Torture time: Try to Bring the SBC TurtleCoin Pool Down With Lots of Hash Rate
In the earlier post, I promised to share with you how we tried to bring the OrangePI One Plus board down with lots of TurtleCoin Pool hash rate.
This test was triggered by two factors.
The first factor was my own curiosity. I had a conservative guess on how much hash rate a single board computer could potentially sustain. I wanted to test my theory.
The second factor was a few remarks in our lovely TurtleCoin community. “Ahh, so you will be the new kind of pool operators. Look, my seventh miner connected, we reached 7kh/s hash rate, pool is crashing”. If you know the potential of ARM-based single-board computers nowadays, you want to show to the world how wrong such view is 🙂
It was a lovely early evening mid-of April when I announced on the TurtleCoin Discord, okey, bring the pool down. I pay out everyone 10K TRTL who participates in the mining tonight and floods us with some serious hash rate.
As you can see in the screenshot below, our community was up for this challenge and soon we reached 1.9 MH/s on the cryptonight-turtle algo.
You read well, it is not a typo. We reached 1.9 MH/s. The board was still performing well. One of our fellow miners and friend who provided 99% of this hash rate kept us flooded for almost 14 hours.
As you can see in the pool frontend (as well as in the following screenshot), that day (16th April) the pool mined at the 1.8 – 1.9 MH/s hash rate 13 blocks in 14 hours, almost 1 block per hour.
More importantly, not a single block orphaned. This result exceeded all of our expectations. Why?
In order to make it very hard for the SBC to cope with this much hash rate, we decided to install and run all components on one board: the TurtleCoin node itself, the pool software consisting of payment, api, pool and stats workers, and the Redis DB.
The only component we hosted on a separate hardware was the actual HTML site for the pool front-end (however, the statistics were still provided by the API provided by the SBC).
The CPU cores oscilated on average between 30-40%. The main load was caused by the TurtleCoin daemon itself, not by the hashing load sent by the miners to the pool.
You can see stats from the HTOP in the screenshot above. Please note the Load Average in the top right shows wrong values, the bars indicators and CPU / MEM percentages in the left pane are correct. This was a known bug in the OrangePI Ubuntu build at the time we performed this test and was fixed in the later builds.
How did the board do in our view compared to our expectations?
First, we did not expect the board to be able to cope with almost 2 MH/s hash rate. We have estimated the tipping point around 1 – 1.1 MH/s. This was a guess, and it was a pretty bad one.
Second, despite the board coping so well with this much hash rate, we expected it to crash after a few minutes up to an hour. This has not happened. In fact, the board never crashed during the test, no blocks orphaned, and the uptime of the pool as of today is:
root@OrangePi:~# uptime 21:00:54 up 111 days
As @thinkpol summarized it in the Discord chat: “You achieved everything with this test, except for one thing – you did not manage to bring the board down.”
This kind of summarizes for us the potential, the stability and sturdiness of ARM-based single board computers. These do have a proven place in blockchain, crypto and mining applications and our test proved exactly that.
We want to hear from you! Did you took an SBC even further in torturing it in the crypto world? We are curious, let us know please!