Gigabit ethernet vs. ?

[KCINREBAK]

currently using a Gigabit Ethernet card. Does firewire have a faster transfer rate for MB/s?. Currently getting around 50 MB/s transfer bewteen computers did some google research just wondering how I can get something way faster thanks

 

[zigzag3143]

currently using a Gigabit Ethernet card. Does firewire have a faster transfer rate for MB/s?. Currently getting around 50 MB/s transfer bewteen computers did some google research just wondering how I can get something way faster thanks

Firewire is somewhat faster, but if you really want to go fast its fiber optic.

What do you need to go that fast for?

Ken

 

[ultraplanet]

If you are only getting 50 Mbit/s it sounds like not all of your network is gigabit. Your hub or switch and also the computer on the other end all need to be gigabit enabled. Your network will only go as fast as the slowest component.

 

[KCINREBAK]

Lots of HD movie transfers thats what I needed the speed for.

 

[Zahl]

Sustained 50MB/s is quite good. I find hard drives to be the most limiting factor in transfer speeds imho

 

[pparks1]

Well, firewire 400 runs at 400 megabits per second and firewire 800 runs at 800 megabits per second. Gigabit Ethernet runs at 1000 megabits per second....so technically gigabit ethernet is faster.

Theoretically, 100 megabits per second comes out to roughtly 12.5 megabytes per second and gigabit comes out to roughtly 125 megabytes per second. These are in a perfect world with no overhead whatsoever. For reference purposes, I usually factor in about 85% of the possible as the real world max...so 85% of 125 megabytes per second is 106 megabytes per second.

As somebody else pointed out. There aren't that many mechanical hard drives which can sustain data transfer rates of 100 megabytes per second. Most are in the 60-80 megabyte per second rate. So, if your rates across the network are 50 megabytes per second or more...that's about as good as it's going to get.

 

[ultraplanet]

pparks1 can you help me calculate this? I just can't get my head around all the conversions. Can you tell me megabits per second or is it megabytes per second?

I was able to transfer from my workstation to a network server:

Size on disk: 19.8 GB (21,283,614,720 bytes)
in... 7 minutes 42 seconds

 

[DC187]

I did some testing of gigabit network speeds of Windows Server and a Debian server running Samba 3 to several workstations running a variety of Windows OS's. The two servers were of identical hardware, here is what I found:

1 big flat file of 4.8GB – Server to Client – Windows = 75MB/s average, Samba = 130MB/s average
1 big flat file of 4.8GB – Client to Server – Windows = 40MB/s average, Samba = 85MB/s average

Directory of 20x 50MB files – Server to Client – Windows = 65MB/s average, Samba = 90MB/s average
Directory of 20x 50MB files – Client to Server – Windows = 60MB/s average, Samba = 80MB/s average

So overall I found Samba based File Servers are quicker at distributing files.
Not really an answer to the original question but people may find the results interesting none the less.

 

[pparks1]

pparks1 can you help me calculate this? I just can't get my head around all the conversions. Can you tell me megabits per second or is it megabytes per second?

Generally speaking, network speeds are rated in megabits per second. So, with ethernet, we have had 10 megabits per second, 100 megabits per second and with gigabit we have 1000 megabits per second.

Generally speaking, a megabit per second is noted as Mbit/s, or Mb/s or Mbps. A megabit per second equates to

• 1,000,000 bits per second
• 1,000 kilobits per second

To convert bits to bytes, you want to divide the # of bits by 8 to get bytes. Therefore, a megabit per second equates to

• 125,000 bytes per second

With a gigabit per second, the notation is Gbit/s, Gb/s or Gbps. A gigabit per second equates to

• 1,000,000,000 bits per second
• 1,000,000 kilobits per second
• 1,000 megabits per second

Again, to convert bits to bytes, divide the bits by 8 to get bytes. Therefore, a gigabit per second equates to

• 125,000,000 bytes per second

I was able to transfer from my workstation to a network server:

Size on disk: 19.8 GB (21,283,614,720 bytes)
in... 7 minutes 42 seconds

Doing the math here, 46,068,430 bytes per second is what you are getting. As you said before, that is nearing 50 megabytes per second.

Depending upon your hard drive speeds, that might be about the best you are going to see. You would have to benchmark your hard drives to see how fast your workstation hard drive is. If it only gets 60MB/s then you certainly won't get 100MB/s from your network The theoretical max of gigabit ethernet is 125MB/s...but about 80% is all that will ever be utilized is 100 MB/s

 

[BeechV35Pilot]

If you are only getting 50 Mbit/s it sounds like not all of your network is gigabit. Your hub or switch and also the computer on the other end all need to be gigabit enabled. Your network will only go as fast as the slowest component.

Is this true if the slowest device is at the end of the network instead of in the middle of it and I was only transferring between the gigabit devices?

In other words, if I have this configuration:

Desktop -> Switch -> NAS -> Router to Internet

and all devices are gigabit cabled with Cat5e except for the router (which is 10/100 Mb), would the router outside of my transfer scenario cause the entire network to be slower? I also experience much slower transfer rates than I thought I would see on 10/100/1000 network devices.

This area of computing knowledge has always been one of my weakest.

 

[DC187]

No, the router should not affect the transfer rates of other devices on your network.

 

[ultraplanet]

Generally speaking, network speeds are rated in megabits per second.

So if we are talking megabits and not megabytes we need to multiply 46,068,430 by 8?

Making 368,547,440 bits per second or 368 megabits per second?

 

[pparks1]

Generally speaking, network speeds are rated in megabits per second.

So if we are talking megabits and not megabytes we need to multiply 46,068,430 by 8?

Making 368,547,440 bits per second or 368 megabits per second?

Yes, you certainly can do the math, converting the bytes into bits to get to 368 megabits per second.

For most people these terms just get confusing, because network devices talk in the bits realm, while with hard drives and such people are used to talking in bytes per second.

 

[pparks1]

In other words, if I have this configuration:

Desktop -> Switch -> NAS -> Router to Internet

and all devices are gigabit cabled with Cat5e except for the router (which is 10/100 Mb), would the router outside of my transfer scenario cause the entire network to be slower? I also experience much slower transfer rates than I thought I would see on 10/100/1000 network devices.

This area of computing knowledge has always been one of my weakest.

In this case, your router shoud not impact your speeds of your local network since all of your local network devices are plugged into the switch.

With regards to speed, it's important to make sure that all adapters are configured and getting full speeds and that full duplex is in place. If you have a device that is only negotiating 100 megabit per second speeds, your speed to that device or from it will be negatively impacted.

And as we have said before, you have to take into account the speed of the hard drive in your computer or laptop. Regardless of how fast your network speeds are...if this hard drive cannot move data faster than the network...it will become the bottleneck. Let's say you have a Western Digital 1TB hard drive that has read speeds of 80 megabytes per second. So, if you need to copy from that drive across the network, you won't go any faster than 80 megabytes per second. And you will likely go a bit slower than that...as your OS and such also runs from that drive and this will negatively impact your performance as well. So, let's say you get 75% of your max hard drive speed during the transfer...well that is (80 x .75) = 60 megabytes per second. So, that is really the best that you are going to get and with a little extra overhead on the network and such...you will probably only get 80-85% efficiency...so (60 x .85) = 51 megabytes per second.

You say that you aren't getting the speeds that you thought that you would with your current 10/100/1000 configuration. What types of speeds are you seeing and what did you expect to see?

 

[BeechV35Pilot]

You say that you aren't getting the speeds that you thought that you would with your current 10/100/1000 configuration. What types of speeds are you seeing and what did you expect to see?

I am seeing about 13 MB/sec, or about 104Mb/sec. I guess I was expecting more along the lines of 30-40 MB/sec.

What utility do you recommend to test the hard drive's speed?

 

[pparks1]

I am seeing about 13 MB/sec, or about 104Mb/sec. I guess I was expecting more along the lines of 30-40 MB/sec.

What utility do you recommend to test the hard drive's speed?

Yeah, that does seem slow. That's only a slight bit faster than 100 megabits per second.

I usually use HDTune to measure my hard drive speeds.

I usually get about 10.5MB/s at home on my network, which is only 10/100 and my home server only has a 100Mbit/s card....so that is what I would expect.

for you, I would expect at least 30-40MB/s on a gigabit network with Cat5e, all ports at 1000 and full duplex.

What switch are you using?

 

[smarteyeball]

What utility do you recommend to test the hard drive's speed?

HD Tune

 

[BeechV35Pilot]

for you, I would expect at least 30-40MB/s on a gigabit network with Cat5e, all ports at 1000 and full duplex.

What switch are you using?

It is an 8 port Dell PowerConnect 2708.

Just ran HDTune Pro (trial) read test and the results were: About 50MB/sec on the 80GB system drive and 80MB/sec on the 500GB data drive.

 

[pparks1]

Might want to take a look at these threads;
Troubleshooting Slow Read/ Transfer Speeds for a Windows Home Server | ServeTheHome.com
Technical Blog for Jim Beveridge: Dell PowerConnect 2708 Performance

This switch really should be providing better performance than what you are seeing?

 

[H2SO4]

Interesting thread

Factors to consider:

- The choice of network protocol(s) used for testing. For example, FTP is generally much faster than SMB. The former emphasises speed while the latter is "chatty" and rich in the sense used by the UI shell.

- Even if you standardise on a particular protocol for testing purposes, say SMB, the way it's employed is very important. Explorer may sometimes use a different SMB transfer mode to CMD.exe, making the latter faster for simple file transfers.

- Windows-to-Windows, the transfer will negotiate the use of SMB2, at least if both sides are Vista or later Windows variants, and that's much faster than the SMB1 used by XP and most versions of samba, especially over high latency networks. SMB1 has a 64KB limit on the "message" size, which defeats the benefit of window scaling - the sender still has to pause and wait for ACKnowledgement every 64KB. With say 100ms network latency, that produces a theoretical maximum of only 640KB/sec.

- The NIC hardware and NIC driver, assuming the machine is not disk-bound. In other words, for those using enterprise-level disk setups, or something like multiple velociraptors in a RAID 0 stripe, the limiting factor becomes the NIC hardware & driver. Everything else being equal, that $1000 server NIC is going to do a lot better than the $29 Realtek jobbie. In fact, true "gigabit" tends to be attainable only with server-grade NICs.

- The hardware and OS horsepower of the other transfer endpoint. It doesn't matter that you've got a fire-breathing i7 with 47TB of RAM and 16 velociraptors hooked up via pure silver cables if you're copying from a budget NAS whose processor is a 486SX25.

- The type of content being copied. Many small files require far more overhead to manage than a few monstrous ISOs or something similarly large. Hence, effective throughput will frequently be inversely proportional to the average file size.

- Anti-virus filter drivers. If every chunk of the transfer is to be inspected and verified, the overall speed won't be nearly as fast as you'd obtain without AV in the picture. Also, different AV products will use different mechanisms to do their job, leading to different levels of throughput performance.