Ethernet speed

[DeepPurple23]

I recently purchased a Netgear SC101 to use in my home set up, two Laptops and the set up was easy and everything seems to be working fine, but... When I open file manager and select the drive, it takes forever to open the directors, not to mention how long it took to copy the data from my old USB external drive on to the new drive.

So here's my question, is it me (my settings) or is a Ethernet connect running at 100 Mbps significatly slower than an USB 2.0 external drive?

I would have thought the networked solution would have be far better.

Let me know what you think.

DeepPurple :?

 

[Madwand]

Regular ethernet is 100 Mb/s, which is significantly slower than high speed USB 2. Modern ethernet is gigabit, which is potentially faster than high speed USB 2, but in practice this varies. Most consumer NAS devices, even when they support gigabit ports in the back, don't get much faster than 100 Mb/s ethernet (aka "fakegigabit").

Decent modern desktop computers with gigabit should be able to exceed USB 2 speeds in practice, at least for those (parts of the) drives where the drives themselves are capable of exceeding USB 2.

 

[Iceblue]

So here's my question, ...is a Ethernet connect running at 100 Mbps significatly slower than an USB 2.0 external drive?

Leaving out protocol overhead, USB 2.0 runs at 480Mbps and your ethernet runs at 100Mbps.

Direct drives are the fastest (ATA, either S or P variety)
Firewire and USB 2.0 drives are next fastest
NAS drives are next fastest.

You can bump the NAS higher on the scale by upgrading your LAN to 1000 Mbps, but check the Tom's chart - I don't think your NAS box would benefit that much.

(I left out SCSI, since that is a niche market these days.)

 

[Madwand]

You can bump the NAS higher on the scale by upgrading your LAN to 1000 Mbps, but check the Tom's chart - I don't think your NAS box would benefit that much.

It's a cheap NAS with only 10/100 ports, so the benefit would be about nil, give or take 0.

 

[Iceblue]

You can bump the NAS higher on the scale by upgrading your LAN to 1000 Mbps, but check the Tom's chart - I don't think your NAS box would benefit that much.

It's a cheap NAS with only 10/100 ports, so the benefit would be about nil, give or take 0.I figured he could check his own specs... :roll:

 

[jjw]

As stated above, most NAS are pretty much junk.

Nobody should expect gigabit rates out of a piece of hardware just because it has a port in the back. But the performance of most NAS lucky if its just over 100mbit. (hence fakabit speeds) Even synthetic benchmarks do not achieve gigabit line speeds.

To me it appears that there is some 'other' bottleneck in these systems. It certainly isn't the Network. If the HDD drive can do 70 Mbyte/sec, why can't the NAS do at least 1/2 of that. But maybe I am out of line considering that Tom's DIY NAS is not that great either: Link

 

[Iceblue]

I love they way devices are hypercriticized against some ideal without considering the market, who the customers are, and the price.

The fastest NAS in Tom's chart achieves in the vacinity of 350Mbps on a 1000Mbps network. This was a DIY NAS, certainly not all customers are going to be looking for a project; they just want shared storage.

The highest performing manufactured NAS achieved about 280Mbps, but it cost ~$700 without any hard drives.

The highest performing single-drive NAS appliance achieved about 110Mbps but costs only about $180 including the drive.

Sure, the device this OP bought is slower than these; it is also a lot cheaper, and is a nice concept for its place in the market.

And all of these are pretty amazing compared with the price of the REAL high performance NAS servers for the enterprise market.

 

[Madwand]

In comparison with cheap desktops, which could do around 30-40 MB/s file transfers, the bulk of consumer NAS devices at this time deserve criticism for their performance, especially those cases where the gigabit ports at the back have net performance little different from 100 Mb/s.

Regarding the cheap single-drive 100 Mb/s boxes -- they're significantly out-performed by simple USB/firewire boxes as the OP noted as the basis for this thread, so again the performance is sub-par.

Tom's DIY NAS box was poorly designed, placing both the NIC and storage controller on the PCI bus. It didn't perform well either. It's sustained performance was much less than 350 Mb/s or 43.75 MB/s. Cheaper simpler boxes can perform better IMO, and DIY servers with good design better still.

As to how people spend their money -- it's up to them. If people want to spend hundreds for under-performing cute boxes and stay uninformed about the subject area and performance potential, it's a reasonable business opportunity -- one which the manufacturers of NAS boxes are appropriately capitalizing on.

However, just as OP saw, it's not nice to find out about poor performance, after spending the money, so IMO it's better to be clear about the performance issues, which exist, rather than trying to sweep them under the cover of falsehoods such as gigabit port = uber performance and that you must spend large amounts for higher-end performance.

Again, performance is not everyone's concern, and will be deliberately overlooked in many cases in the quest for cute convenient cheap boxes, which is just fine. But when people have performance concerns, as does the OP here, it's better to be clear about what's what.

 

[Iceblue]

However, just as OP saw, it's not nice to find out about poor performance, after spending the money, so IMO it's better to be clear about the performance issues, which exist, rather than trying to sweep them under the cover of falsehoods such as gigabit port = uber performance and that you must spend large amounts for higher-end performance.

Again, performance is not everyone's concern, and will be deliberately overlooked in many cases in the quest for cute convenient cheap boxes, which is just fine. But when people have performance concerns, as does the OP here, it's better to be clear about what's what.

Don't buy them; build your own. Whatever floats your boat, but it is worth pointing out two things:

The OP was not "deceived" about the performance of the NAS he bought, he just did not pay attention. He actually thought a 100Mbps ethernet network would be faster than USB 2.0. This has nothing to do with drive performance. Build the best designed DIY NAS you want, and it can't push bits over a 100Mbps ethernet faster than 100Mbps, which is gonna be 1/4 the speed of a cheapo ATA drive installed in a $19.95 DIY USB box.

The Netgear he bought showed read performance in Tom's charts of about 50Mbps, which is respectable for that class of device, and nowhere near what is easily and cheaply achieved with USB 2.0.

Second, there is more to a NAS than merely an attached drive, otherwise we would all be just doing file sharing with our PCs. You dismiss them by calling them "cute" when the actual attraction of these devices is that they are small, quiet, and don't require all the rigamarole associated with a full Windows PC.

I agree with you on one point: for the most part, the 1,000 Mbps ethernet ports on the consumer devices are pure marketing fluff. One of the few exceptions to that I know of is the Buffalo LinkStation Pro, which at least shows performance above 100Mbps over the gigabit interface.

In case you are wondering, no, I do not own a NAS appliance myself. But I have set them up for others.

 

[jjw]

I'm familiar with the DIY NAS on tom's. I was a little shocked at the performance the first time I saw where it landed.

I am a current owner of a Infrant Readynas 600. I am let down by the performance of the box. The box only achieves ~ 70% of the network bandwidth for 100Mbit and 54Mbit (wireless) networks. For Gigabit performance is about 2/3 of what Tom's reports. My system has been online for about 1 year and likely has some significant fragmentation.

What the box does do is keep me from having to set up a fileserver, keep it patched, etc. It has always worked slowly. What I didn't expect is that it has been nearly impossible to back up the device as a result, It took 4 DAYS to backup the device to a USB drive attached directly to the NAS 8O I'm pretty sure that even USB1 would have been better.

Back to the DIY NAS; I have a feeling the real limitation of this system is the fact that the disk controller and the NIC are hanging off the PCI bus. A more modern motherboard would likely perform 'much' better. Kinda explains why the max throughput is the same for the DIY with and without a RAID card.

I'm pretty sure you could build a pretty stellar box for the $600 I paid for the Infrant (without drives)

 

[DeepPurple23]

First I want to thank everyone for their replies it has been helpful in my process. Second I want to clear up the fact that I was neither "deceived" nor failed "to pay attention". I can tell which is higher 480 or 100, but the reality of "peak" vs "sustained" transfer rate and difference in handling small files and large files is significant and that is where my logic and what I got failed to meet my needs. The SC101 is fine for what it is, but I do find it odd that you could put 1.5TB in the thing and then live with a sustained transfer rate of about 4.5 Mbps. My DSL line transfers faster than that. In reality to the common person, where are the other 95 Mbps of bandwidth? All of which has become a mute point as I have replaced the unit with eSATA device and now transfer speeds are great, I have to switch cables when I want it on the network, but that is not too much of a hassle.

Lastly, I will agree with most stated here about the "stated" speeds. You see it in all forms of consumer electronics, it's the whole peak versus sustained issue, however, think about this. If your network is a 1 Gbps and your drive can only transfer at 70 - 100 Mbps, why is the network still the bottleneck?

Thanks for all of you help.

DeepPurple23

 

[Madwand]

Congratulations on finding another solution so that this is no longer a problem for you.

I would have found 4.5 MB/s painful, let alone 4.5 Mb/s. 4.5 Mb/s is much worse than I would have guessed (and from the rest of the thread, you could see that I'm not especially optimistic).

To ask the obvious question: Are you sure you're not reading MB/s somewhere as Mb/s?

b = bit
B = Byte = 8 bits

4.5 MB/s = 36 Mb/s. Still not great, but at least conceivable for a slow 100 Mb/s device.

 

[megamojo]

As Madwand pointed out, you're getting confused about speeds.

b= bit
B = Byte = 8 bits

So, 100 Mb/s = 12.5 MB/s

When we said harddrives go at ~70 MB/s they meant MB, not Mb. 70 MB/s is 560 Mb/s.

Oh, and to keep things in perspective, the claimed 480 number of USB 2.0 is in Mb/s, not MB/s.

 

[PashaM]

i'll ask my question here, i have 2 pcs connected through a regular cheap Netgear router 4 ports. During file transfers my speeds never go above 1-2MBps which i consider very slow for a 100Mbps network. Whats wrong with my setup? or is it normal? my PCs are at 4 feet apart.

 

[jjw]

1-2 MByte/sec is slow... Are the 100 Mbit lights lit on the router (most routers have 2 tone leds, i.e. yellow = 10 Mbit, green = 100 Mbit)

Do you mean 1-2 MByte/sec downloads from the internet or PC to PC. Internet speeds will be tied to your ISP; you get what you pay for.

It could be a bad cable, or maybe if the cables are 'really short' there can be problems. Can't remember the source, but cables under 7' can have problems.

 

[megamojo]

Again, please be specific about speeds. Are you talking about megabytes (MB) or megabits (Mb)? There is a factor of 8 involved, so it makes a big difference.

 

[Iceblue]

Again, please be specific about speeds. Are you talking about megabytes (MB) or megabits (Mb)? There is a factor of 8 involved, so it makes a big difference.

True, but if he has a 100Mbps ethernet LAN, 1-2 mega-whatevers is pretty slow. 1-2MBps is 8-16Mbps - still pretty pathetic for a 100Mbps LAN.

 

[megamojo]

I agree, but at least 1-2 MB would be in the realm of possibility. 1/10 of your potential isn't nearly as bad as 1/100.

 

[Madwand]

i'll ask my question here, i have 2 pcs connected through a regular cheap Netgear router 4 ports. During file transfers my speeds never go above 1-2MBps which i consider very slow for a 100Mbps network. Whats wrong with my setup? or is it normal? my PCs are at 4 feet apart.

The way to find out is to do some speed analysis -- check the networking speed in isolation, check the drive speed in isolation, put them together, try different file transfer methods and check again. Keep in mind that the final speed will not be the minimum of the drive speed and the network speed, it will be less. How much less depends on a lot of complex factors.

iperf can be used to check network speed.

E.g.

server: iperf -s
client: iperf -c server -l 64k -t 21 -i 3 -r

(I use iperf version 1.7.)

Disk read speed can be checked with something like HDTach.

Beyond that it gets more complex; I'd suggest looking at as some tech support notes / reviews and following their tests in detail. Here's one. It's not perfect, but at least it's detailed and a basis for comparison:

http://www.infrant.com/forum/viewtopic.php?t=265

Also, you have to use large files for transfer speed tests -- when using small files, the system won't be as efficient, and can also give very erratic performance results due to file caching issues. I'd suggest files at least as big as your RAM, but understand that this can be slow/frustrating under 100 Mb/s networking esp. when it's under-performing, so probably use something much smaller -- in the range of 100 MB.

Jumping ahead to potential problems, sometimes drives degrade to PIO mode instead of DMA, and this slows down the entire system and of course file transfers.

 

[PashaM]

Again, please be specific about speeds. Are you talking about megabytes (MB) or megabits (Mb)? There is a factor of 8 involved, so it makes a big difference.

Mhen i say Mbps i mean Mbps and when i say MBps i do mean MBps... i have a pretty good networking knowledge base...

I do use large files for transfer tests , and i am talking about PC-to-PC speeds/transfers.
What i suspect is the posible EM interferances, i have a lot of wires, extention cords, resepticle-splitters and a power-bar on the floor area where my CAT5s are laying. PCs are, like i mentioned about 5 feet apart. Could it be cheap NICs or even with cheap NICs i should be able to get at least a third of 100Mbps. What are the speeds some of you guys having, in similar conditions (pc-to-pc)?

 

[Madwand]

What are the speeds some of you guys having, in similar conditions (pc-to-pc)?

8-10 MB/s is fairly common with standard 100 Mb/s -- this data rate tends to be much less stressful overall to the computer than gigabit -- all components, CPU, bus, hard drive, and network should be able to get close to optimal ethernet data rates. There are always some inefficiencies, which is why I say 8-10 MB/s, not 12 MB/s.

It's much easier to approach saturating a 100 Mb/s network than it is a gigabit network. Most of us do not hit 80-100 MB/s with gigabit -- most drives can't do it, some PCI NICs can't do it, etc.. But some of us can do 80-100 MB/s with consumer gear (with tuning and RAID arrays), which hints at how much easier it should be for us to saturate 100 Mb/s.

 

[PashaM]

8-10? thats not bad at all, it makes it 80-90% capacity the rest of it is just protocol overhead. I wish i had that

 

[megamojo]

On a 100 Mbps network I can usually come pretty darn close to saturating it, and on my gigabit network I usually get around 40 MB/s, and I've seen as high as 60.

 

[RomanoX5]

I use Mobilicity and my download on torrent soft goes... Yes 7kB.
Contemplating suicide or genocide with that. Any help around?

 

[The_Prophecy]

This topic has been closed by The_Prophecy