I understand this timeline correctly:
1. You had tested the Pro on old testbed
2. You had also done some testing of the Pro on new testbed
3. You now retested the Pro on new testbed, and with latest 4.2.3 firmware
For this latest retest, you included the Pro results from the old testbed (1) for comparison. But would it also be helpful to include the first Pro results from the new testbed (2) as well, to gauge the performance increase solely from the new firmware?
In any case, it's good to see the top-notch performance here. I'm still using a couple of older ReadyNASes with the Infrant CPU, but want to upgrade to something faster (if only the ReadyNAS has network camera recording capability...).
I can't wait to compare new testbed results of the ReadyNAS NVX (Atom?), as well as the QNAP TS-439/639 (Atom) and TS-809 (Core 2 Duo). Would it be possible to re-test the QNAP TS-509 as well, seeing as it's been one of the previous performance leaders?
For your ReadyNAS Pro retest yesterday with the new 4.2.3 firmware (and on the new testbed), do I understand this timeline correctly:
1. You had tested the Pro on old testbed
2. You had also done some testing of the Pro on new testbed
3. You now retested the Pro on new testbed, and with latest 4.2.3 firmware
For this latest retest, you included the Pro results from the old testbed (1) for comparison. But would it also be helpful to include the first Pro results from the new testbed (2) as well, to gauge the performance increase solely from the new firmware?
In any case, it's good to see the top-notch performance here. I'm still using a couple of older ReadyNASes with the Infrant CPU, but want to upgrade to something faster (if only the ReadyNAS has network camera recording capability...).
I can't wait to compare new testbed results of the ReadyNAS NVX (Atom?), as well as the QNAP TS-439/639 (Atom) and TS-809 (Core 2 Duo). Would it be possible to re-test the QNAP TS-509 as well, seeing as it's been one of the previous performance leaders?
I thought of leaving the first retest data in the database. But decided that one retest was enough.
You can get an indication of what Atom will do by reading Intel Atom vs. VIA C7: Which Makes a Faster, Cheaper NAS?. Basically, I expect Atom-based NASes to deliver in the 30 - 40 MB/s range.
QNAP is going to provide a TS-509 Pro from retest when they can free one up. They tell me that demand is high.
You can get an indication of what Atom will do by reading Intel Atom vs. VIA C7: Which Makes a Faster, Cheaper NAS?. Basically, I expect Atom-based NASes to deliver in the 30 - 40 MB/s range.
QNAP is going to provide a TS-509 Pro from retest when they can free one up. They tell me that demand is high.
Dennis Wood
Senior Member
We've got the six drive READYNAS hooked up with dual LAN connections on the same network as the QNAP TS509, also connected dual LAN.
I'll have complete test numbers up to compare with all my TS509 tests, but the READYNAS is very impressive on large file writes to/from 2 different workstations. Both workstations use Vista SP1 with one using a 3 drive RAID 0 and the other, a 4 drive RAID 0 array.
On a 12GB file read from the NAS, the Netgear unit starts at 120MB/s until workstation cache is exhausted and settles in at about 106MB/s. On the the same 12GB written to the NAS the unit settles in at about 95MB/s. We haven't done comprehensive load testing yet to see where the numbers fall but initial tests are at about 130MB/s aggregate sustained read to two different workstations over the dual LAN connection...that's about 65MB/s each simultaneously. This is the best performance we've seen yet by far from any combination of RAID 0/5 between workstations, NAS units etc. The sustained write speeds are very impressive, particularly if you look back at all my testing so far. The only time I've gotten close to those numbers on writes has been when writing from a RAID 0 (4 drives) workstation, to another RAID 0 (4 drives) workstation. To compare, a workstation eSATA connection to a single WD 1TB Green drive tops out at about 60MB/s read/write which means the READYNAS unit is the first NAS I've tested that exceeds (by almost 80%) a local eSATA drive's performance on both reads and writes. Consider that the NAS does this with fault tolerance while the eSATA drive has none and the picture looks better yet.
The corresponding QNAP TS509 (latest firmware, RAID5) numbers on the same 12GB file set read/written from the TS509 using the same workstations are ~ 80MB/s read, and ~ 54 MB/s write.
On an interesting side note, one of the Vista SP1 workstations in testing was recently restored from a different workstation's backup image (same hardware) and began behaving strangely as during file transfers over the LAN, CPU (C2D at 2.5GHz) was pegged at 100% and network speeds throttled to under 60MB/s. There was nothing in the process log to suggest why this was happening, and even a bare bones service/process configuration fixed nothing. Rather than struggling, I just rebuilt it with a fresh OS, installed the same software, AV etc., and suddenly all the performance issues disappeared. It makes me wonder how many folks out there are seeing similar issues with Vista.... If you're seeing 100% CPU during file transfers, there's a good chance something is amiss.
I'll have complete test numbers up to compare with all my TS509 tests, but the READYNAS is very impressive on large file writes to/from 2 different workstations. Both workstations use Vista SP1 with one using a 3 drive RAID 0 and the other, a 4 drive RAID 0 array.
On a 12GB file read from the NAS, the Netgear unit starts at 120MB/s until workstation cache is exhausted and settles in at about 106MB/s. On the the same 12GB written to the NAS the unit settles in at about 95MB/s. We haven't done comprehensive load testing yet to see where the numbers fall but initial tests are at about 130MB/s aggregate sustained read to two different workstations over the dual LAN connection...that's about 65MB/s each simultaneously. This is the best performance we've seen yet by far from any combination of RAID 0/5 between workstations, NAS units etc. The sustained write speeds are very impressive, particularly if you look back at all my testing so far. The only time I've gotten close to those numbers on writes has been when writing from a RAID 0 (4 drives) workstation, to another RAID 0 (4 drives) workstation. To compare, a workstation eSATA connection to a single WD 1TB Green drive tops out at about 60MB/s read/write which means the READYNAS unit is the first NAS I've tested that exceeds (by almost 80%) a local eSATA drive's performance on both reads and writes. Consider that the NAS does this with fault tolerance while the eSATA drive has none and the picture looks better yet.
The corresponding QNAP TS509 (latest firmware, RAID5) numbers on the same 12GB file set read/written from the TS509 using the same workstations are ~ 80MB/s read, and ~ 54 MB/s write.
On an interesting side note, one of the Vista SP1 workstations in testing was recently restored from a different workstation's backup image (same hardware) and began behaving strangely as during file transfers over the LAN, CPU (C2D at 2.5GHz) was pegged at 100% and network speeds throttled to under 60MB/s. There was nothing in the process log to suggest why this was happening, and even a bare bones service/process configuration fixed nothing. Rather than struggling, I just rebuilt it with a fresh OS, installed the same software, AV etc., and suddenly all the performance issues disappeared. It makes me wonder how many folks out there are seeing similar issues with Vista.... If you're seeing 100% CPU during file transfers, there's a good chance something is amiss.
Last edited:
Dennis Wood
Senior Member
Zero issues with the GS108T and TS509 using LACP other than the somewhat clunky web interface on the switch. Looking at port stats on the little switch it's worked as advertised in 802.3ad. I have not tested with the READYNAS PRO, but I'd be about fairly confident they'd be OK together being that both are NETGEAR products
Similar threads
Sign Up For SNBForums Daily Digest
Get an update of what's new every day delivered to your mailbox. Sign up here!