Google Cloud update
13 Aug 2016 22:27:15 UTC
Topic 199994
(moderation:
Hey guys,
I'm always looking to do more science, so when I heard that you could create a VM on Google Cloud for free and crunch some data, I jumped on it.
I created a Compute Engine with 8 vCPU's and 7.2Gb of Ram. I am running Ubuntu 16.04. I added Boinc with the command line, and also added E@H and linked in my account. It isn't particularly fast, but it runs 8 cpu tasks concurrently. In 3 days, I have added 51282 credit, with a lot of work pending. I think it could be optimized to run faster, but I'm newish to linux as it is.
I would be happy to help anyone who wants to start one but doesn't necessarily know linux that well. You get 2 months for free/$300 of computing, and although you need to enter a credit card, It won't charge you unless you choose to upgrade. I'm fairly certain the credit card is to prevent abuse of the trial by having some way to limit the number of free trials you can set up.
Crunch on!
Jordan
I'd love to learn how to do
)
I'd love to learn how to do this. How do you even start with this? Could be a great way to learn how to set up cloud computing. I've never been successful in setting that up... Let me know!
sounds great, but here no
)
sounds great, but here no credit and no credit card :D
Well i tried to get a VPS for free but failed. Still searching...
A debit card would work
)
A debit card would work Rantanplan.
Adam Socki, The General way to start would be to go to cloud.google.com, get to the free trial screen (click try it free). login to your google account/ make one, then follow the steps to get your free trial. they won't charge you unless you choose to upgrade at the end of your 60 days. Then create a compute engine with Linux 16.04, 8 vCpu's and 7.2gb of ram. Once that's done (might take a few min) you should be able to click on the upper left side of the screen on the three bars drop down menu, go to your compute engine and in VM instances page there should be an instance. on the right there should be an SSH button next to 3 verticle dots. click the dots and select open in new tab. this will open an ssh window into your computer. I'll type more in the morning...
Jordan
Once you have SSH'd into your
)
Once you have SSH'd into your VM, Type in "sudo apt-get update", It will ask for your password, when you type it in nothing will show up on the screen. this is normal. assuming you entered it correctly a whole bunch of stuff will start filling the screen. it's connecting to the Linux servers and updating things in the system. it probably won't need it but its always a good idea to be updated.
Then type in "sudo apt-get install boinc-client" it should then fetch the boinc program from Berkley and install it.
Then, you can use "boinccmd" in the terminal. A good reference page is https://boinc.berkeley.edu/wiki/Boinccmd_tool
If you use an account manager, you can attach it with "boinccmd --join_acct_mgr (url of manager, http://bam.boincstats.com/) and then either (Username and Password - warning, this can be temperamental if not entered exactly right, it won't hurt anything but can be frustrating) or (weak authenticator)
I would try username and password, if it doesn't work try it in quotes.
If you don't use an account manager, then you will need to attach to the project with "boinccmd --lookup_account (URL, for instance http://einstein.phys.uwm.edu/) then your project "email" and "password"
This will give you your account key
Then enter "boinccmd --project_attach (URL) (account_key)"
then it should be attached. at this point you should be able to essentially ignore it for the next 60 days.
if, however you feel the need to have more input, try other boinccmd lines, like "boinccmd --project (url) update" or "boinccmd --get_tasks"
Thanks for donating to science. If you have any questions, just ask and I will do my best to help out. Also don't forget Google isn't just a cloud VM to compute data, there is alot of cool info out there.
Jordan
BTW guys, this is my cloud
)
BTW guys, this is my cloud computer https://einsteinathome.org/host/12380842
Jordan
Jordan, thank you very much
)
Jordan, thank you very much for sharing that absolutely interesting info!
I don't really know how to use Unix, but I managed to get my Google VM running and crunching
One of the distros available was openSUSE Leap 42.1. Could it be slightly "lighter" and give some microcosmic advantage in processing speed? Or do the different distros give same results in those virtual machines? Ubuntu felt safer to create after all, because I had instructions here... and the commands worked like charm. Thanks!
I have had several Google
)
I have had several Google compute engines and I think we can not do much to make them running faster. I was very satisfied with their crunching speed.
I'm relatively new to Linux
)
I'm relatively new to Linux myself, so I can't speak to any other distros. And I fully agree with Sasa, The compute speed is very resonable. After all, it's not running GPU tasks.
I'm glad I could help, And Thank you for helping support science. As Bill Nye would say, "If you take the time you can, I daresay it, Change the World!"
Jordan
Jordan... if you look at the
)
Jordan... if you look at the CPU utilization graph, is your Compute Engine (VM Instance) utilizing CPU up to near 100% ? Mine keeps running only 5 tasks concurrently and CPU utilization is only 62%. I haven't figured out how to change that. There is enough free RAM available for more tasks...
edit: Now I know. The location for those hosts had to be set to 'generic'. I went to project Preferences, 'Computing' and tuned new settings for 'generic' (advanced settings) and saved changes. Then I went to Google site and opened an SSH connection to the compute engine / instance and commanded a project update. More tasks were downloaded and 3 more tasks initialized. 8 tasks are running now concurrently and CPU utilization is almost 100%.
When I had attached these compute engines (VM instances) to a project and they loaded project preferences the setting "Set used for new computers" was 'Home' at that time. I had saved a lower CPU setting for computing preferences of 'Home': "On multiprocessors, use at most: 74% of the processors" . That's why the engines were running only 5 cores/tasks. I tried to change that setting to 100% afterwards and command an update, but it didn't work. Nothing happened. I also tried changing the locations to 'School' and update, but it didn't work either.
It seems (in this case at least) the initial compute settings can be changed afterwards only, if the location for those hosts is set to 'generic'. A "Preference set" of any other location didn't go through.
A similar option also exists
)
A similar option also exists on Amazon's EC2 computing platform. (As far as I understand, the free trial on the google cloud is for businesses only outside the US.) On amazon's platform you can set up two instances (one linux, one windows) for free for a year. Installing boinc is straightforward. One thing to note is that the free options are, as one would expect, slow. I mean really slow. A Gamma-ray pulsar binary search #1 task on the linux machine takes 46 hours, giving an average credit of 340. Overall, it contributes as much as a raspberry pi 2. But then it will run for free for a year. The windows machine is even slower when searching for gamma ray pulsars, but it was faster than the linux for gravitational wave tasks. And if you need a performance boost, there is always the option to pay for a fast instance...