Quote from: mdude77 on January 19, 2014, 01:29:34 AM

Quote from: gyverlb on January 19, 2014, 01:03:33 AM

Quote from: mdude77 on January 18, 2014, 11:52:05 PM

I've never seen anyone say good things about QoS. In fact, I've seen lots of people say turn it off, it hoses things, and I've seen it here in this thread.

I really don't care about what people say but about what is doable. I've configured QoS on my Linux server, it works, end of story. If people use QoS on their Linux server and it doesn't work for them I can provide tips.

That might be the difference. The average person isn't going to use QoS on a Linux server. The average person is going to have a cheap router that touts QoS, but the reality is it doesn't work well.

Quote from: gyverlb on January 19, 2014, 01:03:33 AM

Quote from: mdude77 on January 18, 2014, 11:52:05 PM

Who doesn't use the bandwidth for anything else?

You misunderstood. QoS allows bandwidth for other traffic but for it to work you must have some headroom, if you need all of your available bandwidth for other traffics you can't use p2pool, if you only need what's above 256-512kbps in both directions then you can fit p2pool and bitcoind in with various levels of efficiency.

I know what QoS is. I have theoretical 768k up. If I need p2pool and bitcoin to use 2/3's of that, I'm in trouble.

Quote

You don't have to ask here, the information is already public. Go to p2pool.info and you'll have data for a theoretical 100% efficient p2pool node.

I don't care what theoretical is. I want to know what people are really getting!

Quote

Given that you mentioned a cube I suppose you have an ASICMiner cube. Unless they have a different miner implementation than the blades you are probably out of luck on p2pool. IIRC the blade miner could only do getwork and poorly: it didn't work well on several pools including p2pool because of their implementation. If this didn't change you are probably better off letting the beast mine on whatever pools you could find that worked fine with it.

If I was to attempt it I'd run it through a stratum proxy, as I'm doing now. But as I've implied a few times, there's no way I'm pointing 38gh/s at 1m share difficulty.

M

I would have to say if you have a good switch and aren't using a mickey mouse home one you should be good to go. Home managed switches have very low forwarding rates thus they can't move higher volumes of traffic over them compared to the enterprise class managed switches. So if you a good switch and are using good cabling, I would certainly enable jumbo frames before messing with the QOS. My firewall, Dell PowerConnect 6224 and my openSUSE (p2pool server) have jumbo frames enabled.