TheRoDent
Cool Ideas Rep
Why pings will never be awesome:
I've dug around quite a bit, reading some of the specs of the IPWireless and UMTS technology, and have written this explanation of why I believe that pings will never be awesome on MyWireless. Note, that when I mean awesome, I mean, good enough for gaming casually, but not professionally. I don't think it will ever come close to ADSL or even ISDN's stability/speed.
Note, this is all based on my understanding of the technology and reading up on the signalling interface that MyWireless uses. I could be wrong here and there, but I don't think by much.
MyWireless makes use of UMTS TDD signalling technology, more commonly known as TD-CDMA (used hereafter).
Let's examine the signalling technology:
The CDMA means "code division multiple access", and the TD portion indicates that it is "time division duplexed".
The first issue, is the frequency used to communicate with the tower. Frequencies are licensed commodities in nearly all countries. With normal cellular comms, a device typically uses TWO frequencies: one to transmit, and one to receive. TDD signalling means that only a SINGLE frequency is ever used, which is good for operators since they then have to license less frequencies, which is cheaper.
Normally, when 2 frequencies are used there would be no contention over when to send or when to receive, but it's expensive to license. TD-CDMA uses a single frequency to transmit and receive.
Now, just consider it logically. With a single frequency, your modem and base station need some sort of agreement, on when to send, and when to receive. This is done by the "time division duplexing". Your modem and tower transmit, and receive over a single frequency at <i>predetermined time slots</i>.
You also <i>share</i> this frequency, with the other people connected to this tower. So in addition to the TDD contention for when to receive/send, another multiplexing domain - CDMA (code division multiple access) is used to arrange communications between the tower and all the modems it is serving.
The tower assigns each modem a code, and the modems encode their streams across the frequency using that code, and vice versa. The other modems talking on the same frequency with different codes, essentially look like "noise" to a device who is receiving data. The code itsself is a pattern of bits. The receiver looks for that bit pattern.
As long as the receiving device has the right code, it can pick up its conversation out from all the others. The base station, naturally does a LOT of this decoding, since it has to variably listen for each connected users' coded connection. This means the base station needs quite some horsepower, depending on how many connections there are. The CDMA portion of the air interface will introduce latency depending on the number of people connected at a time, and also just raw processing overhead.
This is what Sentech mean when they say it's a "shared service". You share the frequency to the tower with all the other people connected to it, and the more people are connected the more latent it gets.
On top of all this complexity, the specification for TD-CDMA (otherwise known as UMTS-TDD) assumes a frame length of 10ms. This is the size (length in time) of a completely encoded packet. During a "frame" there are 15 slots to either send, or receive, which is what gives UMTS-TDD good bandwidth characteristics, but unfortunately introduces the 10ms latency, seeing as a frame is only processed after the 10ms length. The scheduler at the base station will do NOTHING until the time period for the frame has passed.
So, the air interface is governed by 3 factors: Time, frequency, and code. Additionally the frame length adds a fixed latency.
Let's take a look at some rough (hypothetic) calculations for a typical PING packet, SINGLE trip, meaning we need to double the final result to get a return trip time:
This assumes a ping to say, www.is.co.za
So, the the optimistic MINIMUM travel time over the air interface, for a packet will be 20ms. That's as low as we can expect it to go. And this is assuming 100% signal probably, with no error correction or retransmissions required.
That's just a ONE-WAY trip still! Let's continue the calculation with rough optimistic, subjective, guesstimates from what I know about local internet conditions and routers.
Right, that gives us a TOTAL, single trip minimum time of 30.5ms, so let's round it to 31ms for convenience because so far I've been very optimistic. The ping return packet will also travel 31ms via this entire route, giving us a <i>bare minimum return time of 62ms</i>.
This was also a very short route, basically 3 hops, and not much else. The reality is that there are at least 10-12 hops between a Sentech user and games.saix.net which will increase latency even more.
I don't think we'll ever see a ping below 60ms. Add to this the normal overhead on internet routes, and you'll see that we're always going to end up with at least 80-120ms pings. Take into account that the number of users connected to a tower, and how much their transmitting is affecting the CDMA and we get the random "spiking" when pinging.
Bottom line, the air interface is pretty latent, on it's own. Combine it with other users, and general internet overhead and it's plain to see that ISDN, and ADSL will always win the ping battle.
Also note, that the latency I've explained here does NOT affect the available bandwidth in all situations. As we all know the actual bandwidth available via Sentech is pretty damn good.
UMTS-TDD/TD-CDMA is a pretty complicated technology, and requires a lot of signal processing. 802.11 on the other hand is an extremely simple air interface, using multiple frequencies, but naturally it doesn't SCALE as well with many users as UMTS-TDD does, due to it's simplicity.
The TD-CDMA spec is under constant evolvement, and a simple firmware upgrade for the modems, and software upgrade for the towers should allow different signalling characteristics, which may change all this, but I think IPWireless is pretty "stable" with the technology and the 1999 specification (which is what they're using), at the moment.
So, in the meantime if you want to game with your buddies, perhaps slapping up a local 802.11 network and sharing an ADSL between say 4, or 5 gamers is the best way to go in terms of price, and ping consistency.
I guess it's apparent from this post that I had a lot of spare time on my hands today.[
] I didn't pull this information out of a hat. I spent lots of time reading the specs and digging through various UMTS sites. I would love for someone from IPWireless or Sentech to comment on this, or correct me if I have oversights somewhere. I think the basic principle is pretty much close though. Give or a take a few milliseconds.
<center><h5><font color="red">Oo. MyWireless <s>Hacks</s> Tweaks & Tech Info.oO </font id="red"></h5><h6>Have you checked the fawking FAQ?</h6></center>
I've dug around quite a bit, reading some of the specs of the IPWireless and UMTS technology, and have written this explanation of why I believe that pings will never be awesome on MyWireless. Note, that when I mean awesome, I mean, good enough for gaming casually, but not professionally. I don't think it will ever come close to ADSL or even ISDN's stability/speed.
Note, this is all based on my understanding of the technology and reading up on the signalling interface that MyWireless uses. I could be wrong here and there, but I don't think by much.
MyWireless makes use of UMTS TDD signalling technology, more commonly known as TD-CDMA (used hereafter).
Let's examine the signalling technology:
The CDMA means "code division multiple access", and the TD portion indicates that it is "time division duplexed".
The first issue, is the frequency used to communicate with the tower. Frequencies are licensed commodities in nearly all countries. With normal cellular comms, a device typically uses TWO frequencies: one to transmit, and one to receive. TDD signalling means that only a SINGLE frequency is ever used, which is good for operators since they then have to license less frequencies, which is cheaper.
Normally, when 2 frequencies are used there would be no contention over when to send or when to receive, but it's expensive to license. TD-CDMA uses a single frequency to transmit and receive.
Now, just consider it logically. With a single frequency, your modem and base station need some sort of agreement, on when to send, and when to receive. This is done by the "time division duplexing". Your modem and tower transmit, and receive over a single frequency at <i>predetermined time slots</i>.
You also <i>share</i> this frequency, with the other people connected to this tower. So in addition to the TDD contention for when to receive/send, another multiplexing domain - CDMA (code division multiple access) is used to arrange communications between the tower and all the modems it is serving.
The tower assigns each modem a code, and the modems encode their streams across the frequency using that code, and vice versa. The other modems talking on the same frequency with different codes, essentially look like "noise" to a device who is receiving data. The code itsself is a pattern of bits. The receiver looks for that bit pattern.
As long as the receiving device has the right code, it can pick up its conversation out from all the others. The base station, naturally does a LOT of this decoding, since it has to variably listen for each connected users' coded connection. This means the base station needs quite some horsepower, depending on how many connections there are. The CDMA portion of the air interface will introduce latency depending on the number of people connected at a time, and also just raw processing overhead.
This is what Sentech mean when they say it's a "shared service". You share the frequency to the tower with all the other people connected to it, and the more people are connected the more latent it gets.
On top of all this complexity, the specification for TD-CDMA (otherwise known as UMTS-TDD) assumes a frame length of 10ms. This is the size (length in time) of a completely encoded packet. During a "frame" there are 15 slots to either send, or receive, which is what gives UMTS-TDD good bandwidth characteristics, but unfortunately introduces the 10ms latency, seeing as a frame is only processed after the 10ms length. The scheduler at the base station will do NOTHING until the time period for the frame has passed.
So, the air interface is governed by 3 factors: Time, frequency, and code. Additionally the frame length adds a fixed latency.
Let's take a look at some rough (hypothetic) calculations for a typical PING packet, SINGLE trip, meaning we need to double the final result to get a return trip time:
This assumes a ping to say, www.is.co.za
Code:
+ base overhead 0ms (drivers, USB cable, PPP calculations, etc)
+ air frame overhead 10ms
+ Time division (guard period) ~5ms (The guard period in the specs
depends on chip rate)
+ CDMA overhead 5ms (a minimum guessed factor of how busy the tower is,
and a pretty optimistic one at that.)
= Total AIR interface overhead 20msSo, the the optimistic MINIMUM travel time over the air interface, for a packet will be 20ms. That's as low as we can expect it to go. And this is assuming 100% signal probably, with no error correction or retransmissions required.
That's just a ONE-WAY trip still! Let's continue the calculation with rough optimistic, subjective, guesstimates from what I know about local internet conditions and routers.
Code:
+ Microwave link from base to
wherever sentech takes it: 2ms
+ Router overhead first hop: 0.5ms
+ Link overhead to Internet
Solutions 7ms (hah, RIIIIGHT...)
+ IS Router overhead 0.5ms
+ Router->Host overhead 0.5ms
= Total IP route overhead 10.5msRight, that gives us a TOTAL, single trip minimum time of 30.5ms, so let's round it to 31ms for convenience because so far I've been very optimistic. The ping return packet will also travel 31ms via this entire route, giving us a <i>bare minimum return time of 62ms</i>.
This was also a very short route, basically 3 hops, and not much else. The reality is that there are at least 10-12 hops between a Sentech user and games.saix.net which will increase latency even more.
I don't think we'll ever see a ping below 60ms. Add to this the normal overhead on internet routes, and you'll see that we're always going to end up with at least 80-120ms pings. Take into account that the number of users connected to a tower, and how much their transmitting is affecting the CDMA and we get the random "spiking" when pinging.
Bottom line, the air interface is pretty latent, on it's own. Combine it with other users, and general internet overhead and it's plain to see that ISDN, and ADSL will always win the ping battle.
Also note, that the latency I've explained here does NOT affect the available bandwidth in all situations. As we all know the actual bandwidth available via Sentech is pretty damn good.
UMTS-TDD/TD-CDMA is a pretty complicated technology, and requires a lot of signal processing. 802.11 on the other hand is an extremely simple air interface, using multiple frequencies, but naturally it doesn't SCALE as well with many users as UMTS-TDD does, due to it's simplicity.
The TD-CDMA spec is under constant evolvement, and a simple firmware upgrade for the modems, and software upgrade for the towers should allow different signalling characteristics, which may change all this, but I think IPWireless is pretty "stable" with the technology and the 1999 specification (which is what they're using), at the moment.
So, in the meantime if you want to game with your buddies, perhaps slapping up a local 802.11 network and sharing an ADSL between say 4, or 5 gamers is the best way to go in terms of price, and ping consistency.
I guess it's apparent from this post that I had a lot of spare time on my hands today.[
] I didn't pull this information out of a hat. I spent lots of time reading the specs and digging through various UMTS sites. I would love for someone from IPWireless or Sentech to comment on this, or correct me if I have oversights somewhere. I think the basic principle is pretty much close though. Give or a take a few milliseconds.<center><h5><font color="red">Oo. MyWireless <s>Hacks</s> Tweaks & Tech Info.oO </font id="red"></h5><h6>Have you checked the fawking FAQ?</h6></center>
]
]
