Questions about Repeater

[addison]

Hello everyone.

This forum is really great!
Well, I have studied Ultra, Chunk and Repeater for two weeks.

Firstly, I download Repeater's source code. Then compiled it.
Actually, I have given both window's version and Linux's one a try.

I've noticed that in the windows one, it defines the MAX_LIST size as 20.
For the Linux one, its default value is 100, while the author comments that:
Maximum number of active session can be [1...1000] at the same time.

What's the 'active session' means?
Does it means active IDs?

If the answer is YES, I'm really confused about Repeater.
Because, one ID stores in the repeater will ony take 1K memory.
We can add additional Memory if we need large numbers.
So why Repeater limit the connection numbers?
If anyone knows any other reasons, please share.

BTW, In terms of the performance, Repeater Linux version is faster than windows one. How could Linux done like that?


Thanks in advance.
addison

[supercoe]

What's the 'active session' means?
Does it means active IDs?

I've never maxed out even the Windows repeater but AFAIK the answer is yes.


So why Repeater limit the connection numbers?

Most likely for bandwidth reasons.


BTW, In terms of the performance, Repeater Linux version is faster than windows one.

Both were tested from the same connection/physical location? They seem equal to me in terms of performance...

[addison]

A big thank you to you!

Sorry for asking the questions directly.

I assume that you have your own server which is running Repeater, yes?
So can you say something about your server's hardware information?

I've never maxed out even the Windows repeater.

May I ask you that, what's the maximum connection numbers in your server at the same time?
If I have 100K computers need to connect Repeater at the same time, What about the hardware requirements?
And also as you mentioned the 'bandwidth' issues?

Both were tested from the same connection/physical location?

Laboratory testing in our college, the answer would be yes.
The repeater windows version is download from sourceforge, while the linux one is from this URL http://koti.mbnet.fi/jtko/uvncrepeater/
Are we using the same one?

Thanks in advance.
addison

[B]

Holy feces on a stick. You want to put 100,000 connections on a <b>single</b> repeater? I have a feeling that's going to be impossible. But I don't really know what I'm talking about here.

For performance predictions I think you're going to need a <b>lot</b> more data than a few posters here can tell you. Most of us support no more than a few dozen connections, at most. You should probably model your setup with a test bed of at least 100 stations, and see what kind of performance you get. The repeater's a very simple packet relay, from what I understand, but you're talking about a pretty large scale. I suspect you're going to need an array of repeaters. Most flexibly, you could run them in VMs locally and/or in a "cloud" service.

I would also suggest doing a quick needs analysis -- do all these stations really need to be connected via a repeater, or can you just access them directly?

Anyway, just some suggestions. Good luck.

[shadowfax]

Just one little note... An active session is, or should be, a viewer and a server communicating with each other where an active ID could be a server or a viewer waiting to pair with the other end of the connection.

Let me explain....

Although the MAX_LIST is defined to a value of 20, there are two arrays (One for servers and another one for viewers), thus making a total of 40 available incoming connection slots. If you connect 20 viewers with IDs 1001 up to 1020 and 20 servers with IDs 1021 up to 1040, you would have 40 active IDs (each waiting for a pairing connection). However, there would be no active session. In the best scenario, where the ratio is 1:1, meaning 20 viewer with the same id as the 20 servers, there would be a maximum of 20 active sessions.

[B]

But the big question for the OP is what is a reasonable MAX_LIST size based on a typical server scenario so far as RAM, disk, CPU, and network utilization.

I'm assuming that, memory aside, the size of those arrays doesn't matter nearly as much as the real active session count?