



|
Post by trog @ 09:09am 15/09/11 | 32 Comments
![]() F&F sources have it that Blizzard are using a new method of client/server communication that provides a secure and tamper-proof method of using the local client to handle its own calculations, while sending the data to the server for verification. The Client doesn’t require the server to respond, hence removing Latency from the equation.Such a development would be obviously great for Australians, who are presumably massively unlikely to see local servers on our shores for Diablo 3. They've included a link to a YouTube video, purportedly from an Australia, which shows lag free fast-paced combat (but as someone that doesn't play Diablo, I have no idea what is going on, so maybe some helpful people can comment to explain):
|
|
Posted 09:17am 15/9/11
Posted 09:20am 15/9/11
Posted 09:29am 15/9/11
Posted 09:32am 15/9/11
Posted 09:50am 15/9/11
Now go get me a beta key Ausgamers!
Posted 10:16am 15/9/11
Posted 10:28am 15/9/11
The risk of a more client-oriented model is cheating - so it's good they addressed that directly.
Also, if you missed the link to the D3 skills calculator, very interesting: http://us.battle.net/d3/en/age?r=%2Fd3%2Fen%2Fcalculator%2Fbarbarian
Posted 10:50am 15/9/11
Posted 10:59am 15/9/11
zzzz
Posted 11:07am 15/9/11
maybe a little game changing for RPG games that typically use TCP and have the server processing the information, but hardly a new way of doing networking for games.
Posted 11:41am 15/9/11
Posted 12:17pm 15/9/11
Posted 12:25pm 15/9/11
I have a friend-of-a-friend who is in it.
Posted 12:26pm 15/9/11
Posted 12:39pm 15/9/11
Honestly one of the biggest problems with blizzard servers was their shitty forced tunnelling around america, the data simply took the longest route between the offshore cable to the server's because from what I understood, they had a deal with the owner of that cable to save on the data bandwidth costs. Hence why wow tunnels worked to some degree, because they circumvented that allowing the internet to at least get you a quicker route.
Honestly my ping to different sides of america is 200-300, but my ping to blizz servers without a VPN connection for the better half, is 400+
If they reduced that it'd be better than any netcode.
Posted 12:41pm 15/9/11
That's not how wow tunnels work. There's almost no routing benefit.
I was surprised to find this out too.
Posted 01:39pm 15/9/11
Yeah there is, most of them have bought servers in the cities where USwest or USeast servers are held, and so you miss that crazy route that it naturally takes.
unless you're using your tunnel wrong, or instead you've used worse than normal tunnels.
Posted 02:01pm 15/9/11
Posted 02:19pm 15/9/11
Posted 02:25pm 15/9/11
Posted 02:26pm 15/9/11
Posted 02:56pm 15/9/11
This, would have explained it better but I was in a bit of a rush.
There might be some SLIGHT routing benefit if you're on a shit ISP, but its all about Nagles. You can really see this with the twin lag meter in the recent patch, your chat server connection without tunnelling isn't much worse than with, but your game data one changes dramatically.
You can get almost the same benefit by running a SOCKS server on your local network and routing through that, because SOCKS disables Nagles.
This is cos US data is cheap, and because routing from USWest to a USWest server via USEast would be dumb.
Posted 02:52pm 15/9/11
Posted 12:12am 16/9/11
Since the server is verifying, there HAS to be a way to "undo" a client's packet. Possibly they just DC the whole user and rollback to before the error was found. But if not I really don't want to be the guy who has to do a solid implimentation.
We obviously also have to get info about the other players and loot - so these elements will hit us with the standard "haha aussies" penalty. There is *NO* way they'll let loot generation be client side. Assuming a 200 ping, we still have a 400ms (+reaction time) round trip time to see and then pick up loot (server sends packet, you react, you send packet) vs an american with say, 20ms RTT. (10 to see it, reaction and then 10 to send).
I'm also pretty scared of what could happen if they start trying to client side predict the other players. The edges around this will be much rougher than anything that could be realistically done for the 'current' player.
All in all, glad they're trying to do something, because netcode has been getting worse and worse over the years. But I'm going to wait until someone with actual judgement and the ability to pick subtle details gives it a thorough analysis. (A random guy in F+F who may or may not believe 30 FPS is the most you can notice is not such a person.)
Posted 03:31am 16/9/11
From what I read, this isn't speeding up transactions, instead it's just lessening the amount. In otherwords the bandwidth will be lowered (cheaper).
Is it going to be faster because there will be less packets in some kind of queue? I thought they already had been running something like UDP over TCP style implementation which reduced those kinds of things.
Overall, from the limited reading it sounds like people on dial-up or mobile connections should be able to play because there shouldn't be a large need for data bandwidth as it's been minimised.
Posted 03:51am 16/9/11
Shouldn't be a big problem though, cos if I remember correctly loot is personalised now? So like, it doesn't just spew everything out and its a mad grab for loot, only you can see your loot and nobody else can steal it. I think.
Posted 07:01am 16/9/11
Yup. Also, your own magic find only affects your drops so it has a few more implications as well.
Posted 10:38am 16/9/11
In related news, my friend (melb fringes) is apparently in F+F and he says it runs pretty well. He usually whinges about frames/lag/etc worse than I do, so sounds like D3 is going to turn out solid in that department.
Posted 11:37am 16/9/11
Posted 12:33pm 16/9/11
better watch out. bethesda might sue you for having scrolls in your site name.
Posted 01:01pm 16/9/11
On the other hand, using a proxy tunnel usually eliminated a particularly bad hop in the route my game packets were taking to reach the destination server. In most cases it was an oversubscribed link between one of the peering ISPs in between me and the datacentre my game server was hosted from that was eliminated, and which reduced both my latency and packet loss.
For some games I was able to eliminate packet loss by using my work VPN for the same reason, although it did add about 35 latency on top of the usual trip time.
The thing with those peering links is that they are often not directly related (at least, from an organisational perspective) to your own ISP, so it can be hard to find a support path to get the link fixed.
Posted 01:08pm 17/9/11