A cool article i found...
Lag. Someone is always blaming the fact that they got killed on lag. Usually, this person is also pointing fingers. Or someone else points the finger. Or someone joins the room and starts talking about how person X or Y has high latency. This often can make for quite a horrid little conversation...
So I thought I'd put together some of my own thoughts (and drag in the combined knowledge and experience of my co-workers) to write up a little document on this subject as it relates to Ghost Recon. By way of explaining the background of the people that contributed, we are all working as computer analysts for a medium sized telecommunications company. The 4 of us have 35 years of technical experience and work with this sort of thing on a daily basis. And gaming networks are built from scratch at one or other of our houses about every other month, complete with Quake Servers, Proxy Servers and about a mile of CAT-5 cable. In other words, we ain't no slouches and we have pretty good handle on how this whole thing works.
Enough with the introductions. Just a few guidelines on this. First of all, for the sake of making this simple, I am not going to discuss any non-Apple based system. I am taking about Apple/OS X cause that the technology that's on top right now. Finally, we have not done much packet sniffing on this game. We have no idea the size of the packets or exactly the type of data they contain. We have got an idea of how much data is transferred and what type of information is being transferred. We just haven't busted open the packets.Now, for the show.
Latency-- the time it takes for a data packet to move across a network connection
This is one of the most common things that causes lag. It is also the most misunderstood. Ghost Recon works on the principle that the clients each run their own game with input from the Host, which coordinates the action between the clients. This means that the only latency that really counts is that between each client and the Host individually.
For the purposes of my examples in this document we will be using the following layout. Chicago is the Host. There are clients in DC and New York in the east, San Francisco and Seattle in the West, Madison(WI) in the Midwest (that's where I am) and one lonely Cannuck somewhere far north of the border.
Now, the Latency between all the clients and the host, at this moment, happens to be pretty good (200ms or less). When in the Zone the clients all see the Host as Green. What this means is that a single packet of data takes <200 milliseconds to reach the host and then return. This also means that the connections are all good and, if everyone is on a fly @$$ system, the game should move smoothly. It is important to note that latency can change at any moment. Client's connection to the Host, if the internet is being used, is dynamic. This means that the path a packet takes to reach its destination can change instantly. Routers go up and down on the iNet all the time. Now, this doesn't mean that latency is going to wildly fluctuate. It may, but it will likely remain about the same. The only parts of the connection to the host that are static are the actual phone/line connections to the internet service provider.
So, when you enter a room on GameRanger, you can see the list of players and a 'Latency Indicator'..(PING). This indicator shows YOUR latency to that individual. It does not show those people's latency to the Host. I have heard some people state that it can give an indication of a possible problem. While this is technically true, it would be more accurate to say that it occasionally gives an indication of a potential problem. Let me explain. If Chicago is hosting a game and San Francisco, Madison and DC are all playing with latency <200ms, the game is going well. New York joins. When he is in the chat room , he notices that he has a Green connection to Chicago, DC and Madison but his connection to San Francisco is red (>500ms). When he gets into the game he complains about this, thinking the Host should boot the San Fran player. In fact, this makes no difference to the game because the clients do not send data directly to one another, and that is what the Latency Indicator is measuring.
Ghost Recon is built in such a way that only the client to Host latency need be measured. All data is sent from each client to the server which then sends the data to each client. It is possible, and I have seen this, for 2 people to be unable to ping each other (due to the packet expiring in transit) but they both are getting <200ms pings to the Host. All other things being equal, the game will run just fine in this case.
Client connection type or speed to the internet makes little difference either. It is time to Host that matters. If the Client has a <200ms ping to the host with a 56k modem it will look and act the same as a T1 or ASDL with a <200ms host latency. The speed to host is what matters and the packets being sent are so small and seldom (in a computer's sense of time) that the connection speed has little impact. In other words, unless you are the host, bragging about your connection type does nothing to 'prove' you aren't experiencing lag.Besides, your connection to your ISP can be glass, but your ping to Host can still suck. Keep in mind this is speaking from the Client side. The Host is another issue all together.
The Host connection needs to have more speed than the average Client. This is because of the number of packets inbound. The server is getting data from ALL the clients constantly and then repacking this data and sending it out to all the Clients. This is as opposed to the Client sending its one packet out and getting a few back from the server. A Host that has a slow network connection needs to either limit the number of Clients or become a Client.
The Host also need to have a fast machine. The faster the better. G4 400 or higher is what I would recommend for 8 or more players. More RAM helps as well. Anyone who hosts on a connection slower than 56k is wasting their time. T1, ADSL, or Cable modem is preferred if you're hosting.
Client hardware is far more tolerant. G4 400mhz to G5 2.0ghz on up for a processor, over 256 MEGs RAM and a cable modem works so long as your ISP is pristine. Faster is better though. I recommend a 17" monitor minimum, you really need to be able to see the little visual queues.
A slow machine can cause lag if it is unable to render things as fast as it should. Usually the impact on the game that a slow machine will have is a long load time, as opposed to lag. I have played this game on a G3 266, 128 MEGs of RAM, with a good 16mb Videocard over a ISDN 128k connection and had the players agree that there was no lag----- The only complaint was the long load time- but we lived through that. MMMMM?Server side lag-hate that almost as much as Campers.
A full installation of the software helps as well. This speeds up the load time and keeps the game flowing. I've heard skuttlebutt about the problems with this in multiplay, but I've been running a full install of 1.04 since I got the game and I have had few problems.
So what can we take from all this. I'll break it down.
1. Ping to Host is latency.
2. If you ain't the Host, then your ping to other Clients means nothing.
3. Lag is in the game.
4. Load time is before the game (when the map is loading).
5. Lag is not equal to Load time.
6. Load time is not equal to lag.
7. Load time is usually hardware or installation related.
8. Lag is usually latency related.
9. Connection type does not contra-indicate latency difficulties.
10. Latency CAN change over time.
There ya go kiddies.....
