On a lan whats the speed diffrence with a 10mb card vs a 100mb card?

[TheLizardMan]

whats the REAL data trasfer rate i can get on a 100mb lan with a 10mb and 100mb card? i have a 10mb card and am thinking of getting a 100.

 

[ttn1]

mb stands for megabit. There are 8 bits to a byte. So 10 mb/s is equal to 1.25 MegaBytes/Second. The 100mb card will be 10 times faster or 12.5 MB/s. If you have cards that are capable of full-duplex you can get up to 25MB/s. One drawback though, if you have older harddrives the drive speed, not the network, will limit the transfer.

With the ide drives I have, older 5400 RPM type, "REAL" transfer rates are around 1-2 MB/s. From a server with SCSI drives, to my ide drive machine, I can download at about 9-10 MB/s.
ttn1

 

[ktwebb]

Actually, 100 Mb is not 10X the speed of 10BaseT. All those numbers are theoretical limits. You wont get 1.25 MB from a 10BaseT LAN or 12.5 MB from a 100 Mb LAN. THere are alot of variables that determine your data transfer rate, but you can expect anywhere from 3 to 5 times the speed going from a 10 to 100 Mb network. That is point to point. THe real advantage of 100 Mb is when you start adding nodes to your LAN.

 

[spidey07]

KT, your smoking crack.

By definition 100 base-t is 10 times faster than 10 base-t. All aspects of the CSMACD spec are scaled by a factor of 10.

As far as true speeds that depends largely on frame size and application. For example:

Packets/sec rate for small frames (64 bytes) = 14,880 pps or 7.618 Mb/s for 10BaseT and 76.18 Mb/s for 100BaseT

Packets/sec rate for largest frame size (1518) = 812 pps or 9.860 Mb/s for 10BaseT and 98.60 Mb/s for 100BaseT

The difference between 76 Mb/s and 98 Mb/s is pretty significant. Really just proving how important frame size is.

cheers!

 

[ktwebb]

Well, I wasnt speaking about the definition of the two data rates. What a jerkoff. Of course 100 is 10 times 10. I am talking real world. You can spout all the specs you want. My post is based on my experience from real world benchmarking, not specification standards. But you go ahead and worry about your CSMACD specs and calculations based on packet sizes if you wish. This guy wanted reality, not specs and theoretical limits read from a RFC. What a putz.

 

[spidey07]

Nothing like a friendly debate.

I too have been benchmarking and building networks for about 7 years and have worked with 100 Base-T before the IEEE passed it in 1995. I remember the old discussions of "Well, A computer can never drive 100 megabits in the first place so why even use it?". That point was disproved years ago.

Things such as drivers, application (NCP vs SMB vs FTP vs HTTP) and OS weigh heavily on througput.

Four PCs running three FTP sessions (12 total) to a decently powered Solaris workstation sustained 93 Megabits/sec utilization on the solaris NIC. If running in a single session (2 machines with one session) mode then sure, you're never going to get much higher than 60 percent due to windowing and acks.

BUT, moving from 10Base-t to 100Base-t does give you ABOUT 10 times the throughput. Please remember I'm talking about network speeds and these are very dependant on DISK speed, memory, processor, pci bandwidth etc. When testing I try to take these out of the picture by using hardware that is faster than the NIC

 

[ktwebb]

I will say this. I made the assumption, perhpas mistakenly that the original poster was using windows. I am only guessing that he isnt talking about using Sun Boxes, but hey, maybe I am wrong. Stanger things have happened.

"Four PCs running three FTP sessions (12 total) to a decently powered Solaris workstation sustained 93 Megabits/sec utilization on the solaris NIC. If running in a single session (2 machines with one session) mode then sure, you're never going to get much higher than 60 percent due to windowing and acks."

If you read the post just above this quote, I mentioned that your real advantage starts showing as you add nodes.

"BUT, moving from 10Base-t to 100Base-t does give you ABOUT 10 times the throughput"

We'll just have to agree to disagree on this one. I have too have been setting up networks for a while. Not 7 years. About 4. Never seen 10X the performance in any scenario I have ever used. Regardless of hardware/software configurations. Vary seldom have I seen more than 5 times the speed from a two PC FTP transfer, only when nodes get added. Work almost exclusively with windows, save a few Linux server/windows client setups, and a few with Mac environments, so I cant speak to other OS performance benchmarks. But hey, mileage will vary I suppose. As far as a friendly debate. Sorry for the name calling, but when you come at me with a smoking crack comment the bulldog takes over. No harm, no foul.

 

[CTR]

Spidey, KT:
Maybe you should say it is 10X the capacity, instead of 10x faster. Speed is measured and dependent on HW/SW, while capacity is a calculated value based on given HW/SW specs. That way you CYA if LizardMan or whoever comes back and says, "I only getting 23Mbps throughput on my network!"

Just a thought. I've been setting networks up for about 6 years now, and I have learned that CYA is more important than anything else.

 

[spidey07]

KT,

Agree to disagree? Probably. We are looking at this question in different contexts. I think you are responding based solely on a single session file transfer in which case, YES you might only see a 5-6 times improvment in raw data transfer.

This single session suffers from the same delays at a 10 megabit LAN only with 100 megabit the delay is proportionately larger compared to the serialization delay on the physical layer. The acknowledgement time stays constant while bandwidth increases. It is this ACK time that prevents you from getting higher results.

Try this sometime:
With a node to node transfer start up 10 different FTP sessions to the source of the data and transfer some huge files on a 10 meg enet. Compare this to the same test using 100 meg enet. You'll be pleasantly suprised on the host end. We've had to do this kind of test many times studying Quality of Service mechanisms for various media. The QoS doesn't do much for the single session (ack time again) but does wonders for managing the multiples. It is hard to congest a link with a single session.

Oh well, no biggie on the name calling. I'm like a reed.

 

[ktwebb]

No need to try it. Been there, done that. No arguments. I go back to my comments about adding nodes. I never really questioned bandwidth scalibility. Multiple PC's transfers or extra FTP sessions is where you see the most improvement and might get close to the bandwidth limitations. Guess I should have clarified my original post and subsequent ones for that matter. The question was "whats the REAL data trasfer rate i can get on a 100mb lan with a 10mb and 100mb card". Fairly ambiguous and I took the single session perspective. Again I made assumptions based on what I read because the question was vague and offered little distinction. Oh well, this conversation has become tiresome. Now is the time on Sprockets where we dance. Peace

 

[spidey07]

true...true

 

[coloumb]

Any reasons why my home lan would not perform as well as the lan at work:

Home Lan:
Computer #1:
WindowsME
PIII-667 w/384MB ram
7200 RPM IDE Master and Slave
10000 RPM SCSI Drive used for swap file/storage
Intel 82559 10/100 NIC card

Computer #2:
Windows98
PIII-700 w/128MB ram
7200 RPM IDE Hard Drive
Netgear 10/100 NIC [can't remember the model #].

Hawkings 4 port 10/100 hub which is connected to a Linksys single port Gateway router connected to Cable Modem. [both computers are connected to the hub]

FTP server [guild ftp because it's free] on computer #1. WSFTP Lite on Computer #2. Highest rate of data transfer over the LAN is approx 450Kbps. This is the highest data tfr rate as downloading from ftp.cdrom.com.

Setup at Work:

Computer #1: Dell Optiplex PIII-500 [IDE Hard drive, 128mb memory, Intel 82559 NIC] running WindowsNT 4.0

Computer #2: HP Vectra PP200, 32MB memory, WindowsNT 4.0, HP 10/100TX NIC [ISA].

Both computers are connected to the company LAN.

Computer #1 running GuildFTP. Computer #2 running WSFTP Lite. Highest date tfr rate 8.4Mbps [according to WSFTP].

Other than the OS's, why would my network at home not match or exceed the network tfr rate at work?

 

[spidey07]

That is strange. Maybe start a new thread on this specific problem.

Keep in mind though that the network gear at work is probably "professional" grade. The home stuff just can't come close to "real" network gear from cisco, nortel, 3com, etc.

spidey

ps - do you use a proxy or socks proxy with that firewall. If so that can cause your pitifull speeds.