Cat6 vs Cat5 - what should we be using...

[DoFoT9]

Cat5e is theoretically capable of 1Gb/sec (gigabit) speeds but you never reach it in real life.

interesting statement, do you have anything to back it up? im sure you know that a general everyday user computer wouldnt be able to max it out - maybe if somebody had a RAID setup they might be able to give it a run for its money (and maybe get around 100MB/s-110MB/s).

 

[Queso]

interesting statement, do you have anything to back it up? im sure you know that a general everyday user computer wouldnt be able to max it out - maybe if somebody had a RAID setup they might be able to give it a run for its money (and maybe get around 100MB/s-110MB/s).

Ethernet never manages to reach anywhere near it's theoretical maximum, even over optical fiber.

 

[DoFoT9]

Ethernet never manages to reach anywhere near it's theoretical maximum, even over optical fiber.

it would still be interesting to see some solid proof, do you have any?

the best speeds i can get are 50MB/s, but that is because of the limitations in the disks read/write speeds and not because of the ethernet.

 

[benlangdon]

it would still be interesting to see some solid proof, do you have any?

the best speeds i can get are 50MB/s, but that is because of the limitations in the disks read/write speeds and not because of the ethernet.

and how much do you pay for your internet.

 

[cube]

Testing your 100 Mbps cabling against the internet only makes sense in places like South Korea.

 

[DoFoT9]

and how much do you pay for your internet.

Testing your 100 Mbps cabling against the internet only makes sense in places like South Korea.

i guess i should have said this is while transferring files on my local network sorry...

BTW FTR: i pay $100 for ADSL2+ 25GB cap

 

[cube]

i guess i should have said this is while transferring files on my local network sorry...

I was referring to the people posting screenshots of internet speed tests from their home.

 

[yg17]

it would still be interesting to see some solid proof, do you have any?

It's like that for ANY transfer medium. USB 2.0 never reaches its theoretical maximum of 480 Mbps, Firewire 800 never reaches its theoretical maximum of 800 Mbps, SATA never reaches its theoretical maximum of 3 Gbps, etc. There is always overhead involved in these things. You'll never get 1 Gbps on a gigabit network.

And Cat5e handles gigabit just as well as Cat6. If you're wiring up a house and putting cable in walls, use Cat6 to future proof it, but if you're just hooking up 2 pieces of equipment a few feet apart with an easily replaceable cable, save your money and get Cat5e.

 

[DoFoT9]

It's like that for ANY transfer medium. USB 2.0 never reaches its theoretical maximum of 480 Mbps, Firewire 800 never reaches its theoretical maximum of 800 Mbps, SATA never reaches its theoretical maximum of 3 Gbps, etc. There is always overhead involved in these things. You'll never get 1 Gbps on a gigabit network.

oh of course ther is overhead. but how much, that is the question.. are we talking 1/2 of what the data packets are? twice the size? (this is excluding packet loss of course).

USB2.0 wont ever reach 480Mb/s because of the crappy HDDs that are put into them, same with FW400&800, also a similar story with SATAII.

according to wiki, the full frame size for an ethernet packet is 64bytes up to 1518bytes. the packet size can be anywhere from 46bytes up to 1500 bytes.

if you take the smallest packet possible, 14byte mac header + 46byte data + 4byte checksum. the overhead is roughly 1/5th the data being sent. this would mean that one would expect 100MB/s theoretical data transmission rates max.

anyway, i want to see proof for this...tests would be nice

 

[uberamd]

I ran a test on a corporate 10/100 Cat6 network and here is what I got:

Downloading a file from server to my system over Cat6: 95,875 kbps which is 11.70349MBps. That translates to 93.62792 Mbps, 6.37208 Mb shy of capping 100Mbps.

 

[yg17]

oh of course ther is overhead. but how much, that is the question.. are we talking 1/2 of what the data packets are? twice the size? (this is excluding packet loss of course).

If you ignore overhead and have hard drives on both ends capable of sustaining those speeds, you still won't reach 100 Mbps (or 1 Gbps if that's what you're using). You may come close, but there is always going to be something preventing it, whether it's outside noise, a slight defect in the cable, etc. Perhaps if you stuck the entire network inside a faraday cage so there's no outside interference, and built a high quality, extremely expensive cable you could peak out at 100 Mbps, but in the real world, there is always a limiting factor.

USB2.0 wont ever reach 480Mb/s because of the crappy HDDs that are put into them, same with FW400&800, also a similar story with SATAII.

Hard drive speeds are half the problem. Again, take overhead out of the equation and assume you have hard drives capable of reaching those speeds. You still won't. USB 2.0 bursts, meaning it transfers at a slower speed, occasionally bursting up towards 480 Mbps, but the average speed over a transfer would still be much lower. Firewire is sustaining, meaning it starts transferring at a higher speed and stays around that speed for the length of the transfer. Even then, it will never quite get up to 400 (or 800) Mbps, but that is why people claim that Firewire 400 is better than USB 2.0 even though USB is faster on paper, because average transfer speeds during a transfer tend to be higher.

Misleading? Yeah, but we've become used to it, just like the whole 1,000 vs 1,024 bytes in a kilobyte for hard drive sizes.

anyway, i want to see proof for this...tests would be nice

I don't have any tests, just a Cisco network certification which means I've studied this and know what I'm talking about, or at least I have a piece of paper from Cisco that says I do

 

[killerrobot]

interesting statement, do you have anything to back it up?

I ran a test on a corporate 10/100 Cat6 network and here is what I got:

About the only place worried about having a 1GB backbone are large corporations, private companies, large public institutions, etc...
The 1GB backbone isn't for the internet (that will always be limited to the outside connection) but for the companies' intranet. Still, the maximum speeds as others have said, will always fall a little short.

 

[yg17]

About the only place worried about having a 1GB backbone are large corporations, private companies, large public institutions, etc...
The 1GB backbone isn't for the internet (that will always be limited to the outside connection) but for the companies' intranet. Still, the maximum speeds as others have said, will always fall a little short.

There are internet connections available faster than 1 Gigabit if you have the money, but once you're up to those speeds, you're definitely using fiber so Cat5 vs Cat6 becomes irrelevant.

 

[uberamd]

About the only place worried about having a 1GB backbone are large corporations, private companies, large public institutions, etc...
The 1GB backbone isn't for the internet (that will always be limited to the outside connection) but for the companies' intranet. Still, the maximum speeds as others have said, will always fall a little short.

I work for a major university, and all of the ports are still 100Mbps. We are just starting to roll out wireless N, but even for large businesses, gigabit isn't all too prevalent. At least in many of the businesses I have seen. There is gigabit rolled out, but generally its not rolled out to all of the end users.

 

[Queso]

I work for a major university, and all of the ports are still 100Mbps. We are just starting to roll out wireless N, but even for large businesses, gigabit isn't all too prevalent. At least in many of the businesses I have seen. There is gigabit rolled out, but generally its not rolled out to all of the end users.

That's because in most places it simply isn't required. Looking at Solarwinds now our maximum utilisation on a LAN uplink today has been 6%. Over the past month we've maxed out at 14% when one of the users must have been doing a large copy from their desktop.

It makes me wonder why we waste so much time putting QoS on our gigabit links. Even maxing out on the data, VoIP and VC stuff we're not even touching the sides of available bandwidth.

 

[killerrobot]

There are internet connections available faster than 1 Gigabit if you have the money, but once you're up to those speeds, you're definitely using fiber so Cat5 vs Cat6 becomes irrelevant.

True, but unless you own the outside connection, you're still at someone elses mercy so to say.

 

[uberamd]

True, but unless you own the outside connection, you're still at someone elses mercy so to say.

Plus, I can't imagine too many servers are able to send or receive your data at 1Gbps anyway. Downloading from multiple sources would benefit you, but from a single server, pointless.

That's because in most places it simply isn't required. Looking at Solarwinds now our maximum utilisation on a LAN uplink today has been 6%. Over the past month we've maxed out at 14% when one of the users must have been doing a large copy from their desktop.

It makes me wonder why we waste so much time putting QoS on our gigabit links. Even maxing out on the data, VoIP and VC stuff we're not even touching the sides of available bandwidth.

At our university all on campus residence halls (dorms and apartments) still only have 10Mbps ports. They do this for 2 reasons. First of all most of the halls are old. Second, it automatically restricts network use. Users can't push over 10Mbps if they live on campus, and that cuts down on bandwidth abuse.

But even like that, our pipe is generally always full with those darn freshman downloading torrents, music, and what not. Every student has an external IP, so if someone is downloading a illegal torrent, its almost guaranteed they will get caught, and booted off the network until they attend a meeting about illegal downloading that our campus puts on every week due to the volume of people getting caught. Heck, I got caught by 3 companies in a 1 week period (Blizzard, Lucas Arts, and I forget the other). Its 3 strikes and you are permanently disconnected, luckily all 3 infractions came at the same time so it counted as my first strike

Anyway, where I was going with this is that our campus implements old technology without QoS restrictions because 10Mbps naturally provides a speed cap, for the ever so crowded residence all pipe. Gigabit to each room without a cap would be a disaster.

At work, I can cap 100Mbps easily over the summer when there are no students bogging the pipe down.

 

[yg17]

At our university all on campus residence halls (dorms and apartments) still only have 10Mbps ports. They do this for 2 reasons. First of all most of the halls are old. Second, it automatically restricts network use. Users can't push over 10Mbps if they live on campus, and that cuts down on bandwidth abuse.

That means all it takes is 100 users to start downloading crap to max out the pipe, that's not that many, especially in the evening and weekends when people are bored.

I worked IT at a university too for awhile, and I don't remember how big the pipe coming into the dorms was, but each room port was 10 mbps, when I left, they were in the process of upgrading the residential network with gigabit switches but the dorm room ports were set to 100 mbps, but they still did bandwidth throttling to keep everything in check. The more you downloaded, the slower your speeds went for a period of time, but I don't believe it would ever drop you below 1 mbps.

 

[geoffreak]

This thread cracks me up.

In networking, the best you can have in actual speeds is 80% of the theoretical. The reason behind this is because before and after each byte sent, there is a starting and stopping bit. With 8 bits in a byte, that equates to 10 bits sent for every byte, or 80% efficiency.

 

[JGruber]

On our campus at the high school I work for, we have 1GB to the computer, and 10GB between switchs, our WAN is 10GB to our main district office, and 100MB to the county.

We very rarely peak any of the fiber connections, and this is on a campus with 1,400+ computers accessing the network at any given time.

 

[uberamd]

On our campus at the high school I work for, we have 1GB to the computer, and 10GB between switchs, our WAN is 10GB to our main district office, and 100MB to the county.

We very rarely peak any of the fiber connections, and this is on a campus with 1,400+ computers accessing the network at any given time.

I believe you mean Gb, and Mb

 

[JGruber]

I believe you mean Gb, and Mb

I have the habit of typing GB/MB/KB, etc... been doing it for years.

 

[uberamd]

I have the habit of typing GB/MB/KB, etc... been doing it for years.

Ah. Thats one thing I became very anal about because when you work on networks like I do, Mb and MB carry such vastly different meanings that to confuse them could cause major problems.

 

[JGruber]

Ah. Thats one thing I became very anal about because when you work on networks like I do, Mb and MB carry such vastly different meanings that to confuse them could cause major problems.

This is very true, I tend not to say it as much as before, but I am getting better

 

[pooryou]

Same rule applies. You will never reach the theoretical speed of 1Gb/sec.

So what? You will still see a noticeable performance increase over 100base-T.