Disappointing performance HomePlug AV 500 and TS-419P II

[Doug in Calgary]

I recently installed a TS-419P II. While copying an immense media library from a PC hard drive to the NAS, the best download obtained was between 6 and 7 MB/s (48 to 56 Mb/S). Details of configuration and install:
-QNAP TS-419P II with two RAID1 volumes. The volume to which I am copying consists of two Seagate Barracuda XT 2 TB 7200RPM 64MB Cache drives in RAID1
-NAS connected to GigE router via two Ethernet cables (the second cable is a fall back).
-GigE router connected to TrendNet HomePlug 500 Mb/S power over Ethernet plug.
-PC connected to another TrendNet HomePlug 500 Mb/S power over Ethernet plug.
-can measure real world bandwidth between two power over Ethernet plugs at between 120 and 220 Mb/S
-CPU utilization on NAS never exceeds 22% during copy.

Is this the best performance I can expect? How can I determine the bottleneck(s)?

[schumaku]

Use a Gigabit Ethernet connection instead. %00 Mbit/s sounds good for home plug, just like 450 Mbit/s for WLAN. A Home Plug AV 500 net delivers about 87 Mbit/s when the installation is permitting, and not too much reflction loops are plugged (ie. wrapped-up extension cords); for comparision, an optimal 450 Mb/s WLAN does deliver up to about 55 Mb/s (free air, no walls, minimum reflections, ...).

Taking the 81.9 Mbit/s as a base does translate to a maximum of about 10 MByte/s when things are almost perfect. FInd a reasonable good comparison of HPAV500 devices SmallNetBuilder: HomePlug AV 500 Adapter Roundup - Admin, Performance, Closing Thoughts

Borrowed form that same compariosn (by courtesy of SmallNetBuilder) - remember Mbit/sec - and not blaiming or preferring one or the other manufacturer:

Complain with the HomePlug AV 500 makers resp. thier marketing departements: Marketing numbers of WLAN and HPAV(500) are blunt lies.

-can measure real world bandwidth between two power over Ethernet plugs at between 120 and 220 Mb/S

Numbers form the low-level information as privided by the HPAV500 chipsets I guess - or single task, single stream numbers. Just very few side traffig will massively reduce the possibilities. Welcome to the world of shared media (beeing air or a power line infrastucture!

-CPU utilization on NAS never exceeds 22% during copy.

So the NAS is not near to be the limitaitons - it's always the processor limiting the throughput possible.

Would you mind if a moderator is adding HPAV 500 to the subject?

[Doug in Calgary]

OK I did an experiment. I moved the NAS to the same GigE router as the PC. I now see downloads in the 10 -12 MB/s range. The NAS CPU utilization is around 30%. Is the bottleneck the Seagate HD?

[schumaku]

Another well-known number - one of the network elements in the line (PC / computer network interface, Ethernet switch/Router-embedded switch only Fast Ethernet (100 Mb/s) - this would explain the top at about 12 MB/s or 100 Mbit/s

I have seen you have some standard Gigabit Ethernet hardware (reouter, switch) of course:
- Check all LAN cables are fully wired with four pairs (eight strands) each.
- Check the PC LAN interface does run on effective 1000 Mb/s.
- Check if the router and dedicaed GbE switch LED do show GbE connections, ie. LED color, Link LED color, ...

[father_mande]

Hi,

Just for information my own result

network Gigabit
TS-459 Seagate in Raid5 (not empty, lot of heavy application run ... )
Devolo AV500 (firmware up to date)
PC Windows 7 64 bits (4 CPU / 8GB mem / 2 singles hitachi disk)
no specific optimisation

PC -> QNAP (transfer of 4 file 1GB each test CPL and Gigabit in 5 mns elapse, 4 transfer at the same time ( // ) to optimize the use of bandwidth.
direct link : ~65MB/s
thru CPL : ~15MB/s
Devolo cokpit said 463 Mbit/s / 475 Mbit/s so theoretical ~55MB/s ...

so yes ... the CPL cut the performance ... (even if it's reasonable when you don't want to have cable ...)

Philippe.

[Doug in Calgary]

Father Maude's results suggest that I am seeing a bottleneck somewhere other than the POE as he is seeing signficantly higher bandwidth on both.

[schumaku]

Please try to understand that the performance of a WLAN or HomePlug AV 500 (generically aka. CPL ... Cable Power Line, or PLC Power Line Communication) depends on the quality of the media, the wiring, the layout of the wiring. Some kind of installations are absolite poison for power line performance. The perfromannce does vary widely, it also depends on the number of PLC devices active on the same media - this can be the same network ID, or on a different network ID. Even worse - some older/different PLC technologies (not HP, HP AV, HP AV 500 - these are compatible) are destructive for the HP technology variant when co-existing on the same physical wires. The only truth is: A Gigabit conneciton does transport up to 125 MB/s, a Fast Ethernet connection does transport up to 12.5 MB/s. Of course, you can ignore the SmallNetBuilder HP AV 500 test, and contine to belive the marketing B.S. or the naked chipset-supplied bandwidth lies.

I now see downloads in the 10 -12 MB/s range. The NAS CPU utilization is around 30%. Is the bottleneck the Seagate HD?

As you now are testing a direct network conneciton that does top-of at 12 MB/s - what is a typical upper imit of Fast Ethernet (there is some protocol overhead) - and the NAS CPU has a lot of spare perforance left. Almost convinced, one of the link works at 100 Mbit/s (FE) only.

We're not hee ot fight about tehcnologies and what is possible - we can just show you the many things that can (and will) go wrong - beeing on "obvious" Ethernet tehcnologies, or on the lesss-obvious HP (or WLAN) technology. Over here (or say outsde of North America and US-oriented countries), there are all three phases entering a building, and also smaller houses, appartements, ... Most outlets are wired n a split-phase or single-phase method (one phase plus null), other outlets (machines, largers dishwashers, tumblers, ...) offer all thee phases plus null. This lead to the situation, that there is no direct copper connection between the phases on different outlets - there whee you plug the HP/HP AV/HP AV 500 or whatever PLC devices. So all the network carriers (each can be seen as a RF transmitter) used to transport the data just coupling RF-wise between multiple phases. Again, here it depends if htere are sufficient wires with the three phases togehter in one cable to couple the signals. The situation can be enhanced by installing phase couplers - which allow a good part of the RF energy and bandwith to pass between the different phases. This can commonly enhance the performance by 200..300%. On the opposite, you find HomePlug technologies devices with power distributors built-in. Why? Bwcause of the PLC device must be located near or almost direct plugged to the outlet. Any kind of "christmas-tree-like" extnesion cords, probably paired with spaghetti layout, rollerd up, ... and the PLC device plugged somewhere thee is aain massively lowering the performance again.

Just like WLAN ... in the same room thre are some interferences and reflection, when WLAN passes a wall and/or a floor, it depedns on the material, the structure, the amount of water in the material, ....this does lead to very different performance possibilities.

Conclude: There are no HP AV 500/HP AV/HP or WLAN systems that are performing the same. The possible results are varying widely on these technologies. Just because of Pieres (or my...) PLC network does perform much better - you can't expect your performing the same.

Back to Ethernet-basd teting for now please....

[Doug in Calgary]

The network card on the PC is rated at 1 GB/s as is the Linksys E3000 router. I'll dig around a bit to see if some setting is limiting either to 100 Mb/s. I've also got an Asus 1 GB Ethernet switch that I can try in place of the Linksys.

What is the real world max performance I could expect from the NAS? This link: http://www.xbitlabs.com/articles/networ ... html#sect0 suggest in the 100 MB/s range.

[father_mande]

Hi,

The network card on the PC is rated at 1 GB/s as is the Linksys E3000 router. I'll dig around a bit to see if some setting is limiting either to 100 Mb/s. I've also got an Asus 1 GB Ethernet switch that I can try in place of the Linksys.

What is the real world max performance I could expect from the NAS? This link: http://www.xbitlabs.com/articles/networ ... html#sect0 suggest in the 100 MB/s range.

Please don't compare benchmark (better state for all) and real life ... but in all case you must have a better result ... target 75% of a benchmark test can be a good value ... AS MAXIMUM ... not with others activities, even if you don't see it ...

Philippe.

[schumaku]

That challenging benchmark does provide a good overview on different activities - check the the picture with the graphs for the various tests.

[Doug in Calgary]

The Linksys router was set to 100 Mb/s. I changed it to GigE and can now get uploads in the 40-50 MB/s range. I also rearranged the PoE network (used different plugs) and was able to get uploads over in the 8-10 MB/s range. Thanks for all the advice.

[ovbg]

I find this very interesting as I am getting absolutely terrible Homeplug performance in my apartment. I have three D-Link Homeplug devices. One connects to my Router, one is in the basement with my NAS, and a third is in another part of my apartment with a wifi access point (as our walls are quite thick)

Essentially, I am lucky to get 2MB/s. Some days it drops to 1MB/s and if I move the wifi access point/Homeplug unit to the Hallway in the middle of the apartment, I get only 200KB/s.

I know I should not expect terribly high performance, but surely being 100-200 times lower than the "stated" amount, something is seriously going wrong. **, I would be happy to be only 5% of the stated 200MB/s, as that would be 10MB/s and fast enough for most of the tasks I want. But 1MB/s is taking it a bit too far.

* I live in a brand new apartment (in Germany), so the wiring is pretty good
* All devices are from D-Link and so should be compatible (200MB/s max)
* I have tried with just two Homeplug devices and this makes no difference (i.e. disconnecting the wifi Homeplug device) and this makes no difference.
* I have tried with the basement unit back upstairs (and again only two Homeplug devices) but it makes no difference. I thought this would as it seems the basement power point is on a different breaker switch to the rest of the apartment.
* At one stage, I had the NAS upstairs, with the Homeplug in the bedroom, and all of a sudden I reached a staggering 6-7MB/s. Fantastic, except I can't leave it there since the NAS makes so much noise and it no longer is "off-site" for backup purposes.
* I have tried switching off all sorts of devices (unplugging from the wall) but it makes no difference.
* All Homeplug adapters are plugged directly into the wall (not into multi-adapters or on extension cables).

A rather odd thing is that when I first tested it, I got 25MB/s down to the basement. That was before I bought the NAS though. Maybe other people in the apartment also have similar devices and they are all interfering with mine?

* Is there anything else I can try to improve performance. As said, I would be happy with 10MB/s?
* What are these Phase coupler solutions? Would this be of advantage here in Germany? And would I need one on each Homeplug?

One thing I can't do is run ethernet. You have no idea how much I would love to do that. But, we would need to somehow feed that through the wall which will be far too expensive for us. I don't even think that is possible in Germany on a new apartment build as they don't have cable ducts here.

[schumaku]

There are two competing technologies on the market: On one hand HomePlug (1.0) / HomePlug AV / HomePlug AV 500 (AV2) from the HomePlug Assciation (HPA) vs. the DS2 technologies from the Universal Powerline Association (UPA). While Design of Systems on Silicon (DS2) went out of business some years ago, the UPA cesed from the net, however Marvell accquired the technology is is still selling chipsets. As soon as you have both technologies on the same or a very near power network - the perfomance goes bust.

Basement to appartement: Are both power outlets connected to the same power counter? If yes, it can happen these are simply on two different phases, so a phase coupler might help, creating an RF-coupling between the thee phases. Here in Switzerland, the rooms associated fix to a specifc apparement (ie cellar, storage, or hobby rooms) must be wired to the same counter nowadays.

When in the basement a different power counter is used ... well, technically you still can install a power phase coupler, when I remember right, some coupler have four phase points (plus a common null).

New apparetement and no Ethernet or Fiber installed? Pitty... Good options to install an Ethernet cable are a) Cable TV wiring ducts, or b) Telephony wirhing ducts. Try to get a duct/cable plan from the installation company or the appartement management company.

[ovbg]

Thanks for your reply schumaku,

So, if someone in the same apartment building was using a DS2 system, that could be causing me the problem. Hmmm, I wonder if they would also be experiencing the same drop in speed? I will have to send an email out to everyone.

I am presuming the basement is on the same power counter. It's all pretty modern (the building was completed last year). But the odd thing is, is that I get the same or near low speeds if I have the Powerline adapters and NAS upstairs. The fastest I got was 7-8 MB/s with the NAS in the bedroom and around 6 MB/s using an outlet in the same room as the computer and it's Homeplug. So, I am not convinced the problem is related terribly to the basement. After all, even 6MB/s when everything is in the same room but different sockets is still slow.

Sadly, there is no way to run a cable to the basement, so if I were to run ethernet cables in the apartment, it would help in one situation, i.e. fast speed all throughout the apartment itself, but no help for the NAS in the basement which is my "off-site" solution.
I did think about having ethernet running to the basement during the build, but the building company didn't want anything to do with it. As for inside the apartment, it was an extra cost that we just couldn't afford, and I assumed such a small apartment would have no Wifi problems. I mean, it's only a few steps from one side to another, but the middle wall is thicker than I anticipated.

I know, I should have still gone with ethernet, and spent the money I didn't have at the time as it would have been cheaper in the long run. But I didn't, so I am stuck with this situation. The thing is, I just never expected my Powerline adapters to be working 200x slower than stated on the box. Who would have?

If anyone can think of anything else I can check, please let me know.

[schumaku]

The fastest I got was 7-8 MB/s with the NAS in the bedroom and around 6 MB/s using an outlet in the same room as the computer and it's Homeplug.

Some 50..70 Mbit/s are feasible with HomePlug AV (the theoretical 200 Mbit/s stuff) in one direction when the power network is fine here, with a phase coupler in place. Compare to the SNB HPAV500 roundup performance http://www.smallnetbuilder.com/lanwan/l ... l=&start=2 ... they show even lower numbers like ~40..45 Mbit/s for HPAV or ~80..85 Mbit/s for HPAV500. That could be caused by additional parts of the RF spectrum filtered in the US vs. the European specs, ie. because in the US there are some more HAM-Radio bands available and protected.

The point is that the RF does - however - poorly pass the electricy meters (power meters). Any transfer you see to the installation in the basement is almost likely just by capacitive wire-to-wire coupling. This works rather nice, as long as it's (from the pure a RF view) nice and clean environment.

Yes, it's commonly known that interoperation of HomePlug and DS2 badly s...s, because of both systems don't "tolerate" the other similar (but not same) carrier frequences used. Compare to a classic yellow or thinwire Ethernet, when the CSMA/CD mechanism fails or is suppressed: The devices dodn't detect the other technology carrier (CS), will let multiple devices access (MA) at the same or very similar time (Ethernet does add a random delay for example), nor does recognize collisions with the other technology. This will simply lead to a lot of low-level RF-activity on all system, makin the situation even worse.

Even more, already using the same technology (all HomePlug based [even if the HP 1.0 85 Mbit/s and the HPAV/HPAV500 are not interoperable, but they tolerate each other]) does create interferences - as we're using kind of a shared media. Similar when operating multiple WLAN AP on the same frequency.

In any case: It's slow like **** - even when everything is in perfect shape and HP can perform well. When you can accept that to be ways below a wired Fast Ethernet (or 1/10 of a witched Gigabit Ethernet) you can consider some optimisation. Hard to do if you are not permitted (or simply not experienced ) to work on the electrical installations, so experimenting with phase couplers might be difficuclt. Similar trouble exist when it comes to add a CAT6 (or better a hich-quality CAT6/7) cable ot the low voltage (Cable TV, or telephony) installations - assuming there are some decent metric M20 or M25 ducts installed, and not the historical thin stuff.