[1200 TH] EMC: 0 Fee DGM. Anonymous PPS. US & EU servers. No Registration!

[zvs]

i noticed the IPs still respond, but can't mine anyway

208.110.68.114, 208.110.68.115, 208.110.68.116


probably because of this wholesale internet

69.30.209.1

it gets 4000ms ping time


10  wholesale-internet-inc.10gigabitethernet1-3.core1.mci2.he.net (216.66.78.90)  112.809 ms  115.204 ms  116.400 ms
11  69.30.209.1 (69.30.209.1)  3952.636 ms !H  3892.535 ms !H  3891.844 ms !H


10  wholesale-internet-inc.10gigabitethernet1-3.core1.mci2.he.net (216.66.78.90)  113.323 ms  115.929 ms  113.640 ms
11  69.30.209.1 (69.30.209.1)  114.734 ms  115.683 ms  116.227 ms


well, that's working now

there, now it's all up

[1713579274]

1713579274

[1713579274]

1713579274

[1713579274]

1713579274

[1713579274]

1713579274

[mc_lovin]

only 1 minute after i switched my power to elsewhere.

what can be done to prevent this in the future?  downtime sucks.

[goxed]

CGminer just switched back to eclipse.

 [2012-08-1X XX:41:52] Accepted fdc63791.c93704e4 GPU 0 pool 1
 [2012-08-1X XX:41:52] Pool 0 http://us2.eclipsemc.com alive
 [2012-08-1X XX:41:52] Switching to http://us2.eclipsemc.com
 [2012-08-1X XX:41:54] Accepted afe6b602.03c418bf GPU 0 pool 1

[kano]

only 1 minute after i switched my power to elsewhere.

what can be done to prevent this in the future?  downtime sucks.

Use cgminer ... properly

I have 9 pools on mine (5 + solo + 3 backups of the original 5)
It would take 5 pools to fail, for me to even notice anything but a blip.

Edit: and if you don't like the tiny switch blips when things fail, you can set load balance to share your work around the pools all the time rather than just when things fail, so the switch over blip will probably be even smaller - but the individual payouts will be proportioned from each pool, of course, since the total shares will be distributed around all the pools
(though IMO, solo included in load-balance would not be ideal, due to my small hash rate

[Inaba]

So the new FS and other boxes were suppose to be installed last week but the new DC has dropped the ball timeline wise... and the FS took a dump again tonight.

I went to swap out the hardware and the replacement board was crapped out (I tested the damned thing last week, WTH!)... so I cast a spell of ressurection on the current FS and brought it back to life... again.  I have a query out to the new DC as to what the status of the new servers are.  If they come back with something I don't like, I'll get some different replacement hardware and swap out the FS tomorrow.  Hopefully it doesn't decide to take a dirt nap again tonight until I can get stuff replaced.

I apologize for the downtime.   I'm not sure why the mining server stopped responding, but it showed it was rebooted around the same time, which is not right.  I have to investigate that further, but you should have been able to keep mining, even though the website is down. 

I will be making some changes to the DNS right now that should prevent that from happening in the future, so even if the website goes down, the mining servers should stay operational.

[arobincaron]

Thanks Inaba.  Your hard work is much appreciated here!

[Inaba]

Ok.. DNS issues should be resolved and if the web server takes a dive, the mining servers should remain functional.  I found a couple problems with the DNS and have corrected them.

[DobZombie]

I haven't been getting txt messages lately when my miner goes down. Excellent service to have.

[evanesce]

I absolutely love the text message feature Inaba.  Its so cool to get texted when a block is found.

[Inaba]

DobZombie: Send me a PM with your pool name and I can take a look.

On another note:

For you faster miners out there, I would like to request that you guys help me out testing the new Difficulty 10 server.  http://diff10.eclipsemc.com:8337  ... everything is the same, it's just a new hostname.  That server will be spitting out Difficulty 10 shares.  It seems to be working fine in my internal testing, but I would like to get some more miners on it and see how it goes.

I don't recommend it for slower miners though, as you may end up with more stales than usual.

[kano]

Hmmm - thought about that ... then started writing a reply ... then deleted the reply ... then decided, what the heck I'll try it with cgminer and see what happens
I know cgminer supports higher difficulty shares - now to see it in action!

Anyway, the causes of stales are:
1) working on a nonce range that completes after an LP notification is received
2) pools all tell the miners about the LP after the LP has occurred (of course) ... but how long after (how slow the pool is) determines how bad the stales are ... there will always be some delay, but the size of the delay will determine how bad the stales are (on average)

No miner would process a 10 difficulty nonce range (they don't exist), it would simply do multiple 1 difficulty nonce ranges and only return the ones with the required 10 difficulty (and roll-n-time helps with that drastically also)

... yeah I tried to get the nonce size changed as it should be done ... but no one liked the idea

[Inaba]

Well, up until now, the submitted work has been a small portion of the system load.  With the way I process shares, coupled with anonymous and authenticated PPS on top of DGM, submitted shares has actually become a bottleneck and reducing the amount of submitted shares could realize sizable, tangible gains for EMC.  It may not be as good for some other pools, but for EMC it's a good fit.  I've got a few other tricks up my sleeve coming down the pipe as well.

Is it working properly for you though?  Working fine for me here.  I've had 0 stales in the past 4 hours at 1400 MH/s... the LP hits and cgminer seems to just checkout new work and carry on.

[kano]

Well my BFL permanently died yesterday, so I've only been able to switch 2xIcarus to it to see what happens - so yeah with a U of ~1 a minute it may take a while to see how it's going

Edit: heh interesting - of course it's now looking for 1999.... or less

[Inaba]

So it occurred to me that a higher difficulty might affect intensity settings in cgminer. 

How does a diff10 setting affect what your intensity should be set to?  Lets say you have a 5970 at 800 MH/s... I9 is about right for that, but if it's diff 10 is that going to screw you?

[JayCoin]

So it occurred to me that a higher difficulty might affect intensity settings in cgminer. 

How does a diff10 setting affect what your intensity should be set to?  Lets say you have a 5970 at 800 MH/s... I9 is about right for that, but if it's diff 10 is that going to screw you?

When I ran p2pool with testing different share difficulties, Hashrates stayed consistant on cgminer with the same intensity setting.

[Inaba]

Yeah, hashrates will stay the same, but will it increase your stales is the question, since, if cgminer was actually doing a diff 10 nonce, you wouldn't make it all the way through.  However, as Kano has said, I guess cgminer is actually going through diff 1 nonce and taking the diff 10 shares and submitting them.  Makes perfect sense, so perhaps it doesn't matter.

It sure blows your U and /m out of the water though.

[wogaut]

Inaba,

I just noticed, I'm one of your diff10 testers 

I see how U can be higher with diff10, but overall I'm not sure why it wouldn't take longer to find a valid block at diff-10 instead diff-1.
Can you elaborate?

[kano]

Inaba,

I just noticed, I'm one of your diff10 testers 

I see how U can be higher with diff10, but overall I'm not sure why it wouldn't take longer to find a valid block at diff-10 instead diff-1.
Can you elaborate?

The time to find a block is dependent on the pool hash rate trying to find it.

Though ... I guess I should have asked if the 10diff shares are part of the same pool or not


Anyway, surprise! The BFL was dead due to USB issues ... and ... it's alive (after relegating my Icarus down to the garage and playing with USB cables for half an hour with the BFL)
OK all my hashing going to EMC 10diff now (2 rigs) to see how it goes for a while

I get multiple LP messages each LP but other than that it seems all OK
(I presume EMC does the LP like that normally?)

[Inaba]

I'm not sure if I understand what you're asking, but I'll try to answer it:

It won't take longer to find a valid block because the real difficulty is over 2 million, so whether or not you find a diff1 or diff10 (or diff1000) share won't make any difference to how often you'll find a 2 million difficulty block.  The only reason we do difficulty 1 blocks is to make accounting simple for slower miners.  Back before GPU mining and even during the era of GPU mining, if you were trying to solve a > difficulty 1 block, it's entirely possible you would not get a single share before a LP invalidated your work, so you'd get no credit.  Moving to the lower possible difficulty lets everyone from the slowest to the fastest an equal chance of submitting a share and getting credit for it.

Anyway, there seems to be something wrong with the diff10 calculations and would have ended up overpaying a few people.  I've disabled diff10 shares for the night.  I will take a look at it tomorrow when I'm not so tired and poke at it some more.  The server is still up and running, just serving diff1 shares now.  No need to change anything for anyone who's changed over.

[Inaba]

Inaba,

I just noticed, I'm one of your diff10 testers 

I see how U can be higher with diff10, but overall I'm not sure why it wouldn't take longer to find a valid block at diff-10 instead diff-1.
Can you elaborate?

The time to find a block is dependent on the pool hash rate trying to find it.

Though ... I guess I should have asked if the 10diff shares are part of the same pool or not


Anyway, surprise! The BFL was dead due to USB issues ... and ... it's alive (after relegating my Icarus down to the garage and playing with USB cables for half an hour with the BFL)
OK all my hashing going to EMC 10diff now (2 rigs) to see how it goes for a while

I get multiple LP messages each LP but other than that it seems all OK
(I presume EMC does the LP like that normally?)

Hah sorry Kano... something is slightly wrong with the DGM formula for 10diff and I'm way too tired to try to wrap my brain around what that might be.  I will have 10diff back tomorrow when I get a chance to look closer at it.

As for the LP, are you sure you're not connecting to both EMC servers, one as a backup or anything?  That would give you two of the same LP, no?  I'm not seeing multiple LPs with the same block on this end? 

Although now that I think about it, it could be cause I was working on the server and when I restart it, it sends the LPs out again in case it missed anyone during the time it went down and restarted.