NETGEAR is aware of a growing number of phone and online scams. To learn how to stay safe click here.
Forum Discussion
Performance halved at a stroke (ReadyNAS Duo, 4.1.8) ...
In a busy few weeks, my little ReadyNAS Duo's performance seems to have halved at a stroke. The problem seems to be a consequence of upgrading to RAIDiator 4.1.8 (release) and performing a full factory reset but ... just in case ... here's a full rundown of its recent life ...
* After swapping my ageing PC for a shiny new iMac (really shiny, by the way), I noticed a very modest (but welcome) *increase* in performance from the ReadyNAS.
* Running Windows 7 (directly through Boot Camp, fully Windows Update'd) and RAIDiator 4.1.8 beta (for Lion compatibility) the little Duo could saturate my Gigabit Ethernet to 25-30% (transferring data at around 250-300Mb/s, then). Everything was running great ...
* Then a mini-disaster ... after three years use the ReadyNAS power supply gave up. The unit shut down without warning and wouldn't power up again.
* Some excellent online support put me back on track. "Some time later" and with a new power supply connected (an identical Netgear part) I was up and running again.
* The unit started first time and all was well ... everything was accessible and usable ... but restarting the unit once more (from front-view) the ReadyNAS unexpectedly froze on boot.
* Blue light flashing at 1Hz, fan at medium-volume, no activity or drive lights and invisible to RAIDar, I gave it four/five hours ...
* Not responding to physical (power) button presses, I had to pull the plug. Fortunately, the unit picked itself up and booted without issue. Scouring downloaded log's I could see *no trace whatsoever* of the failed boot.
* To eliminate any possibility of data corruption or similar badness (and with a new release-version of RAIDiator 4.1.8 now available) I updated the unit firmware, performed full factory reset and reconfigured the unit to match its old settings (all from front-view). Perhaps this was a beta glitch?
* Now running solidly and without issue, I copied data back to the unit using Windows Explorer and a local backup. It should have close to zero fragmentation, then.
Doing this, it was clear performance had halved. The unit now saturates just 10-15% of the network ... transferring at around 100-150Mb/s only (no RAIDar or front-view pages open).
This doesn't look like a hardware problem ... the drives look fine (a matched Seagate ST3500630AS pair, their SMART reports look good and their temps are normal), the network looks fine (private LAN, both Mac (on-board Broadcom NetXtreme) and NAS are wired to the same Netgear WNDR3700 (gigabit router, latest firmware), no concerning errors ... a few TCP retransmits which are not unusual) and there are no errors or warnings in the logs. But for a new power adapter, nothing has changed hardware-wise.
The old and new front-view settings match near-exactly (same network settings, same shares, same users and access rights, one or two *fewer* services running than before, no bittorrent/photo's/remote/vault, full data journaling, oplocks enabled, fast CIFS writes enabled, jumbo frames disabled), the unit isn't resynching or indexing media, there are no scheduled backup jobs (internal or time machine) and ... to be absolutely sure ... I've also confirmed performance several times over the last week to eliminate any other invisible, transient activity ... no luck :?
Performance seems to have halved at a stroke with no clear reason.
It looks very much like a RAIDiator 4.1.8 issue but I'm happy to help troubleshoot. Any chance a Jedi (or other expert) can help?
Thanks,
Justin
* After swapping my ageing PC for a shiny new iMac (really shiny, by the way), I noticed a very modest (but welcome) *increase* in performance from the ReadyNAS.
* Running Windows 7 (directly through Boot Camp, fully Windows Update'd) and RAIDiator 4.1.8 beta (for Lion compatibility) the little Duo could saturate my Gigabit Ethernet to 25-30% (transferring data at around 250-300Mb/s, then). Everything was running great ...
* Then a mini-disaster ... after three years use the ReadyNAS power supply gave up. The unit shut down without warning and wouldn't power up again.
* Some excellent online support put me back on track. "Some time later" and with a new power supply connected (an identical Netgear part) I was up and running again.
* The unit started first time and all was well ... everything was accessible and usable ... but restarting the unit once more (from front-view) the ReadyNAS unexpectedly froze on boot.
* Blue light flashing at 1Hz, fan at medium-volume, no activity or drive lights and invisible to RAIDar, I gave it four/five hours ...
* Not responding to physical (power) button presses, I had to pull the plug. Fortunately, the unit picked itself up and booted without issue. Scouring downloaded log's I could see *no trace whatsoever* of the failed boot.
* To eliminate any possibility of data corruption or similar badness (and with a new release-version of RAIDiator 4.1.8 now available) I updated the unit firmware, performed full factory reset and reconfigured the unit to match its old settings (all from front-view). Perhaps this was a beta glitch?
* Now running solidly and without issue, I copied data back to the unit using Windows Explorer and a local backup. It should have close to zero fragmentation, then.
Doing this, it was clear performance had halved. The unit now saturates just 10-15% of the network ... transferring at around 100-150Mb/s only (no RAIDar or front-view pages open).
This doesn't look like a hardware problem ... the drives look fine (a matched Seagate ST3500630AS pair, their SMART reports look good and their temps are normal), the network looks fine (private LAN, both Mac (on-board Broadcom NetXtreme) and NAS are wired to the same Netgear WNDR3700 (gigabit router, latest firmware), no concerning errors ... a few TCP retransmits which are not unusual) and there are no errors or warnings in the logs. But for a new power adapter, nothing has changed hardware-wise.
The old and new front-view settings match near-exactly (same network settings, same shares, same users and access rights, one or two *fewer* services running than before, no bittorrent/photo's/remote/vault, full data journaling, oplocks enabled, fast CIFS writes enabled, jumbo frames disabled), the unit isn't resynching or indexing media, there are no scheduled backup jobs (internal or time machine) and ... to be absolutely sure ... I've also confirmed performance several times over the last week to eliminate any other invisible, transient activity ... no luck :?
Performance seems to have halved at a stroke with no clear reason.
It looks very much like a RAIDiator 4.1.8 issue but I'm happy to help troubleshoot. Any chance a Jedi (or other expert) can help?
Thanks,
Justin
87 Replies
Replies have been turned off for this discussion
- If the Duo and the NV+ are the only models affected I think it's safe to assume that there is a performance issue with the Sparc processor. Hopefully the people at Netgear are aware of the problem.
- So, I was wondering why most/all of us here are running Mac's?
We all see the 4.1.8 performance problem consistently and clearly, across the whole range of protocols (CIFS, AFS, NFS) and with the widest range of clients we could expect (Macs, Windows PC's, Linux boxes included) but what we don't see is a simple, dedicated PC network (home or office) and a ReadyNAS ... no doubt the most common configuration out there. Something must set us apart, no?
In a tiny Eureka moment I figured this can't be the clients themselves and must be something unique to our ReadyNAS configuration ... something unique to folks supporting Macs ... and as I don't run Time Machine the AFP/Bonjour services must be at fault.
After a little firmware changing, reconfiguration and benchmarking, I suspect this may well be the case.
Here how the (good) 4.1.7 firmware performs (*1) ...
- Under 4.1.7 with AFP and Bonjour disabled, I measure read speeds of 19.9 MB/s
- Under 4.1.7 with AFP and Bonjour enabled, I measure identical read speeds of 19.9 MB/s
That's our baseline. Here's how our (broken) 4.1.8 firmware compares ...
- Under 4.1.8 with AFP and Bonjour disabled, I measure read speeds of 18.45 MB/s. A small drop but ...
- Under 4.1.8 with AFP and Bonjour enabled, read speeds are reduced to just 8.2 MB/s. A massive drop. (*2).
Furthermore, disabling the services and re-testing restores performance to the 18.45 MB/s range. (*3)
I'll be the first to say this isn't definitive testing (if someone could repeat this independently, I'd be really grateful) but it does seem to suggest the AFP and/or Bonjour services in 4.1.8 are are causing the problem. Next, I'll be checking the old 4.1.8 Beta to see if it suffers similarly as ... if it's OK, I'll be switching over.
Thanks,
Justin
*1 All these tests were run under Windows 7 using CIFS, a fixed-size file and a stopwatch. I did disable the ReadyNAS' media/streaming services to ensure they don't skew anything by reindexing content after firmware changes but ... other than that ... the config here should match my original post. Given the test methodology, all these measurements are approximate but should be fine for the job.
*2 Running these tests a few times, I did notice the results here could be inconsistent. On at least one occasion, I enabled the AFP and Bonjour services to find read-speeds comparable to the preceding step. Rebooting the ReadyNAS returned performance to its sub-standard 10 MB/s level.
*3 I also tested write speeds which were more consistent. Approximately 13 MB/s for 4.1.7 and 10.8 MB/s for 4.1.8, these don't seem to be affected significantly by AFP/Bonjour. Justin S wrote: So, I was wondering why most/all of us here are running Mac's?
We all see the 4.1.8 performance problem consistently and clearly, across the whole range of protocols (CIFS, AFS, NFS) and with the widest range of clients we could expect (Macs, Windows PC's, Linux boxes included) but what we don't see is a simple, dedicated PC network (home or office) and a ReadyNAS ... no doubt the most common configuration out there. Something must set us apart, no?
Having complaints from a fellow user (running 30 macs from 1 NV+) on 4.1.8 about speed, I stumbled upon this thread.
I run an NV myself with 4.1.8 and need it for Lion. Only services running are FTP, AFP and SRync, though only AFP is actively used. Checked performance options are full data journalling disabled, optimized for os x. I'm on a DGS_1005D gigabit jumbo frame router. The NV has 4x Seagate ST21000542AS drives. Copy of a single dmg file of 1.2 GB (average over 3 measurements each)
Bonjour on
jumbo on read 32.4 MB/s write 22 MB/s
jumbo off read 17.1 MB/s, write 10.9 MB/s
Bonjour off
jumbo on read 21.8 MB/s write 18.2 MB/s
jumbo off read 21.8 MB/s write 13.3 MB/s
Processor load was about 0.4 (as per MenuMeters) across all measurements. I think this throws a spanner in the works of your conclusion?ard wrote: Bonjour on
jumbo on read 32.4 MB/s write 22 MB/s
jumbo off read 17.1 MB/s, write 10.9 MB/s
Bonjour off
jumbo on read 21.8 MB/s write 18.2 MB/s
jumbo off read 21.8 MB/s write 13.3 MB/s
Processor load was about 0.4 (as per MenuMeters) across all measurements. I think this throws a spanner in the works of your conclusion?
Hi Ard,
Thanks for testing and sharing stats ... it feels like pooling our results might be the only way we can actually trace the issue. I can't say I have anything as solid as a conclusion but I do have a hunch which could use confirming or disproving with some independent tests so this is good stuff.
Your results make very interesting reading too as they seem to confirm Bonjour is impacting performance ... a seemingly positive impact with Jumbo Frames enabled and a negative one if (like me) you have them disabled. No spanner in the works, then ... the drop in performance when Bounjour is active vs inactive (Jumbo Frames disabled) is (while less significant than I've seen) consistent with my own testing.
Jumbo frame should increase throughput and your 'Bonjour On' performance results reflect this quite nicely. With Bonjour off, however, you seem to be see better performance with Jumbo frames disabled than enabled which is very strange. As a service discovery/name resolution protocol (which isn't responsible for file transfer activities itself) I wouldn't expect to see Bonjour have any impact on performance at all.
Thanks,
Justin- Quick update guys ...
Justin S wrote: Here how the (good) 4.1.7 firmware performs (*1) ...
- Under 4.1.7 with AFP and Bonjour disabled, I measure read speeds of 19.9 MB/s
- Under 4.1.7 with AFP and Bonjour enabled, I measure identical read speeds of 19.9 MB/s
That's our baseline. Here's how our (broken) 4.1.8 firmware compares ...
- Under 4.1.8 with AFP and Bonjour disabled, I measure read speeds of 18.45 MB/s. A small drop but ...
- Under 4.1.8 with AFP and Bonjour enabled, read speeds are reduced to just 8.2 MB/s. A massive drop. (*2).Justin S wrote:
Next, I'll be checking the old 4.1.8 Beta to see if it suffers similarly as ... if it's OK, I'll be switching over.
I've just checked performance running 4.1.8 Beta and this behaves exactly like 4.1.7 ... good news ...
- Under 4.1.8-T9 with AFP and Bonjour disabled, I measure read speeds of 19.9 MB/s
- Under 4.1.8-T9 with AFP and Bonjour enabled, I measure read speeds of 19.9 MB/s
Write speeds are identical to 4.1.7 too at around 13MB/s. No performance drop like that seen with 4.1.8 Release.
Looks like I'll be running with the beta software for a while but that's far from ideal ... a temporary solution at best. With this (limited) testing it seems like performance problems were introduced between the 4.1.8 Beta and its release, affecting devices with AFP/Bonjour enabled.
As before, I'd be grateful if anyone could confirm this themselves (or contrast it with different behaviour) independently.
Thanks,
Justin
PS. Thanks Dfiler .... your own switch back to 4.1.8-T9 gave me good confidence this would be OK - Oh dear :(
"Upgraded" to 4.1.8 about 2 days ago, what a mistake. Transfer speeds for files to and from the device plummeted from 20/25MBps to 8/11 MBps over a 1Gbps connection :shock:
Nothing else at all has changed. I use a Windows 7 x64 desktop and both devices are configured for 1000MB Auto Negotiation. Usual tests including running with just a crossover cable between the devices did not resolve.
I tested by grabbing a 4GB file and trying to copy it to the device, and from the device (it reckoned about 7 hours to copy it)
Re-applied RAIDiator 4.1.7 and I'm back with my 20/25MBps speeds :roll:
Think I'll stay away from this upgrade until it's fixed!! - Hi Muttzcuttz,
Glad to hear you were able to rollback to 4.1.7 to restore performance. So far we've seen mostly of Macs with performance problems leaving me to wonder whether the issue relates to functionality like AFP, Bonjour or Time Machine.
Running Windows 7, you may be able to shed a little more light on the problem. Do you have AFP, Bonjour or Time Machine enabled?
Thanks,
Justin - Hi,
here are my test results (in hours):
ReadyNAS 1100 Firmware 4.1.17:
rsync backup from Linux Server to 1100 takes 3,5 hours
ReadyNAS 1100 Firmware 4.1.18:
rsync backup from Linux Server to 1100 takes 5,5 hours
--------------------------------------------------------------------------------------------------------------
ReadyNAS NVX Firmware 4.2.17:
rsync synchronization from 1100 (Firmware 4.1.17) to NVX (Firmware 4.2.17) takes 3,5 hours
ReadyNAS NVX Firmware 4.2.19:
rsync synchronization from 1100 (Firmware 4.1.17) to NVX (Firmware 4.2.19) takes 5,5 hours - More data here: NV+, 1 GB RAM, 4xWD1000FYPS-01ZKB0 1TB drives, 4.1.8 software.
Jumbo frame off:
Bonjour off read 22.2 MB/s write 9.2 MB/s
Bonjour on read 21.7 MB/s write 9.5 MB/s
Bonjour doesn't seem to matter much. But wait a second, there's more!
Jumbo frame on:
Bonjour off read 15.6 MB/s write 1.7 MB/s
Bonjour on read 15.6 MB/s write 1.5 MB/s
Ouch!
To verify that it was not the mac, I copied a 1 GB file across the network to another mac. Read 58.8 MB/s, write 35.7 MB/s. - I'm having the same problem with my non-plus NV but in a predominantly Windows 7/Windows Server environment. I also have a few macs and some Linux VMs. I have 3 Seagate 1.5TB 7200.11 drives in flex RAID5 with another as a hot spare in User security mode. I have a Netgear WNDR4000 router.
In my limited testing the fastest speeds I can get are ~19 MB/s reads and ~12.5 MB/s writes. No changes in Frontview help, including all network options, sharing services, discovery services, time machine, add-ons, etc. Turning on Jumbo frames on the NV and my PC (I normally don't use them) cut the above speeds in half and changing from a DHCP reserved IP to a static IP lowers each by ~2.5 MB/s(?). The NV was factory reset after I installed the final 4.1.8 over 4.1.8-T9. I also reverted to the T9 beta with no change. Haven't tried 4.1.7 yet.
At first I thought it was an auto-negotiation problem but now it seems more like something's broke in the Raidiator driver.
EDIT: mixup on r/w speeds
NETGEAR Academy
Boost your skills with the Netgear Academy - Get trained, certified and stay ahead with the latest Netgear technology!
Join Us!
