Show Posts
Hello! While folding on GPU hardware is a great way to earn Folding@Home points for protein research, if you do not have GPUs, or want to do a lot of folding, EC2 can be an alternate option to rack up those folding points.
Make sure you sign up for a folding@home account and an account at cryptobullionpools.com. You won't get QRB (Quick Return Bonus) for BigAdv WUs until your account has submitted 10 WUs successfully. I recommend you use the google NACL extension to quickly get 10 WUs before you start your EC2 work if you are using a new folding@home account. Make sure your account at cryptobullionpools and at folding@home are IDENTICAL.
Note: there is an above-normal risk for using EC2 on Curecoin. Curecoin EC2 folding is most profitable with Bigadv WUs, which can take up to nearly 24 hours to complete. Unlike mining where shares are submitted multiple times per minute, a partially-completed WU is worthless. If your server is destroyed before a WU is finished, you get no credit for that WU. Due to the nature of EC2 spot pricing, your servers can be bought out from under you. If you were 80% done with a WU and it costed you $6 of computing power to get to that point, then that server is out-bid from you, you lost that $6. You were awarded no points by Stanford, and thus no CURE by the payout pool for folding.
How does EC2 pricing work?
EC2 spot pricing is based on an auction-style bid system. People put a maximum price they are willing to pay per hour for a server of a certain type. If the current spot price is lower than that, they get a server provisioned. The second the bid price exceeds the price they have set for the maximum price they will pay, the server is decomissioned and redeployed, losing any data you had on the system. As a result, you can put a higher bid price to have a higher chance of keeping that system through price increases. Make sure you would be actually willing to pay the price you put in, though, as you may end up paying that. You will never pay more per hour than your maximum bid price.
How does EC2 work?
EC2 allows you to pay for server space by the hour. You can rent a server, fold on it until it completes a bigadv WU and returns it, and then cancel the server, paying for only the hours that the server belonged to you for.
Why spot pricing?
Spot pricing is significantly cheaper than on-demand instances. Spot pricing for a c3.8xlarge (16 cores, 32 threads) is currently around $0.26/hr/server in Oregon. On-demand (guaranteed no shutdown) are $1.68/hr/server.
What is Paravirtualization? What is HVM? Which should I choose?
If you want maximum performance (an additional ~1 or 2%, nothing significant), go with Paravirtualization of Ubuntu 14.04. However, in the very likely chance that you want to create one server and then duplicate it, HVM is for you, as you can easily create and deploy images without having to mess with S3. HVM still performs extremely well.
Getting Started:
I will publish a more comprehensive guide tomorrow complete with images and long-winded explanations. However, for now, a bit of knowledge about EC2 is assumed. You can find tutorials for this kind of stuff all over the forum.
--> Decide how much you want to spend per day, in USD. Remember the above risk. Divide this number by 24 to get the hourly rate you wish to spend. Divide this number by the number of servers you want to run to see what your price per hour per server should be. Currently, in Oregon and North Virginia, a c3.8xlarge instance (the best one) costs around $0.26 per hour. However, setting the price somewhere in the $0.35 to $0.42 range will make it far less likely that a spike in price will cause you to lose your server, and thus your investment back to the start of the current WU.
--> Click on "Request Spot Instance", and choose Ubuntu 14.04. If you are unsure which option to pick, go with HVM.
--> Click next, sort by ECU, and then choose the c3.8xlarge option (which has 108 ECUs)
--> Next, enter the price you calculated above. Make sure it is not lower than the price you see listed in the above boxes. In most cases, you want to set it high enough that small spikes in price will not cause you to lose your server (s).
--> If you need to, configure security and a private key file. Download this file, use PuTTYgen to convert it to a .ppk, then load it into PuTTY as a private key file for authentication. Google is your friend. Tomorrow, I will have pictures, maybe a video.
-->Finish, and request your spot instance(s). Wait for them to be fulfilled (1 minute to 10 minutes).
-->Go to your Instances page, and copy the IP address for each one into PuTTY, making sure your .ppk file is chosen for Auth options. The username is 'ubuntu'.
--> Once you SSH into your server:
In config.xml, use the arrow keys to navigate down to fill out your username and passkey. Double-check this! Control + X to exit nano, press 'y' and enter to save.
Now, you can monitor progress by typing:
Press Control + V to scroll down to the bottom, where you will find the current percentage.
If you want to turn your server off, finish your current WU, otherwise your account loses that WU:
You can open the program 'htop' to monitor CPU usage.
this is pointless now, it could damage the whole folding operation, i thought the point was to submit more WU?Make sure you sign up for a folding@home account and an account at cryptobullionpools.com. You won't get QRB (Quick Return Bonus) for BigAdv WUs until your account has submitted 10 WUs successfully. I recommend you use the google NACL extension to quickly get 10 WUs before you start your EC2 work if you are using a new folding@home account. Make sure your account at cryptobullionpools and at folding@home are IDENTICAL.
Note: there is an above-normal risk for using EC2 on Curecoin. Curecoin EC2 folding is most profitable with Bigadv WUs, which can take up to nearly 24 hours to complete. Unlike mining where shares are submitted multiple times per minute, a partially-completed WU is worthless. If your server is destroyed before a WU is finished, you get no credit for that WU. Due to the nature of EC2 spot pricing, your servers can be bought out from under you. If you were 80% done with a WU and it costed you $6 of computing power to get to that point, then that server is out-bid from you, you lost that $6. You were awarded no points by Stanford, and thus no CURE by the payout pool for folding.
How does EC2 pricing work?
EC2 spot pricing is based on an auction-style bid system. People put a maximum price they are willing to pay per hour for a server of a certain type. If the current spot price is lower than that, they get a server provisioned. The second the bid price exceeds the price they have set for the maximum price they will pay, the server is decomissioned and redeployed, losing any data you had on the system. As a result, you can put a higher bid price to have a higher chance of keeping that system through price increases. Make sure you would be actually willing to pay the price you put in, though, as you may end up paying that. You will never pay more per hour than your maximum bid price.
How does EC2 work?
EC2 allows you to pay for server space by the hour. You can rent a server, fold on it until it completes a bigadv WU and returns it, and then cancel the server, paying for only the hours that the server belonged to you for.
Why spot pricing?
Spot pricing is significantly cheaper than on-demand instances. Spot pricing for a c3.8xlarge (16 cores, 32 threads) is currently around $0.26/hr/server in Oregon. On-demand (guaranteed no shutdown) are $1.68/hr/server.
What is Paravirtualization? What is HVM? Which should I choose?
If you want maximum performance (an additional ~1 or 2%, nothing significant), go with Paravirtualization of Ubuntu 14.04. However, in the very likely chance that you want to create one server and then duplicate it, HVM is for you, as you can easily create and deploy images without having to mess with S3. HVM still performs extremely well.
Getting Started:
I will publish a more comprehensive guide tomorrow complete with images and long-winded explanations. However, for now, a bit of knowledge about EC2 is assumed. You can find tutorials for this kind of stuff all over the forum.
--> Decide how much you want to spend per day, in USD. Remember the above risk. Divide this number by 24 to get the hourly rate you wish to spend. Divide this number by the number of servers you want to run to see what your price per hour per server should be. Currently, in Oregon and North Virginia, a c3.8xlarge instance (the best one) costs around $0.26 per hour. However, setting the price somewhere in the $0.35 to $0.42 range will make it far less likely that a spike in price will cause you to lose your server, and thus your investment back to the start of the current WU.
--> Click on "Request Spot Instance", and choose Ubuntu 14.04. If you are unsure which option to pick, go with HVM.
--> Click next, sort by ECU, and then choose the c3.8xlarge option (which has 108 ECUs)
--> Next, enter the price you calculated above. Make sure it is not lower than the price you see listed in the above boxes. In most cases, you want to set it high enough that small spikes in price will not cause you to lose your server (s).
--> If you need to, configure security and a private key file. Download this file, use PuTTYgen to convert it to a .ppk, then load it into PuTTY as a private key file for authentication. Google is your friend. Tomorrow, I will have pictures, maybe a video.
-->Finish, and request your spot instance(s). Wait for them to be fulfilled (1 minute to 10 minutes).
-->Go to your Instances page, and copy the IP address for each one into PuTTY, making sure your .ppk file is chosen for Auth options. The username is 'ubuntu'.
--> Once you SSH into your server:
Code:
sudo -i
apt-get update
apt-get install htop
mkdir /etc/fahclient
apt-get install nano
wget 1.curecoinmirror.com/config.xml -O /etc/fahclient/config.xml
nano config.xml
In config.xml, use the arrow keys to navigate down to fill out your username and passkey. Double-check this! Control + X to exit nano, press 'y' and enter to save.
Code:
wget https://fah.stanford.edu/file-releases/public/release/fahclient/debian-testing-64bit/v7.4/fahclient_7.4.4_amd64.deb
sudo dpkg -i --force-depends fahclient_7.4.4_amd64.deb
Now, you can monitor progress by typing:
Code:
nano /var/lib/fahclient/log.txt
Press Control + V to scroll down to the bottom, where you will find the current percentage.
If you want to turn your server off, finish your current WU, otherwise your account loses that WU:
Code:
nc 127.0.0.1 36330
finish
You can open the program 'htop' to monitor CPU usage.
With this as people compete for EC2 instances, they'll keep killing each others... and no one will profit. I'd advise everyone to not use this as it could do more harm than good.
Also another announcement either later tonight or tomorrow! 
