Status as of Monday, September 28

Post by **Steve Sokolowski** » Mon Sep 28, 2015 9:15 am

Good morning. Here's what happened over the weekend:

The early part of the weekend was spent finishing up with performance improvements to the database server. We were able to eliminate cascading triggers so that we did not have to re-SELECT the parent table's row in the child trigger. That got the time to insert a share down to 1.1ms, which we determined is the fastest we can get it until Postgres 9.5 is released. There should be plenty of capacity for a while given these improvements. I released this earlier in the morning, which is why there was a short mining server outage.
The mining server was updated to weight the orphaned block penalty by time. Before, the average number of orphans over the last 28 days was used. Now, recent orphans receive twice as much weight, and orphans 27 days ago receive almost no weight, on a declining linear scale. This fix, which was just released, will more accurately compute the orphan rates for some coins, like litecoins, and will increase profitability. It means that recent network conditions will be considered more important than conditions long ago in the past. Coins that have steady orphan rates will see no change in orphan penalties under this new query.
After that, we made some changes to our trading code to account for Cryptsy's increased fees. Now, we account for these fees and assume that we will lose 0.5% of all balances we send to Cryptsy (because that's what it costs to get the money back out). The effect is that Cryptsy's buy or sell prices have to be 0.5% more favorable than the other exchanges in order for us to use them now.
Cryptsy locked a lot of wallets over the weekend, so Chris listed those coins for discontinuation. You can see that list in the "news" forum.
We released a new version of the website that is designed to address concerns brought up by some users about mobile applications. While it would be very time-consuming to completely redesign the website to have "cards" instead of tables, we were able to reduce the number of HTTP calls to the server. For example, loading the miner earnings pages now requires only 12 calls instead of 22. I'm hoping to continue these reductions and get that number down to 2 by today. Try it out and you should notice that the pages load more "at once" and with less delay, even on desktops.
We're going to go back to examining multiple algorithms at some point in October, now that the pool is consistently profitable and we have been able to automate most functions so that the number of customer support issues caused by bugs is much lower.
Since it's a Monday, we'll be doing our normal backups. It will be interesting to see how the backups proceed since this is the first week with the performance enhancements in effect. Please comment on your experience and whether it has improved.