Slot Time

hassantalal785

I have been trying to understand concepts of slot time in Ethernet. However i am unable to understand some concepts :
Q1)I read that as we move from 10 Mbps to 1 Gbps , slot time increases to 4096 bit times from 512 bit times (for 10 and 100 Mbps LANS).Can any one summarize why slot time increases where as bit time decreases from 10 Mbps to 1 Gbps.i am not able to correlate both factors.

Q2)Is slot time and Back off time , one and the same thing ?

Q3)How 512-bit slot time establishes the minimum size of an Ethernet frame as 64 bytes ?

Q4)How this slot time is irrelevant in Full duplex mode?

Roguetadhg

Slot time is the amount of time the data has on the wire. It's the assumed time that a computer would normally wait.

If you have 1 lane road, between two towns. Both mayors decided to make a law that "Every car needs to wait 20 minutes before going down this road" This is the expected time per slot/data. Town 1 Sends a car over. Town 2 Sends a car over. Bam!.

Back off time happens when a collision has been detected. The amount of time the NIC waits before attempting to send again is the Back-Off time.

The two cars collided. Sirens roaring, people tweeting, and chaos ensues. The Emergency crews clean up the mess. At the same time, both towns back off from sending any news cars over to the other town.

Slot time doesn't apply for full duplex modes because the wires are dedicated to TX or RX, Transmit or Receive. There's no wait, no CSMA/CD needed.

Instead of a one-lane street where only 1 car can be on the road at one time, it's a 2 lane road. Cars / Data go both way. You can send the cars one right after the other. Both towns have happier citizens, newer cars, and less accidents.

hassantalal785

But why slot time increases for Gigabit Ethernet ???

Roguetadhg

I don't know. I'm looking.


Edit: Im going to take a stab in the dark here with why it's a longer bit time. I hope someone else could chime in, because I'm not entirely sure.

Firstly -
Slot time is defined as: The time taken for a pulse to travel through the longest permitted length of network medium.


Gigbit distances can span over a lot more cabling than 10/100:
10/100 is about 100 Meters

Gigabit


1000BASE‑CX
Twinaxial cabling
25 meters


1000BASE‑SX
Multi-mode fiber
220 to 550 meters dependent on fiber diameter and bandwidth^[3]


1000BASE‑LX
Multi-mode fiber
550 meters^[4]


1000BASE‑LX
Single-mode fiber
5 km^[4]


1000BASE‑LX10
Single-mode fiber using 1,310 nm wavelength
10 km^[4]


1000BASE‑EX
Single-mode fiber at 1,310 nm wavelength
~ 40 km


1000BASE‑ZX
Single-mode fiber at 1,550 nm wavelength
~ 70 km


1000BASE‑BX10
Single-mode fiber, over single-strand fiber: 1,490 nm downstream 1,310 nm upstream
10 km


1000BASE‑T
Twisted-pair cabling (Cat‑5, Cat‑5e, Cat‑6, or Cat‑7)
100 meters


1000BASE‑TX
Twisted-pair cabling (Cat‑6, Cat‑7)
100 meters




My overall guess is that the permitted distance increases (Fiber) explains the Gigabit time increase.


Used several different Wiki links to piece the theory together. My ICND1/2 Books didn't cover "Slot Time" in the appendix.

MAC_Addy

hassantalal785 wrote: »

But why slot time increases for Gigabit Ethernet ???

10 and 100Mbps has 64 bytes and 1Gbps has 512 bytes.

Roguetadhg

MAC_Addy: where do you find this at?

MAC_Addy

I have an ethernet design sheet/document that's been in my desk for a while now. I had to refer to it to be exact on the bytes, turns out I was correct

Roguetadhg

+ Rep for Correctness

MAC_Addy

Thanks mate, I appreciate it. I'll scan it later and send it to you if you want? I believe it was a free document in my network world subscription.

Roguetadhg

Heck yeah. I'll take a copy.

hassantalal785

MAC_Addy wrote: »

10 and 100Mbps has 64 bytes and 1Gbps has 512 bytes.

These sizes are minimum Ethernet frame size for each type of Ethernet Lan ?

NetworkVeteran

hassantalal785 wrote: »

These sizes are minimum Ethernet frame size for each type of Ethernet Lan ?

No. The minimum Ethernet frame size is always 64 bytes.

Forsaken_GA

NetworkVeteran wrote: »

No. The minimum Ethernet frame size is always 64 bytes.

Yes and no. GigE doesn't change it, but the minimum size it actually transmits is 512 bytes, anything less than that, and it's padded out to 512.

NetworkVeteran

Forsaken_GA wrote: »

Yes and no. GigE doesn't change it, but the minimum size it actually transmits is 512 bytes, anything less than that, and it's padded out to 512.

This technique, if I recall correctly, is called Carrier Extension.

Interestingly, the 10GE and 100GE standards do not provide for carrier extension, since they're meant to always operate in full-duplex mode.

[Edit: I just double-checked. The carrier extension comes after the FCS, so padding is probably not the appropriate term. The minimum frame size remains 64 bytes. I have sniffed 1GE frames on the wire and some tools do not show these extra bytes. I remember having to explain before why a shiny new GE link was not getting as much throughput as expected.]

Forsaken_GA

hassantalal785 wrote: »

I have been trying to understand concepts of slot time in Ethernet. However i am unable to understand some concepts :
Q1)I read that as we move from 10 Mbps to 1 Gbps , slot time increases to 4096 bit times from 512 bit times (for 10 and 100 Mbps LANS).Can any one summarize why slot time increases where as bit time decreases from 10 Mbps to 1 Gbps.i am not able to correlate both factors.

Ok, let me see if I can explain this a bit better -

The slot time is the maximum amount of time it takes data to go from one end of the network to the other, in that time, a NIC can transmit 512 bits. If a NIC sees a frame come in at less than 512 bits, it knows a collision occurred.

The problem is that GigE is actually physically faster on the wire. You can't make it transmit any slower, that would break the laws of physics. So you either shorten the effective cable length to what it would take gige to travel end to end in the span of 512 bits (about 10 meters), or you have it toss out more data in the same time frame, which is why we have bigger slot times.

Q2)Is slot time and Back off time , one and the same thing ?

No

Q3)How 512-bit slot time establishes the minimum size of an Ethernet frame as 64 bytes ?

Easy. 512 / 8 = 64. (I'm not being a smartass, that's the answer. 512 bits is 64 bytes, and that's the minimum amount you have to transit to know a collision hasn't occurred)

Q4)How this slot time is irrelevant in Full duplex mode?

[/quote]

Collisions can't occur in full duplex, so the time slot is irrelevant.

Forsaken_GA

NetworkVeteran wrote: »

This technique, if I recall correctly, is called Carrier Extension.

Interestingly, the 10GE and 100GE standards do not provide for carrier extension, since they're meant to always operate in full-duplex mode.

[Edit: I just double-checked. The carrier extension comes after the FCS, so padding is probably not the appropriate term. The minimum frame size remains 64 bytes. I have sniffed 1GE frames on the wire and some tools do not show these extra bytes. I remember having to explain before why a shiny new GE link was not getting as much throughput as expected.]

Yeah, Padding not in the technical sense that we're used to in order to make a frame or packet the correct, but the concept remains the same, GigE is going to transit 512 bytes as it's smallest frame hehe

Dexter Das

with reference to the car driven between two towns on a single lane , let us see this way.
How a host will come to know that a collision occured, it's only by the 32- bit jam sequence. Now take a 32- bit jam sequence as a mobile van which comes and updates that an accident occured so please stop for some random time till everything is okay.
This random time for which the cars will stop before starting again is analogous to backoff time.
Take into consideration that the host or car starting can only recognize the 32 bit jam sequence at its starting point not in between the road(i.e at the NIC itself).
The timings should be such that the time taken by the jam sequence to reach to the host should be less than the time before which the host(s) starts again.
If that is not the case before the jam sequence have acknowledged the host the cars must have started again and the same collision again.
So , the transsion should be for some time that the jam sequence should reach before it transmits again.
This time is analogous with the slot time.
Guess It would help you.