What Is Effective Capacity, And How To Improve It?

I’ve been mining for a few days now 24/7 on my ODROID-XU4 (article coming soon) setup with two 8TB drives and my Effective Capacity has been knocking around a little bit between 12TB and 14TB. That means that I’m earning between 75-87.5% of my potential BURST from my historical contribution.

Effective Capacity

Historical share is calculated as:

 historicalShare = minerPlotSize / poolPlotSize

So you can see from this, the larger your minerPlotSize, the greater the payout. Thus it’s essential to optimize this as much as possible.

I’ve been mining with a targetDeadline of 16515036. cryptoguru.org has a calculator to determine your starting targetDeadline with a formula of targetDL = 720 * netDifficulty / yourPlotSize. Keep in mind that while your plot size may remain constant, the network difficulty moves around from day to day. Today’s netDiff is 345,103 on their site, so I’m going to try with 350,000 to see if that makes a difference. That gives me a Target Deadline Limit of 15750000 (6m 2d 07:00:00). A little lower than I was using before.

It will take a full 360 blocks to fully take effect. The system aims for around 1 block every 240 seconds, and there are 86,400 seconds in a day, so 360 blocks is a full 24 hours. I’ll let that sit for at least 24 hours and come back with my findings!

How the Estimated Plot Size Is Determined

Say the miner got nConf best confirmed deadlines in the last nAvg rounds (only the blocks with block time less than TMin are used for the estimate). To turn every best DL into a value that has a clear relation with the miners plot size, the DLs are divided by the Difficulty of the round they were confirmed in.

target_i = Deadline_i / Difficulty_i

Those characteristic values target_i are summed up to get the totalTarget. With this the estimated effective plot size (EEPS) is calculated:

EEPS = alpha * 240 * (nConf-1) / totalTarget

The constant 240 is the mean block time (in seconds) that the burst network aims to achieve. The parameter alpha is a bias that always assumes that the miner withheld the (nAvg-nConf) of his worst target values if the miner did not submit DLs in every of the last nAvg rounds. This ensures that miners cannot trick the pool into systematically overestimating the miners plot size. alpha is calculated as:

alpha = 1 - (nAvg-nConf) / nConf * Ln[ nAvg / (nAvg-nConf) ]

This is taken directly from cryptoguru.org

Leave a Reply

Your email address will not be published.