NETGEAR is aware of a growing number of phone and online scams. To learn how to stay safe click here.
Forum Discussion
Trouble shooting slow network speeds on NV
I’ve had slow transfer speeds on my internal network for some time. I’ve upgraded some hardware and run a variety of tests and I still don’t think I’m getting the speeds I should. I’m out of ideas and energy…
Here is my set up:
A new FIOS Actiontec MI424WR Rev. I router, gigabit Ethernet, wireless N - no Jumbo frames I believe. MTU set at originally set at 1492, now 1500. Provisioned at 15/5.
Desktop XP SP3 (static IP):
- Intel Gigabit CT Desktop NIC with latest driver shows Gigabit Full Duplex active. NIC is auto negotiating.
ReadyNAS NV (static IP),
- RAIDiator 4.1.8, stock RAM, 3 2TB drvies in X-RAID
- Drives – 3x’s Hitachi HDS 722020 2TB each, no smart events
- disabled full data journaling, and enabled fast CIFs writes and fast USB disk writes.
- CIFS, NFS and Rsync are enabled in services and for all the relevant shares.
- Frontview shows gigabit is active; NV is auto negotiating.
- MTU was at 1492
Vista Laptop (sp2) (Static IP)
- intel wireless N adapter shows connection speeds of 120-130 Mbps, with a solid signal above 90% from the kitchen about 25 feet from the lap top through two walls.
- - Intel gigabit NIC.
Other devices connected to the network are assigned dynamic addresses from DHCP – several laptops, iphones, printer, sonos network, wii, networked blu ray.
I’ve run tests with both my previous CAT5e cables and again with new CAT 6 cables. There seemed to be some modest improvement with the CAT 6 cables.
Download speeds -- DSLreports speed tests:
- 16-19Mbps up/4.2Mbps down. I'm on east coast and these speeds are for east coast
- Same speeds for both wired desktop and wireless laptop.
- West coast test is much slower and vary 4-6Mbps up/2-4Mbps down. upload speeds varies
-I’m provisioned at 15/5.
IOMeter results with 5e cables:
13MBps Read
16.2MBps Write
IOMeter results with Cat 6 cables and checking “disable journaling” in frontview:
15MBps Read
17.9MBps Write
Here are some network transfer speeds. I've run two tests. One with a single 350MB Avi file and the other with 450 MB of small photo files (280 photos) to see differences. I've tested wired and wireless connections, then again after adjusting mtu on the vista laptop to 1430 (per Botts posts) and after swaping in CAT 6 cables. The wireless was tested from two distances to router. Hopefully, this chart is readable. Columns 2-4 are for the avi file; colums 5-7 are for the phote files.
In addition, I tried a few tweaks after these results
Increase MTU on NV to 1500:
IO Meter results plummeted on the Read score
6 MB/sec Read
11MB.sec Write.
Switched MTU back to 1492:
IO Meter results stayed slow.
Switched back to 1500 and rebooted router and NV:
IO Meter:
11.6 MB/sec Read
17.4 MB/sec write.
At some point in the middle of all this, I started checking the error log. It showed TCP retransmits and some unrecovered retransmits. I started checking the error logs more frequently. This log appeared before any MTU changes (see just below). After resetting logs and checking more regularly, I just see a few TCP transmit errors and this is from the wired connection mostly.
Network Errors [Ethernet 1]
Auto-negotiation 1
Bad packets 0
Disconnect 1
False carrier 8
Idle errors 510
Link failures 0
Receive errors 0
Symbol errors 3
VLAN tags 0
TCP Retransmits 134
Unrecovered TCP Retransmits 6
Bottom line. Wired results are slower than they should be. IO meter read speeds are lower than write. wireless N speeds seem slow too. Can’t seem to figure this out.
Any suggestions?
Here is my set up:
A new FIOS Actiontec MI424WR Rev. I router, gigabit Ethernet, wireless N - no Jumbo frames I believe. MTU set at originally set at 1492, now 1500. Provisioned at 15/5.
Desktop XP SP3 (static IP):
- Intel Gigabit CT Desktop NIC with latest driver shows Gigabit Full Duplex active. NIC is auto negotiating.
ReadyNAS NV (static IP),
- RAIDiator 4.1.8, stock RAM, 3 2TB drvies in X-RAID
- Drives – 3x’s Hitachi HDS 722020 2TB each, no smart events
- disabled full data journaling, and enabled fast CIFs writes and fast USB disk writes.
- CIFS, NFS and Rsync are enabled in services and for all the relevant shares.
- Frontview shows gigabit is active; NV is auto negotiating.
- MTU was at 1492
Vista Laptop (sp2) (Static IP)
- intel wireless N adapter shows connection speeds of 120-130 Mbps, with a solid signal above 90% from the kitchen about 25 feet from the lap top through two walls.
- - Intel gigabit NIC.
Other devices connected to the network are assigned dynamic addresses from DHCP – several laptops, iphones, printer, sonos network, wii, networked blu ray.
I’ve run tests with both my previous CAT5e cables and again with new CAT 6 cables. There seemed to be some modest improvement with the CAT 6 cables.
Download speeds -- DSLreports speed tests:
- 16-19Mbps up/4.2Mbps down. I'm on east coast and these speeds are for east coast
- Same speeds for both wired desktop and wireless laptop.
- West coast test is much slower and vary 4-6Mbps up/2-4Mbps down. upload speeds varies
-I’m provisioned at 15/5.
IOMeter results with 5e cables:
13MBps Read
16.2MBps Write
IOMeter results with Cat 6 cables and checking “disable journaling” in frontview:
15MBps Read
17.9MBps Write
Here are some network transfer speeds. I've run two tests. One with a single 350MB Avi file and the other with 450 MB of small photo files (280 photos) to see differences. I've tested wired and wireless connections, then again after adjusting mtu on the vista laptop to 1430 (per Botts posts) and after swaping in CAT 6 cables. The wireless was tested from two distances to router. Hopefully, this chart is readable. Columns 2-4 are for the avi file; colums 5-7 are for the phote files.
In addition, I tried a few tweaks after these results
Increase MTU on NV to 1500:
IO Meter results plummeted on the Read score
6 MB/sec Read
11MB.sec Write.
Switched MTU back to 1492:
IO Meter results stayed slow.
Switched back to 1500 and rebooted router and NV:
IO Meter:
11.6 MB/sec Read
17.4 MB/sec write.
At some point in the middle of all this, I started checking the error log. It showed TCP retransmits and some unrecovered retransmits. I started checking the error logs more frequently. This log appeared before any MTU changes (see just below). After resetting logs and checking more regularly, I just see a few TCP transmit errors and this is from the wired connection mostly.
Network Errors [Ethernet 1]
Auto-negotiation 1
Bad packets 0
Disconnect 1
False carrier 8
Idle errors 510
Link failures 0
Receive errors 0
Symbol errors 3
VLAN tags 0
TCP Retransmits 134
Unrecovered TCP Retransmits 6
Bottom line. Wired results are slower than they should be. IO meter read speeds are lower than write. wireless N speeds seem slow too. Can’t seem to figure this out.
Any suggestions?
19 Replies
- I suggest that you try a direct connect of the ReadyNas to the XP desktop, and retest the speed. Also look at the network stats and see if those idle errors, etc also occur with direct connect.
I would also measure the desktop to laptop transfer speed over wireless. jperf is a good tool for this. WiFi link speeds are very misleading, they often don't correlate to actual throughput. While you are at it, you might also try a wired test (using the lan port on the laptop and disabling the wifi),
I am puzzled on the router model - it should be something like MI424WR REV. I Can you check? - Thanks for the quick response.
StephenB wrote: I suggest that you try a direct connect of the ReadyNas to the XP desktop, and retest the speed. Also look at the network stats and see if those idle errors, etc also occur with direct connect.
I tried that first, but I have to admit I couldn't get it to work and apparently wasn't understanding some of the posted directions. I probably need some additional guidance on how to make it work. For what it is worth, I haven't seen any other idle errors. Since last night for example, I was doing some work from my wireless laptop and these are the errors I see:
TCP Retransmits 739
Unrecovered TCP Retransmits 122I would also measure the desktop to laptop transfer speed over wireless. jperf is a good tool for this. WiFi link speeds are very misleading, they often don't correlate to actual throughput. While you are at it, you might also try a wired test (using the lan port on the laptop and disabling the wifi),
I understand the wifi link speed doesn't equate with throughput, but I included it to show that it is connected at N speeds. Anyway, I tested transfers in all directions: desktop wired to/from NV; laptop wireless to/from desktop and NV and provided that data in the chart, where it says vista laptop to/from XP desktop or NV. I just used my own math to come up with the figures.I am puzzled on the router model - it should be something like MI424WR REV. I Can you check?
Yes, I just left off it was the MI424WR REV I. All the FIOS routers are MI424WR and the diffeence is the REV.
I am suggesting something like jperf, so you can measure the raw network transfer speed independently from the NAS.mas99 wrote: I understand the wifi link speed doesn't equate with throughput, but I included it to show that it is connected at N speeds StephenB wrote:
I am suggesting something like jperf, so you can measure the raw network transfer speed independently from the NAS.mas99 wrote: I understand the wifi link speed doesn't equate with throughput, but I included it to show that it is connected at N speeds
OK.
Quick question about the direct connect. I've re-read instructions and want to make sure I don't screw up access the the NAS. My router currently assigns addresses by DHCP. In the router, I have made the addresses to the NAS and my desktop static (after they were assigned). Needed that for rsync backups, so the addresses didn't move around.
Now, for the direct connect, do I want to go into frontview, where it still says the address is assigned by DHCPeven though the router now treats it as static, and simply change it temporarily to be static. Or can I leave it as is, change the IPv4 address on the client and simply plug the client into the NAS?- Had success with direct connect of desktop to NV.
Ran IOMeter twice. The Read throughput is worse than originally, write is slightly better. There were two TCP retransmit errors and nothing else.
7 Read/19 Write
10 Read/19 Write
Ran it a 3rd time, with same results and no additional retransmit errors.
Tried drag and drop with the same files as reported initially. The speeds copying from the desktop to the NAS were the same to slightly faster as before, but the NAS to C were slower. In the course of copying both sets of files, the total number of TCP retransmits went up to 7, but no other errors - I prefer timing drop and drag over IOMeter.
It might be useful to experiment with the PC NIC settings while in direct connect. For instance, try turning off any off-loading processing, interrupt moderation, etc.
Make sure rx and tx flow control are turned on. - Over several days, I’ve run direct connects between the Desktop and the NV and the Laptop and NV. The IO results are wildly inconsistent and frustrating, particularly the Read speeds. Laptop results are better, but Read speeds also vary.
Just a reminder, the current settings for the NV are:
- disabled journaling & full data journaling, and enabled fast CIFs writes and fast USB disk writes.
- CIFS, NFS and Rsync are enabled in services and for all the relevant shares.
- Frontview shows gigabit is active; NV is auto negotiating.
- MTU 1500
DESKTOP DIRECT CONNECT. Best speeds were with the above settings.
Best R/W speeds were 15/19MB/sec, but the read speed, even with these same settings, varied from 6 to 15 MB/sec. Avg Read speed seemed to be about 13. There are a few restransmit errors, but not many. I noticed idle time errors, but they seemed to be linked to when I changed setting on the NV or the desktop NIC and the two devises lose their connection temporarily. After I reset error counts, and run a few tests, I’m left with a few retransmit errors and that is all.
If I enabled jumbo frames, the R/W speeds drop -- (8/11MB/sec), turned off offloading (4/18MB/sec), interrupt moderation (4/19MB/sec). Each time I’d reset defaults, the next read speed was terrible. If I ran it again, it seemed to improve a bit, more to the avg. Just now, after getting the better speeds, I let it sit for 3 hours and ran another IO meter test, the Read speed drops to 6 MB/sec. No additional errors reported in frontview. Read speeds are the biggest problem
LAPTOP DIRECT CONNECT. Speeds were better, but Read speed also varied greatly.
Best speed was 27/30MB/sec once out of 20+ tests. However, read speed varied from 11 to 30MB/sec, with some in the 15MB/sec and a few in the 27MB/sec range. With Jumbo frames, Read dropped to 11-15MB/sec. So, jumbo frames doesn’t work in direct connect with either desktop or laptop.
For the Desktop, why are the Read speeds so bad and for the laptop so inconsistent? Why are the Write speeds still below what botts suggests are a range from 25-40 MB/sec? I understand I don’t have jumbo frames on (as they never seem to work), but even so, it should be better.
On the laptop the peak speeds are in the range bott describes, but the Read speeds are wildly volatile and drop well below the expected range.
I’ve invested huge amount of time on my NV over 5 years, and I’ve always had slow speeds. Having upgraded the Desktop NIC card, I assumed that would improve, but there really seems to be a problem separate from the Actiontec Router. I am pretty exhausted from all this. I know my family is.
Off to the Superbowl now.... - Well the NV is a 6 year old product (it's pretty much the same as the NV+ v1 minus a few evolutionary improvements unrelated to performance). Newer ReadyNAS models such as the NV+ v2 and the Ultra 4 are much quicker (though due to the different RAID format you'd need to transfer your data e.g. across your network).
There are some performance improvements in 4.1.9 beta: http://www.readynas.com/forum/viewtopic.php?f=17&t=59222 - mdgm, Does the fact that I bought it in 2006 explain the wildly inconsistent results? I can understand how it could result in slower overall speeds. The laptop was inconsistent too, but also much faster. I'd like to optimize what I have, not start over. This is for a home setup and I don't need the fastest. Last year, with your help, I upgraded all the drives, got a fresh start and it has been working fine for most of my purposes. However, I do get stutter on some videos and the back up take forever. It was the backup speeds that got me to acquire the gigabit NIC along with the newest Actiontec. Any suggestions besides the beta? I'm interested in it, but would rather wait for it to be cleaned up.
mas99 wrote: mdgm, Does the fact that I bought it in 2006 explain the wildly inconsistent results?
No. Were you always testing the exact same thing (i.e. same file transferred same way onto NAS)? If testing very different files one would expect to get different results. What add-ons and other services are you running?mas99 wrote: I can understand how it could result in slower overall speeds.
It does explain why the speeds are a lot slower than the newer models available today.
NETGEAR Academy
Boost your skills with the Netgear Academy - Get trained, certified and stay ahead with the latest Netgear technology!
Join Us!
