On amazon's platform you can set up two instances (one linux, one windows)
I jumped on this one too. I chose to install SUSE Linux Enterprise Server 12 SP1 (64-bit), but I haven't been able to make it install Boinc. I run "sudo zypper update" at first and it updated plenty of things and everything seem okay. Then I rebooted the instance and run "sudo zypper install boinc-client". It does some searches but then it says:
Loading repository data...
Reading installed packages...
'boinc-client' not found in package names. Trying capabilities.
No provider of 'boinc-client' found.
Resolving package dependencies...
Nothing to do.
How could I make it install boinc? I tried also "boinc" and "Boinc-Client". I see someone in internet has been running Boinc with SUSE Linux Enterprise Server 11. Isn't that possible (in an easy way, for a linux-newbie) with this version 12?
I wonder if Ubuntu Server 14.04 LTS could find Boinc. Should I have chosen that distro? Or their own Amazon Linux...
If Boinc can't run on this one, I'll propably just delete this instance and see if there was a possibility to launch a different one for free anymore.
I've tried both Googles Compute and Amazon's cloud services running FAH. I had chosen the same, max CPU cores. The amazon option wasn't worth the time. Very low productivity and I assume the same would translate here. The google compute performed roughly like a Sandy/Ivy CPU of similar cores/speed.
@mmonnin, the point here is to simply contribute more resources to science. I am fully aware it's not much better than a similar computer. Think of it more as adding another computer to crunch for science for free for a month. it is, in fact, just 8 cores running at 2.3ghz. If you don't think its worth the time, then don't do it. but don't come in here saying its not worth other peoples time to help contribute as much as they can. yea, maybe it's not 100,000 RAC but it Definitely Contributes to the Research.
BTW, it only took me (pretty new to linux, including figuring out how to set it up), about an hour. If an hour of your time isn't worth helping science, then you and I have very different views on why we participate in boinc.
@Richie, I have no experiance with SUSE so I'm of no help here, I have only used ubuntu and it was fairly easy.
I jumped on this one too. I chose to install SUSE Linux Enterprise Server 12 SP1 (64-bit), but I haven't been able to make it install Boinc. I run "sudo zypper update" at first and it updated plenty of things and everything seem okay. Then I rebooted the instance and run "sudo zypper install boinc-client". It does some searches but then it says:
Loading repository data...
Reading installed packages...
'boinc-client' not found in package names. Trying capabilities.
No provider of 'boinc-client' found.
Resolving package dependencies...
Nothing to do.
How could I make it install boinc? I tried also "boinc" and "Boinc-Client". I see someone in internet has been running Boinc with SUSE Linux Enterprise Server 11. Isn't that possible (in an easy way, for a linux-newbie) with this version 12?
I wonder if Ubuntu Server 14.04 LTS could find Boinc. Should I have chosen that distro? Or their own Amazon Linux...
If Boinc can't run on this one, I'll propably just delete this instance and see if there was a possibility to launch a different one for free anymore.
@Richie, your problem is that SUSE Enterprise 12 does not include the boinc-client in its own repository. You could 1) download it manually and install it from http://boinc.berkeley.edu/download.php, but it may require other software dependencies that you have not installed, 2) try to find a "contrib" or 3rd party repository for SUSE that includes the boinc-client software; frankly I don't know where to find one, or 3) install a Linux distribution that does include boinc-client in its list of software packages.
I would highly recommend #3. I believe Ubuntu 14 LTS (trusty) does include the boinc-client software.
I didn't get back earlier, but with Yahoo platform I ended up terminating the SUSE instance a few days ago (I just didn't have the skills to install BOINC on it).
I launched an Ubuntu Server 14.04 LTS instance instead. Then I succesfully managed to upgrade it to 16.04.1 LTS and installed BOINC (client only). It runs great, no problems at all. Today I upgraded the distribution to a yet newer internal version... from 4.4.0-34 to 4.4.0-36, just for fun. This has been interesting and I think I've learned at least something while fiddling with these
I run Albert tasks on these. Here are my Yahoo instances:
They crunch Gamma-ray pulsar search #4 v1.17 tasks. After the initial start, Windows instance seems to complete one task in ~34h and for Linux it takes ~43h. First tasks were a little bit faster, because instance starts with 100% CPU cycle time, but then it slows down. After about half an hour the initial CPU credits are used. Then it continues working at base level ( ~10% speed). Pauses and Idle time would slowly load up the CPU credits again, but that isn't happening while crunching.
For that Windows instance I use BOINC Manager and set BOINC to use only 50% of the CPU time. That should be more than enough. I thought maybe that would smooth out the competition between constant credit usage and "trying to give a birth to new credits". At the Yahoo EC2 Management Console Monitoring I can see now the graphs of CPU utilization, CPU credit usage and CPU credit balance have settled in a bit different way for both instances. Anyway, I believe both are running "at full speed".
I updated also My Google Cloud instances to Ubuntu version 4.4.0-36 today.. Their potential is at completely different level and they've been crunching like crazy pigs for a few days! One Gamma-ray pulsar search #4 v1.17 takes about 8h to complete while running eight in parallel !
... Which has lead me to wonder why my desktop Linux host seems to be much slower, even though it shouldn't be. It's running Linux Mint 4.4.0-36 . I think I'm going to try soon if Ubuntizing that lazy pig would change anything
@mmonnin, the point here is to simply contribute more resources to science. I am fully aware it's not much better than a similar computer. Think of it more as adding another computer to crunch for science for free for a month. it is, in fact, just 8 cores running at 2.3ghz. If you don't think its worth the time, then don't do it. but don't come in here saying its not worth other peoples time to help contribute as much as they can. yea, maybe it's not 100,000 RAC but it Definitely Contributes to the Research.
BTW, it only took me (pretty new to linux, including figuring out how to set it up), about an hour. If an hour of your time isn't worth helping science, then you and I have very different views on why we participate in boinc.
@Richie, I have no experiance with SUSE so I'm of no help here, I have only used ubuntu and it was fairly easy.
Jordan
Did I saw it wasn't worth YOUR time? No, I said its not worth my time and hassle of getting the Amazon cloud setup and running. Just the setup was quite a bit harder than Google's. For comparison the Google compute was doing 20k PPD, the amazon service was less than 200 in FAH. E@H RAC would be even less than FAH's PPD to give an idea of how low compute amazon's free cloud actually performs. That slow of a processing power just ties up work that someone else could be doing in much less time. For projects that require work to be done sequentially or have wingmen, it is not worth it to the project to have such slow computing.
... Which has lead me to wonder why my desktop Linux host seems to be much slower, even though it shouldn't be. It's running Linux Mint 4.4.0-36 . I think I'm going to try soon if Ubuntizing that lazy pig would change anything
Thanks for posting this Ritchie, which host is the desktop Linux? One thing that does make a noticeable difference is the number of memory channels and L3 cache size.
Thanks for posting this Ritchie, which host is the desktop Linux? One thing that does make a noticeable difference is the number of memory channels and L3 cache size.
RAM should be working in triple channel (that's what I saw when this system was running Windows). In fact, this host still has also Windows 10 installed. I could join Albert from that side and test what the completion times would be then
Hmmh, I think I had confused myself at some point how the completion times between Windows and Linux tasks should be. Initially I thought Linux should be clearly the faster one of them, but it seems with almost identical hardware and clock speeds Linux version of the Gamma-ray pulsar search #4 v1.17 simply takes longer to complete. That's what also the instances at Yahoo platform suggests... but uhm, I just didn't get it.
This Linux desktop host is running 9 tasks concurrently at the moment and they seem to take ~42k secs to complete, but with 8 tasks it was ~38k. No GPU work.
Here's a Windows 10 host with very similar hardware and same speed settings: https://albertathome.org/host/16489 . It runs 8 tasks in parallel and finishes them in ~30k secs. Also runs GPU work at the same time (Einstein BRP6-Beta-cuda55).
I see the details for Linux host says "Cache 12288 KiB", but for all similar Windows hosts it says "256 KiB". I wonder why.
Ubik wrote:On amazon's
)
I jumped on this one too. I chose to install SUSE Linux Enterprise Server 12 SP1 (64-bit), but I haven't been able to make it install Boinc. I run "sudo zypper update" at first and it updated plenty of things and everything seem okay. Then I rebooted the instance and run "sudo zypper install boinc-client". It does some searches but then it says:
Loading repository data...
Reading installed packages...
'boinc-client' not found in package names. Trying capabilities.
No provider of 'boinc-client' found.
Resolving package dependencies...
Nothing to do.
How could I make it install boinc? I tried also "boinc" and "Boinc-Client". I see someone in internet has been running Boinc with SUSE Linux Enterprise Server 11. Isn't that possible (in an easy way, for a linux-newbie) with this version 12?
I wonder if Ubuntu Server 14.04 LTS could find Boinc. Should I have chosen that distro? Or their own Amazon Linux...
If Boinc can't run on this one, I'll propably just delete this instance and see if there was a possibility to launch a different one for free anymore.
Richie_9 wrote:I wonder if
)
I chose the Ubuntu Server distro. A simple "sudo aptitude install boing-client" worked fine.
I've tried both Googles
)
I've tried both Googles Compute and Amazon's cloud services running FAH. I had chosen the same, max CPU cores. The amazon option wasn't worth the time. Very low productivity and I assume the same would translate here. The google compute performed roughly like a Sandy/Ivy CPU of similar cores/speed.
@mmonnin, the point here is
)
@mmonnin, the point here is to simply contribute more resources to science. I am fully aware it's not much better than a similar computer. Think of it more as adding another computer to crunch for science for free for a month. it is, in fact, just 8 cores running at 2.3ghz. If you don't think its worth the time, then don't do it. but don't come in here saying its not worth other peoples time to help contribute as much as they can. yea, maybe it's not 100,000 RAC but it Definitely Contributes to the Research.
BTW,
it only took me (pretty new to linux, including figuring out how to set it up), about an hour. If an hour of your time isn't worth helping science, then you and I have very different views on why we participate in boinc.
@Richie, I have no experiance with SUSE so I'm of no help here, I have only used ubuntu and it was fairly easy.
Jordan
Richie_9 wrote:I jumped on
)
@Richie, your problem is that SUSE Enterprise 12 does not include the boinc-client in its own repository. You could 1) download it manually and install it from http://boinc.berkeley.edu/download.php, but it may require other software dependencies that you have not installed, 2) try to find a "contrib" or 3rd party repository for SUSE that includes the boinc-client software; frankly I don't know where to find one, or 3) install a Linux distribution that does include boinc-client in its list of software packages.
I would highly recommend #3. I believe Ubuntu 14 LTS (trusty) does include the boinc-client software.
Einstein@Home Project
Ubuntu 16.04 LTS provides
)
Ubuntu 16.04 LTS provides Boinc: sudo apt-get install boinc-client boinc-manager
This should produce an icon from which you can launch boinc-manager and join E@H.
Thanks guys for every advice
)
Thanks guys for every advice and hint!
I didn't get back earlier, but with Yahoo platform I ended up terminating the SUSE instance a few days ago (I just didn't have the skills to install BOINC on it).
I launched an Ubuntu Server 14.04 LTS instance instead. Then I succesfully managed to upgrade it to 16.04.1 LTS and installed BOINC (client only). It runs great, no problems at all. Today I upgraded the distribution to a yet newer internal version... from 4.4.0-34 to 4.4.0-36, just for fun. This has been interesting and I think I've learned at least something while fiddling with these
I run Albert tasks on these. Here are my Yahoo instances:
Linux - https://albertathome.org/host/16953
Windows - https://albertathome.org/host/16954
They crunch Gamma-ray pulsar search #4 v1.17 tasks. After the initial start, Windows instance seems to complete one task in ~34h and for Linux it takes ~43h. First tasks were a little bit faster, because instance starts with 100% CPU cycle time, but then it slows down. After about half an hour the initial CPU credits are used. Then it continues working at base level ( ~10% speed). Pauses and Idle time would slowly load up the CPU credits again, but that isn't happening while crunching.
For that Windows instance I use BOINC Manager and set BOINC to use only 50% of the CPU time. That should be more than enough. I thought maybe that would smooth out the competition between constant credit usage and "trying to give a birth to new credits". At the Yahoo EC2 Management Console Monitoring I can see now the graphs of CPU utilization, CPU credit usage and CPU credit balance have settled in a bit different way for both instances. Anyway, I believe both are running "at full speed".
I updated also My Google Cloud instances to Ubuntu version 4.4.0-36 today.. Their potential is at completely different level and they've been crunching like crazy pigs for a few days! One Gamma-ray pulsar search #4 v1.17 takes about 8h to complete while running eight in parallel !
... Which has lead me to wonder why my desktop Linux host seems to be much slower, even though it shouldn't be. It's running Linux Mint 4.4.0-36 . I think I'm going to try soon if Ubuntizing that lazy pig would change anything
Jordan Kallinen
)
Did I saw it wasn't worth YOUR time? No, I said its not worth my time and hassle of getting the Amazon cloud setup and running. Just the setup was quite a bit harder than Google's. For comparison the Google compute was doing 20k PPD, the amazon service was less than 200 in FAH. E@H RAC would be even less than FAH's PPD to give an idea of how low compute amazon's free cloud actually performs. That slow of a processing power just ties up work that someone else could be doing in much less time. For projects that require work to be done sequentially or have wingmen, it is not worth it to the project to have such slow computing.
Richie_9 wrote:... Which has
)
Thanks for posting this Ritchie, which host is the desktop Linux? One thing that does make a noticeable difference is the number of memory channels and L3 cache size.
AgentB wrote:Thanks for
)
It's this one: https://albertathome.org/host/16921
RAM should be working in triple channel (that's what I saw when this system was running Windows). In fact, this host still has also Windows 10 installed. I could join Albert from that side and test what the completion times would be then
Hmmh, I think I had confused myself at some point how the completion times between Windows and Linux tasks should be. Initially I thought Linux should be clearly the faster one of them, but it seems with almost identical hardware and clock speeds Linux version of the Gamma-ray pulsar search #4 v1.17 simply takes longer to complete. That's what also the instances at Yahoo platform suggests... but uhm, I just didn't get it.
This Linux desktop host is running 9 tasks concurrently at the moment and they seem to take ~42k secs to complete, but with 8 tasks it was ~38k. No GPU work.
Here's a Windows 10 host with very similar hardware and same speed settings: https://albertathome.org/host/16489 . It runs 8 tasks in parallel and finishes them in ~30k secs. Also runs GPU work at the same time (Einstein BRP6-Beta-cuda55).
I see the details for Linux host says "Cache 12288 KiB", but for all similar Windows hosts it says "256 KiB". I wonder why.