Hi all,
There was some email conversation on this list a couple of weeks ago about the HPC SIG, but it didn't really seem to go anywhere or have any solid conclusion, other than "join #centos-devel". Between some freenode instability/spam protections etc. some issues with my usual persistent session, and just sheer quantity of conversation happening there, it's somewhat hard to figure out where anything sits. I didn't see any obvious signs of further conversation there, but there's a good chance I missed it. If there was further conversation and someone would be able to point me towards it off list, I'd love to read it.
As a cloud provider that provides high spec bare metal servers to customers, we're finding a lot of usage and interest from customers with various forms of HPC workloads, everything from GROMACs to Hadoop and beyond. CentOS, likewise, continues to be a popular distribution for our customers, across every hardware and virtual machine specification. If there's ideas on how to make HPC better on CentOS, I'd love to be part of the conversation, and see if there are opportunities to help.
Paul
On Tue, Sep 04, 2018 at 11:41:32AM -0700, Paul Graydon wrote:
There was some email conversation on this list a couple of weeks ago about the HPC SIG, but it didn't really seem to go anywhere or have any solid conclusion, other than "join #centos-devel". Between some freenode instability/spam protections etc. some issues with my usual persistent session, and just sheer quantity of conversation happening there, it's somewhat hard to figure out where anything sits. I didn't see any obvious signs of further conversation there, but there's a good chance I missed it. If there was further conversation and someone would be able to point me towards it off list, I'd love to read it.
As far as I am aware, I am the only 'active' person in the HPC SIG and have not seen any further HPC discussions.
As a cloud provider that provides high spec bare metal servers to customers, we're finding a lot of usage and interest from customers with various forms of HPC workloads, everything from GROMACs to Hadoop and beyond. CentOS, likewise, continues to be a popular distribution for our customers, across every hardware and virtual machine specification. If there's ideas on how to make HPC better on CentOS, I'd love to be part of the conversation, and see if there are opportunities to help.
Right now I still think that OpenHPC is a very good starting point to collaborate on HPC packages for CentOS. Especially if you are interested in bringing in additional packages like GROMACs.
I am also happy to help any other HPC efforts around CentOS. So if you have any ideas how to better integrate HPC into CentOS let me (us) know.
Adrian
On 9/4/18 14:45, Adrian Reber wrote:
On Tue, Sep 04, 2018 at 11:41:32AM -0700, Paul Graydon wrote:
There was some email conversation on this list a couple of weeks ago about the HPC SIG, but it didn't really seem to go anywhere or have any solid conclusion, other than "join #centos-devel". Between some freenode instability/spam protections etc. some issues with my usual persistent session, and just sheer quantity of conversation happening there, it's somewhat hard to figure out where anything sits. I didn't see any obvious signs of further conversation there, but there's a good chance I missed it. If there was further conversation and someone would be able to point me towards it off list, I'd love to read it.
As far as I am aware, I am the only 'active' person in the HPC SIG and have not seen any further HPC discussions.
As a cloud provider that provides high spec bare metal servers to customers, we're finding a lot of usage and interest from customers with various forms of HPC workloads, everything from GROMACs to Hadoop and beyond. CentOS, likewise, continues to be a popular distribution for our customers, across every hardware and virtual machine specification. If there's ideas on how to make HPC better on CentOS, I'd love to be part of the conversation, and see if there are opportunities to help.
Right now I still think that OpenHPC is a very good starting point to collaborate on HPC packages for CentOS. Especially if you are interested in bringing in additional packages like GROMACs.
I am also happy to help any other HPC efforts around CentOS. So if you have any ideas how to better integrate HPC into CentOS let me (us) know.
For EPEL, there is an rpm that made its way in to centos extras to make it easy for people to install & enable those repositories. Is that a good potential way forward for OpenHPC and CentOS (is that even feasible or a good idea?)
Paul
On Wed, Sep 05, 2018 at 07:53:05AM -0700, Paul Graydon wrote:
On 9/4/18 14:45, Adrian Reber wrote:
On Tue, Sep 04, 2018 at 11:41:32AM -0700, Paul Graydon wrote:
There was some email conversation on this list a couple of weeks ago about the HPC SIG, but it didn't really seem to go anywhere or have any solid conclusion, other than "join #centos-devel". Between some freenode instability/spam protections etc. some issues with my usual persistent session, and just sheer quantity of conversation happening there, it's somewhat hard to figure out where anything sits. I didn't see any obvious signs of further conversation there, but there's a good chance I missed it. If there was further conversation and someone would be able to point me towards it off list, I'd love to read it.
As far as I am aware, I am the only 'active' person in the HPC SIG and have not seen any further HPC discussions.
As a cloud provider that provides high spec bare metal servers to customers, we're finding a lot of usage and interest from customers with various forms of HPC workloads, everything from GROMACs to Hadoop and beyond. CentOS, likewise, continues to be a popular distribution for our customers, across every hardware and virtual machine specification. If there's ideas on how to make HPC better on CentOS, I'd love to be part of the conversation, and see if there are opportunities to help.
Right now I still think that OpenHPC is a very good starting point to collaborate on HPC packages for CentOS. Especially if you are interested in bringing in additional packages like GROMACs.
I am also happy to help any other HPC efforts around CentOS. So if you have any ideas how to better integrate HPC into CentOS let me (us) know.
For EPEL, there is an rpm that made its way in to centos extras to make it easy for people to install & enable those repositories. Is that a good potential way forward for OpenHPC and CentOS (is that even feasible or a good idea?)
Not sure if this discussion belongs here... If the package is already part of EPEL I am not sure it needs to be part of OpenHPC. It could if you need the additional features OpenHPC provides. From my point of view the advantages of OpenHPC are the integration of packages in a way common for HPC sites, for all arches and operating systems OpenHPC supports. This makes it easier for OpenHPC users to easily figure out the available software. In addition, which is even more CentOS unrelated, OpenHPC also provides packages compiled with proprietary (ARM and Intel) compilers (which is also common for HPC sites).
Let us discuss this special topic off-list or on the OpenHPC side.
Adrian
On Tue, 4 Sep 2018 23:45:17 +0200 Adrian Reber adrian@lisas.de wrote:
On Tue, Sep 04, 2018 at 11:41:32AM -0700, Paul Graydon wrote:
There was some email conversation on this list a couple of weeks ago about the HPC SIG, but it didn't really seem to go anywhere or have any solid conclusion, other than "join #centos-devel". Between some freenode instability/spam protections etc. some issues with my usual persistent session, and just sheer quantity of conversation happening there, it's somewhat hard to figure out where anything sits. I didn't see any obvious signs of further conversation there, but there's a good chance I missed it. If there was further conversation and someone would be able to point me towards it off list, I'd love to read it.
As far as I am aware, I am the only 'active' person in the HPC SIG and have not seen any further HPC discussions.
As a cloud provider that provides high spec bare metal servers to customers, we're finding a lot of usage and interest from customers with various forms of HPC workloads, everything from GROMACs to Hadoop and beyond. CentOS, likewise, continues to be a popular distribution for our customers, across every hardware and virtual machine specification. If there's ideas on how to make HPC better on CentOS, I'd love to be part of the conversation, and see if there are opportunities to help.
Right now I still think that OpenHPC is a very good starting point to collaborate on HPC packages for CentOS. Especially if you are interested in bringing in additional packages like GROMACs.
I think application level packages (or the more general scientific software building and providing question) is better addressed by something like easybuild or spack.
It makes more sense to me that hpc-sig and/or OpenHPC works with more system near components (file system support, batch systems, module systems, node cloning / provisioning, ...).
/Peter
On Thu, Sep 06, 2018 at 03:13:53PM +0200, Peter Kjellström wrote:
On Tue, 4 Sep 2018 23:45:17 +0200 Adrian Reber adrian@lisas.de wrote:
On Tue, Sep 04, 2018 at 11:41:32AM -0700, Paul Graydon wrote:
There was some email conversation on this list a couple of weeks ago about the HPC SIG, but it didn't really seem to go anywhere or have any solid conclusion, other than "join #centos-devel". Between some freenode instability/spam protections etc. some issues with my usual persistent session, and just sheer quantity of conversation happening there, it's somewhat hard to figure out where anything sits. I didn't see any obvious signs of further conversation there, but there's a good chance I missed it. If there was further conversation and someone would be able to point me towards it off list, I'd love to read it.
As far as I am aware, I am the only 'active' person in the HPC SIG and have not seen any further HPC discussions.
As a cloud provider that provides high spec bare metal servers to customers, we're finding a lot of usage and interest from customers with various forms of HPC workloads, everything from GROMACs to Hadoop and beyond. CentOS, likewise, continues to be a popular distribution for our customers, across every hardware and virtual machine specification. If there's ideas on how to make HPC better on CentOS, I'd love to be part of the conversation, and see if there are opportunities to help.
Right now I still think that OpenHPC is a very good starting point to collaborate on HPC packages for CentOS. Especially if you are interested in bringing in additional packages like GROMACs.
I think application level packages (or the more general scientific software building and providing question) is better addressed by something like easybuild or spack.
Yes, that is why OpenHPC includes easybuild and spack. But I also think that there is value in application level packages coming from something like OpenHPC if the user base is large enough. Especially if reading the motivation of Paul, being 'a cloud provider that provides high spec bare metal servers to customers', pre-built software might be more interesting then build from source like easybuild or spack (although I think spack does (or will do) provide precompiled binaries, which makes it more similar to OpenHPC again...).
It makes more sense to me that hpc-sig and/or OpenHPC works with more system near components (file system support, batch systems, module systems, node cloning / provisioning, ...).
And that is one of the main features provided by OpenHPC.
Adrian
-
Adrian Reber -
Paul Graydon -
Peter Kjellström