TCP Choosing optimal window size? - tcp

I know I need to get the RTT but I really don't get how to calculate it.
For some contextualization:
I have a ns2 script which simulates a server and a receiver with 2 routers in the way. All three links - server to first router, first router to second router, second router to receiver - have different transmission speeds.
The propagation delay is 10ms for the first two links and 3ms for the third one.
I'm sending a 3MB file in 1000 bytes packages (TCP's default packet size), 3146 packages total.
I don't want you to calculate it (the RTT) for me of course, I just want to know how to do it. :\

You can use Agent/Ping and collect RTT. Here is an example TCL snippet
There is only one problem if there is to much trafic de ping packet will be drop (you can see if you collect trace-all)

Related

Network: sending a video (in theory)

Context
We have got un unsteady transmission channel. Some packets may be lost.
Sending a single network packet in any direction (from A to B or from B to A) takes 3 seconds.
We allow a signal delay of 5 seconds, no more. So we have a 5-second buffer. We can use those 5 seconds however we want.
Currently we use only 80% of the transmission channel, so we have 1/4 more room to utilize.
The quality of the video cannot be worsened.
Problem
We need to make the quality better. How to handle lost packets?
Solution proposition
A certain thing - we cannot use TCP in this case, because when TCP detects some problems, it requests retransmission of lost data. That would mean that a packet would arrive after 9 seconds, which is more than the limit.
Therefore we need to use UDP and handle those errors ourselves. How to do it then? How to make sure that not so many packets will be lost as currently, without retransmitting them?
Its a complicated solution, but by far the best option will be to add forward error correction (FEC). This is how images are transfered form space probes where the latency is measures in minutes or hours. Its also use by cell phones where delayed packets are bad for two way communication.
A not as good, but eaiser to implement option is to use UDT. This is a UDP with tcp like retranmission library, but allows you much more controll over the protocol.

Idea to handle lost packets (theory)

Context
We have got un unsteady transmission channel. Some packets may be lost.
Sending a single network packet in any direction (from A to B or from B to A) takes 3 seconds.
We allow a signal delay of 5 seconds, no more. So we have a 5-second buffer. We can use those 5 seconds however we want.
Currently we use only 80% of the transmission channel, so we have 1/4 more room to utilize.
The quality of the video cannot be worsened.
We need to use UDP.
Problem
We need to make the quality better. How to handle lost packets? We need to use UDP and handle those errors ourselves. How to do it then? How to make sure that not so many packets will be lost as currently (we can't guarantee 100%, so we only want it better), without retransmitting them? We can do everything, this is theory.
There are different logic's to handle these things.It depends on what application you are using. Are you doing real time video streaming? stringent requirements?
As you said you have a buffer, you can actually maintain a buffer for the packets and then send an acknowledgement for the lost packets (if you feel you can wait).
As this is video application, send acknowledgements only to the key frames. make sure that you have a key or I frame and then do interpolation at the rx side.
Look into something called forward error correction, fountain codes, luby codes. Here, you will encode the packets 1 and 2 and produce packet 3. If packet1 is lost, use packet3 and packet2 to get the packet1 back at the rx side. Basically you send redundant packets. Its little harsh on network but you get most of the data.

flow control implementation - how

I'm sending 1k data using TCP/IP (using FreeRTOS + LwiP). From documents I understood that TCP/IP protocol has its flow control inside its stack itself, but this flow control is dependent on the Network buffers. I'm not sure how this can be handled in my scenario which is described below.
Receive data of 1k size using TCP/IP from wifi (this data rate will be in 20Mb/s)
The received Wifi data is put into a queue of 10k size10 block, each block having a size of 1K
From the queue, each block is taken and send to another interface at lower rate 1Mb/s
So in this scenario, do I have to implement flow control manually between data from wifi <-> queue? How can I achieve this?
No you do not have to implement flow control yourself, the TCP algorithm takes care of it internally.
Basically what happens is that when a TCP segment is received from your sender LwIP will send back an ACK that includes the available space remaining in its buffers (the window size). Since the data is arriving faster than you can process it the stack will eventually send back an ACK with a window size of zero. This tells the sender's stack to back off and try again later, which it will do automatically. When you get around to extracting more data from the network buffers the stack should re-ACK the last segment it received, only this time it opens up the window to say that it can receive more data.
What you want to avoid is something called silly window syndrome because it can have a drastic effect on your network utilisation and performance. Try to read data off the network in big chunks if you can. Avoid tight loops that fill a buffer 1-byte at a time.

Why does the number of links not matter in the transmission time of circuit switching?

I am reading a book about networking, and it says that, in a circuit switching environment, the number of links and switches a signal has to go through before getting to destination does not affect the overall time it takes for all the signal to be received. On the other hand, in a packet switching scenario the number of links and switches does make a difference. I spent quite a bit of time trying to figure it out, but I can't seem to get it. Why is that?
To drastically over-simplify, effectively a circuit-switched environment has a direct line from the transmitter to the transmitted once a connection has been established; imagine an old-fashioned phone-call going through switchboards. Therefore the transmission time is the same regardless of hops (well, ignoring the physical time it takes the signal to move over the wire, which very small since it's moving at the speed of light).
In a packet-switched environment, there is no direct connection. A packet of data is sent from the transmitter to the first hop, which tries to calculate an open route to the destination. It then passes its data onto the next hop, which again has to calculate the next available hop, and so on. This takes time that linearly increases with the number of hops. Think sending a letter through the US postal system. It has to go from your house to a post office, then the post office to a local distribution center, then from the local distribution center to the national one, then from that to the recipient's local distribution center, then to the recipient's post office, then finally to their house.
The difference is that only one connection at a time per circuit can exist on a circuit-switched network; again, think phone line with someone using it to make a call. Whereas in a packet switched network many transmitters and receivers can be sending data at the same time; again, think many people sending/recieving letters.

Arduino RF sensor network

I'm currently designing a sensor network that will have small ATtiny85 probes that each have a temperature sensor, a barometer, and a humidity sensor. I think I will use these (http://goo.gl/TqaDjl) to communicate as they are low cost and don't need much range. Im not sure though how I will get the probes to communicate with the main control, as the transmitter transmits digitally and I will have +20 probes that all need to send data without signals overlapping or getting messed up every minute. I think the easiest way would be to time the probes so that they don't overlap in transmission but I'm not sure.
Questions:
-Is using RF the cheapest and best option for this system?
-How can I prevent communication overlapping?
-What is the easiest way to send data digitally from an arduino (or ATtiny85)?
I guess I'm late to the party, but I'll offer some insight into collision control with a ton of chattering transmitters on one link, a la 802.11. This is somewhat packetized.
If two transmitters try to transmit at the same time, you're bound to get a mangled mess of rotten bacon at the receivers.
A simplified version of WiFi-style collisions would be good. Basically, it uses preambles that can be detected, and for longer transmissions that have a higher chance of conflicting, it can use shorter request/clear to send packets.
While this is likely overkill, I'd go for preambles. Start by transmitting a steady stream of something recognizable, like in hex, 555533330f0f00ff which is basically alternating 1s and 0s but with changing frequency(0101, then 0011, then 00001111, and so on), a readily recognizable pattern that is unlikely to be given off by stray radiation or noise.
This pattern could undergo a shift so there's a finite set of other preambles that should be bitwise-shifted relative to the original.
If a transmitter detects this preamble, it should STOP and wait. If you limit all packets to a certain temporal length, collisions should not occur if you wait sufficient time between packets. If during the time of one packet, a preamble is heard, then your station should wait for the full length of the transmission(listening to its length and other header fields so it knows how long to wait). Once the packet is done, your station can transmit its preamble.
This is where the WiFi resemblance stops and simpler protocols take over.
Note that if 2 stations are waiting on a packet they can start their preambles almost simultaneously. To resolve this, each station should have a different zero bit flipped in its preamble. If it detects a 1 for that bit, it sees that there's another station preambling, and should back off.
Each station should wait a certain delay(up to you) after each packet so other stations can start their transmissions.
A few sketches of the communication patterns show that this is sufficient for your needs.
Now if it's a master-slave-style system as long as you only have one network it should be easier since there should only be one outstanding request that would involve a slave transmitting.
Those will be by far the cheapest method. As for the best method, there are a variety of choices much better, but more expensive. A network of Xbee modules comes to mind, but those are much more expensive than $1.25 a pair.
Using the RF modules is very do-able however. To prevent communication overlapping, put a RF transmitter and receiver on each sensor node and the main hub. The main hub can send "hey sensor1 give me your data", which gets broadcasted to all of the sensors. However, only sensor1 will realize "hey I am sensor 1, here is my data" which the hub will listen for. Then, the hub will go on and say "hey sensor2 send me your data" and so on and so forth.
I think your original approach may be best. The approach of putting a Tx and Rx on every device may be affordable, but I question if it will work. With 20 devices transmitting on the same frequency, which one will the receiver "hear". Most important, how will a device receive any remote transmitter's signal when its own transmitter is very close? Keep in mind: these are AM radios and will "send" a carrier even if not sending any data. Get a small number of transmitters before trying to go full scale.
To avoid the problem of receiving the one active transmitter among the soup of inactive transmitters, you want only 1 transmitter powered at 1 time. You would control Vcc to one transmitter, turn it on, send the burst of data, and then power it off.
-How can I prevent communication overlapping?
You can't -- you have to accept that there will be occasional overlaps. Add a CRC to the transmitted data so that the receiver can detect garbage.
The timing of the multiple transmitters is surely a project in itself. You surely don't want to run them all at the same transmission period. They may not collide at the beginning, but when two devices did drift together and start colliding, they would stay together and collide for a long time, until the clocks drifted apart.
I would start with something simple. For example with three devices, run the transmissions at 2000 ms, 2200 ms, 2400 ms period (use EEPROM to configure). That way, if a pair happens to collide at one data point, then next transmissions that pair will be 200 ms apart.

Resources