1) Message boards : Number crunching : Use the (free trial) $300 credit from GoogleCloudCompute for your Universe@home and other BOINC endeavors (Message 1092)
Posted 11 Mar 2016 by Matt Kowal
Post:
Google offers a $300, 60 day trial period for their cloud computing platform. The platform allows you to create Linux virtual machines that run on Google's cloud infrastructure.

Offer requires a Google account and an active credit card. When the free trial credit/period ends the instances are paused and you are prompted to upgrade to continue services. Accounts are only charged if the user opts to upgrade to paid service. I successfully ran (5) 2xhighcpu instances for about a month until my credit expired. To avoid any potential downtime, I'd recommend attaching two projects to each instance. I guess I should note that I used my trial before adding Universe@home as a project, so my processing power went elsewhere.

The vanilla link : https://cloud.google.com/free-trial/

1 - go to https://cloud.google.com/free-trial/. Sign in with your Google account and complete the registrations necessary to 'Start your free trial'

2 - create a new compute engine project

3 - create a a new VM instance

4 - create a 'high CPU' instance using the debian or ubuntu image

5 - once Google has created the instance, click on it in the project management console. Select 'Open in browser window' from the SSH drop down to start interacting with the VM instance.

6 - you will see a command prompt where you can begin entering the commands in step 8

7 - for each of the BOINC projects that you want to run in the VM, log in to your account for the project and click on 'Account Keys'. (It is assumed that you already have BOINC accounts created. Creating BOINC project accounts is beyond the scope of this tutorial)

8 - make note of the URL and account key shown on the account keys page

9 - in the SSH terminal enter the following commands:

$ sudo apt-get update
$ sudo apt-get install boinc

for each project that you want to run, enter:
$ boinccmd --project_attach PROJECT_URL AUTH_KEY
replacing PROJECT_URL and AUTH_KEY with the values that you gathered in step 8

$ boinccmd --set_run_mode always
$ boinccmd --set_network_mode always

The following commands are useful for monitoring the state of BOINC:
$ boinccmd --get_simple_gui_info
$ boinccmd --get_tasks
$ boinccmd --get_file_transfers

10 - once BOINC has connected with the servers and downloaded tasks, you should see both CPU cores running near 100%

$ top

11 - your instance has been successfully created and configured. You will see the CPU graph in the project console running at 100%.

12 - looking at the pricing, the instance will cost approximately $0.06 per hour. In order to use up the $300 trial credit within the 60 day trial period, we can run 4 instances. ($0.06/hr * 24 hr * 60 days = $86 per instance for 60 days)

13 - copy your instance and repeat steps 4 through 11 for each instance that you create.

14 - please note that the above pricing model may be outdated. I was able to create 5 instances that collectively exhausted the free credit in about 4 weeks. I was not charged anything.

Happy Computing
2) Message boards : Number crunching : Gpu App? (Message 1044)
Posted 27 Jan 2016 by Matt Kowal
Post:
I would also be interested in running a GPU application.

I send positive thoughts.
3) Message boards : Number crunching : Application for ARM based Linux devices (Message 1018)
Posted 17 Jan 2016 by Matt Kowal
Post:
Running well on my Raspberry Pi 2.
Overclocking to 1050 cpu, 525 cache, 483 mem took my WU times from ~78,000sec to 56,000sec.
Stock is 900mhz,250mhz,450mhz