)]}'
{"reference/constellations/scientific.rst":[{"author":{"_account_id":7546,"name":"Blair Bethwaite","email":"blair.bethwaite@gmail.com","username":"blairo"},"change_message_id":"589f0db7f51b6aec4d37f5ef6a5f3e4dff1e8148","unresolved":false,"context_lines":[{"line_number":3,"context_line":"\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d"},{"line_number":4,"context_line":""},{"line_number":5,"context_line":"This constellation focuses on how OpenStack is used for Scientific Research"},{"line_number":6,"context_line":"including high-performance computing (HPC), high-throughput computing (HTC)"},{"line_number":7,"context_line":"and high-performance data analysis (HPDA) workloads."},{"line_number":8,"context_line":""},{"line_number":9,"context_line":"Status"},{"line_number":10,"context_line":"\u003d\u003d\u003d\u003d\u003d\u003d"}],"source_content_type":"text/x-rst","patch_set":1,"id":"9f6a8fd7_1a0da734","line":7,"range":{"start_line":6,"start_character":10,"end_line":7,"end_character":51},"updated":"2018-05-01 14:38:13.000000000","message":"- This is a very compute focused definition. That\u0027s ok, but cloud technology is also important in science because it is an enabler of collaboration - self-service is the key to unlocking and accelerating this.\n- Furthermore, we\u0027re starting to see more research publications with associated web-services or online datasets, cloud technology is also an enabler of dissemination.","commit_id":"11be4be1aece08117ff3806155d033c5ea257daa"},{"author":{"_account_id":7546,"name":"Blair Bethwaite","email":"blair.bethwaite@gmail.com","username":"blairo"},"change_message_id":"589f0db7f51b6aec4d37f5ef6a5f3e4dff1e8148","unresolved":false,"context_lines":[{"line_number":33,"context_line":"* Vibrant community of scientific users around the world contributing"},{"line_number":34,"context_line":"  to make the platform better for all users."},{"line_number":35,"context_line":""},{"line_number":36,"context_line":"Which OpenStack Projects?"},{"line_number":37,"context_line":"\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d"},{"line_number":38,"context_line":""},{"line_number":39,"context_line":"There are users that focus on baremetal compute (Ironic) and users that"}],"source_content_type":"text/x-rst","patch_set":1,"id":"9f6a8fd7_1aa927dc","line":36,"range":{"start_line":36,"start_character":16,"end_line":36,"end_character":24},"updated":"2018-05-01 14:38:13.000000000","message":"- Agree with Doug, there should be a list/s.\n- But furthermore, should the constellation recommend specific configurations of particular projects (especially the core components)? E.g., tenant networking support is probably pretty fundamental to many science cloud use-cases.","commit_id":"11be4be1aece08117ff3806155d033c5ea257daa"},{"author":{"_account_id":2472,"name":"Doug Hellmann","email":"dhellmann@redhat.com","username":"doug-hellmann"},"change_message_id":"b10987a3d1822606cc5740549d7796dd57045347","unresolved":false,"context_lines":[{"line_number":33,"context_line":"* Vibrant community of scientific users around the world contributing"},{"line_number":34,"context_line":"  to make the platform better for all users."},{"line_number":35,"context_line":""},{"line_number":36,"context_line":"Which OpenStack Projects?"},{"line_number":37,"context_line":"\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d"},{"line_number":38,"context_line":""},{"line_number":39,"context_line":"There are users that focus on baremetal compute (Ironic) and users that"}],"source_content_type":"text/x-rst","patch_set":1,"id":"9f6a8fd7_947bd0f2","line":36,"updated":"2018-05-01 13:37:21.000000000","message":"I like the fact that the prose in this section is feature-focused. I\u0027m trying to think of a better title for it. Maybe if we split \"Desirable Features\" out as prose and then have an \"OpenStack Projects\" section that just has a bullet list for quick reference?","commit_id":"11be4be1aece08117ff3806155d033c5ea257daa"},{"author":{"_account_id":14826,"name":"Mark Goddard","email":"markgoddard86@gmail.com","username":"mgoddard"},"change_message_id":"1aad02d939da61b6f4f484d6f7caf6d617522934","unresolved":false,"context_lines":[{"line_number":34,"context_line":"  to make the platform better for all users."},{"line_number":35,"context_line":""},{"line_number":36,"context_line":"Which OpenStack Projects?"},{"line_number":37,"context_line":"\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d"},{"line_number":38,"context_line":""},{"line_number":39,"context_line":"There are users that focus on baremetal compute (Ironic) and users that"},{"line_number":40,"context_line":"focus more on virtualisation (Nova with KVM). Very often the former is"}],"source_content_type":"text/x-rst","patch_set":1,"id":"9f6a8fd7_cab18f91","line":37,"updated":"2018-05-02 09:23:20.000000000","message":"This includes discussion of the hardware used to build the cloud, which is particularly important for HPC. It also discusses at a high level some techniques for making effective use of hardware. Perhaps hardware considerations are worth breaking apart from the discussion of OpenStack projects?","commit_id":"11be4be1aece08117ff3806155d033c5ea257daa"},{"author":{"_account_id":14826,"name":"Mark Goddard","email":"markgoddard86@gmail.com","username":"mgoddard"},"change_message_id":"1aad02d939da61b6f4f484d6f7caf6d617522934","unresolved":false,"context_lines":[{"line_number":44,"context_line":"It is common for users to request container orchestration, and Magnum is"},{"line_number":45,"context_line":"widely used to deliver that."},{"line_number":46,"context_line":""},{"line_number":47,"context_line":"Some HPC focused baremetal clouds offer low latency networks such as"},{"line_number":48,"context_line":"infiniband, which are usually integrated as provider networks exposed to"},{"line_number":49,"context_line":"baremetal compute nodes."},{"line_number":50,"context_line":""}],"source_content_type":"text/x-rst","patch_set":1,"id":"9f6a8fd7_6af3a374","line":47,"range":{"start_line":47,"start_character":17,"end_line":47,"end_character":26},"updated":"2018-05-02 09:23:20.000000000","message":"I wouldn\u0027t say these are restricted to bare metal, although in the virt case they\u0027ll typically use SR-IOV.","commit_id":"11be4be1aece08117ff3806155d033c5ea257daa"},{"author":{"_account_id":136,"name":"Tim Bell","email":"tim.bell@cern.ch","username":"tim-bell"},"change_message_id":"a2e930872716a971b7e1329c7cad6324c9d4eac6","unresolved":false,"context_lines":[{"line_number":50,"context_line":""},{"line_number":51,"context_line":"Virtualization users want predictable performance that also allows for"},{"line_number":52,"context_line":"good utilization of the purchased hardware. As such, SR-IOV is commonly"},{"line_number":53,"context_line":"used to imporve network performance, and CPU and memory pinning is used"},{"line_number":54,"context_line":"both to reduce virtualization overheads and maximise the isolation"},{"line_number":55,"context_line":"between virtual machines."},{"line_number":56,"context_line":""}],"source_content_type":"text/x-rst","patch_set":1,"id":"9f6a8fd7_4eb59ced","line":53,"range":{"start_line":53,"start_character":41,"end_line":53,"end_character":63},"updated":"2018-05-01 20:20:38.000000000","message":"Using techniques such as NUMA","commit_id":"11be4be1aece08117ff3806155d033c5ea257daa"},{"author":{"_account_id":7546,"name":"Blair Bethwaite","email":"blair.bethwaite@gmail.com","username":"blairo"},"change_message_id":"589f0db7f51b6aec4d37f5ef6a5f3e4dff1e8148","unresolved":false,"context_lines":[{"line_number":50,"context_line":""},{"line_number":51,"context_line":"Virtualization users want predictable performance that also allows for"},{"line_number":52,"context_line":"good utilization of the purchased hardware. As such, SR-IOV is commonly"},{"line_number":53,"context_line":"used to imporve network performance, and CPU and memory pinning is used"},{"line_number":54,"context_line":"both to reduce virtualization overheads and maximise the isolation"},{"line_number":55,"context_line":"between virtual machines."},{"line_number":56,"context_line":""}],"source_content_type":"text/x-rst","patch_set":1,"id":"9f6a8fd7_d7494ab7","line":53,"range":{"start_line":53,"start_character":8,"end_line":53,"end_character":15},"updated":"2018-05-01 14:38:13.000000000","message":"typo","commit_id":"11be4be1aece08117ff3806155d033c5ea257daa"},{"author":{"_account_id":10910,"name":"Stig Telfer","email":"stig.openstack@telfer.org","username":"stigtelfer"},"change_message_id":"5949fd36d3968606682453fef4ebc34a1c6db825","unresolved":false,"context_lines":[{"line_number":53,"context_line":"used to imporve network performance, and CPU and memory pinning is used"},{"line_number":54,"context_line":"both to reduce virtualization overheads and maximise the isolation"},{"line_number":55,"context_line":"between virtual machines."},{"line_number":56,"context_line":""},{"line_number":57,"context_line":"Many existing workloads written for batch compute platforms make use of"},{"line_number":58,"context_line":"parallel filesystem to share data both between computation steps and"},{"line_number":59,"context_line":"output results at the end. As such, there is strong demand for Manila"}],"source_content_type":"text/x-rst","patch_set":1,"id":"9f6a8fd7_06cb124a","line":56,"updated":"2018-05-01 16:56:13.000000000","message":"It might be worth including a mention of GPUs here, especially if we are also considering the requirements for AI use cases.  It could be as brief as \"Scientific workloads may also make use of GPUs.  A scientific constellation should support the effective management of these specialist resources.\"","commit_id":"11be4be1aece08117ff3806155d033c5ea257daa"},{"author":{"_account_id":7022,"name":"Adam Huffman","email":"adam.huffman@gmail.com","username":"verdurin"},"change_message_id":"2822de2e5b1bdefd57650e28af67c32a226a373b","unresolved":false,"context_lines":[{"line_number":55,"context_line":"between virtual machines."},{"line_number":56,"context_line":""},{"line_number":57,"context_line":"Many existing workloads written for batch compute platforms make use of"},{"line_number":58,"context_line":"parallel filesystem to share data both between computation steps and"},{"line_number":59,"context_line":"output results at the end. As such, there is strong demand for Manila"},{"line_number":60,"context_line":"with backends such as CephFS."},{"line_number":61,"context_line":""}],"source_content_type":"text/x-rst","patch_set":1,"id":"5f7c97a3_0e9c5a0d","line":58,"range":{"start_line":58,"start_character":9,"end_line":58,"end_character":19},"updated":"2018-05-10 12:49:33.000000000","message":"filesystems","commit_id":"11be4be1aece08117ff3806155d033c5ea257daa"},{"author":{"_account_id":14826,"name":"Mark Goddard","email":"markgoddard86@gmail.com","username":"mgoddard"},"change_message_id":"1aad02d939da61b6f4f484d6f7caf6d617522934","unresolved":false,"context_lines":[{"line_number":57,"context_line":"Many existing workloads written for batch compute platforms make use of"},{"line_number":58,"context_line":"parallel filesystem to share data both between computation steps and"},{"line_number":59,"context_line":"output results at the end. As such, there is strong demand for Manila"},{"line_number":60,"context_line":"with backends such as CephFS."},{"line_number":61,"context_line":""},{"line_number":62,"context_line":"Release Highlights"},{"line_number":63,"context_line":"\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d"}],"source_content_type":"text/x-rst","patch_set":1,"id":"9f6a8fd7_4a4c9f9d","line":60,"updated":"2018-05-02 09:23:20.000000000","message":"Could also mention Sahara for HPDA","commit_id":"11be4be1aece08117ff3806155d033c5ea257daa"},{"author":{"_account_id":7022,"name":"Adam Huffman","email":"adam.huffman@gmail.com","username":"verdurin"},"change_message_id":"2822de2e5b1bdefd57650e28af67c32a226a373b","unresolved":false,"context_lines":[{"line_number":57,"context_line":"Many existing workloads written for batch compute platforms make use of"},{"line_number":58,"context_line":"parallel filesystem to share data both between computation steps and"},{"line_number":59,"context_line":"output results at the end. As such, there is strong demand for Manila"},{"line_number":60,"context_line":"with backends such as CephFS."},{"line_number":61,"context_line":""},{"line_number":62,"context_line":"Release Highlights"},{"line_number":63,"context_line":"\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d"}],"source_content_type":"text/x-rst","patch_set":1,"id":"5f7c97a3_cea18248","line":60,"range":{"start_line":60,"start_character":22,"end_line":60,"end_character":28},"updated":"2018-05-10 12:49:33.000000000","message":"Should also mention traditional HPC-associated filesystems such as GPFS and Lustre","commit_id":"11be4be1aece08117ff3806155d033c5ea257daa"},{"author":{"_account_id":7546,"name":"Blair Bethwaite","email":"blair.bethwaite@gmail.com","username":"blairo"},"change_message_id":"589f0db7f51b6aec4d37f5ef6a5f3e4dff1e8148","unresolved":false,"context_lines":[{"line_number":58,"context_line":"parallel filesystem to share data both between computation steps and"},{"line_number":59,"context_line":"output results at the end. As such, there is strong demand for Manila"},{"line_number":60,"context_line":"with backends such as CephFS."},{"line_number":61,"context_line":""},{"line_number":62,"context_line":"Release Highlights"},{"line_number":63,"context_line":"\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d"},{"line_number":64,"context_line":""}],"source_content_type":"text/x-rst","patch_set":1,"id":"9f6a8fd7_faf793d3","line":61,"updated":"2018-05-01 14:38:13.000000000","message":"Maybe this belongs in the Future Plans section, but just dumping it here for now:\n\nCommon for science clouds to *not* do PAYG or charge-back and quickly become resource constrained, so there is a wide need for supporting resource reservations as well as the typical on-demand usage.","commit_id":"11be4be1aece08117ff3806155d033c5ea257daa"},{"author":{"_account_id":2472,"name":"Doug Hellmann","email":"dhellmann@redhat.com","username":"doug-hellmann"},"change_message_id":"b10987a3d1822606cc5740549d7796dd57045347","unresolved":false,"context_lines":[{"line_number":65,"context_line":"Here we look at some recent releases and what particular features will"},{"line_number":66,"context_line":"be of interest to the scientific computing community:"},{"line_number":67,"context_line":""},{"line_number":68,"context_line":"* Queens: Volume multi-attach, Application credentials "},{"line_number":69,"context_line":""},{"line_number":70,"context_line":"Future Plans"},{"line_number":71,"context_line":"\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d"}],"source_content_type":"text/x-rst","patch_set":1,"id":"9f6a8fd7_54169857","line":68,"updated":"2018-05-01 13:37:21.000000000","message":"How do you anticipate keeping this up to date over time? Perhaps the relevant SIGs would propose patches?","commit_id":"11be4be1aece08117ff3806155d033c5ea257daa"},{"author":{"_account_id":14826,"name":"Mark Goddard","email":"markgoddard86@gmail.com","username":"mgoddard"},"change_message_id":"1aad02d939da61b6f4f484d6f7caf6d617522934","unresolved":false,"context_lines":[{"line_number":75,"context_line":"this community:"},{"line_number":76,"context_line":""},{"line_number":77,"context_line":"* Improved utilization of a finite pool of resources, while ensuring there"},{"line_number":78,"context_line":"  is also fair sharing of the resources. As such, pre-emptable instances"},{"line_number":79,"context_line":"  and the reservation of resources are of particular interest."},{"line_number":80,"context_line":"* Better support for virtual organisations where authorization and resource"},{"line_number":81,"context_line":"  allocation is delegated to the manager of a virtual organisation, very"}],"source_content_type":"text/x-rst","patch_set":1,"id":"9f6a8fd7_ca242fc9","line":78,"range":{"start_line":78,"start_character":50,"end_line":78,"end_character":62},"updated":"2018-05-02 09:23:20.000000000","message":"preemptable","commit_id":"11be4be1aece08117ff3806155d033c5ea257daa"},{"author":{"_account_id":8878,"name":"Masahito Muroi","email":"masahito.muroi@linecorp.com","username":"masa"},"change_message_id":"85b75a1ecdc164d036c0c193790afa4e09830395","unresolved":false,"context_lines":[{"line_number":76,"context_line":""},{"line_number":77,"context_line":"* Improved utilization of a finite pool of resources, while ensuring there"},{"line_number":78,"context_line":"  is also fair sharing of the resources. As such, pre-emptable instances"},{"line_number":79,"context_line":"  and the reservation of resources are of particular interest."},{"line_number":80,"context_line":"* Better support for virtual organisations where authorization and resource"},{"line_number":81,"context_line":"  allocation is delegated to the manager of a virtual organisation, very"},{"line_number":82,"context_line":"  often this is the Principal Investigator who is responsible for ensuring"}],"source_content_type":"text/x-rst","patch_set":1,"id":"5f7c97a3_09fed436","line":79,"range":{"start_line":79,"start_character":6,"end_line":79,"end_character":34},"updated":"2018-05-08 04:44:31.000000000","message":"Related to the Blazar project?  We\u0027re welcome to support scientific computing.","commit_id":"11be4be1aece08117ff3806155d033c5ea257daa"},{"author":{"_account_id":14826,"name":"Mark Goddard","email":"markgoddard86@gmail.com","username":"mgoddard"},"change_message_id":"1aad02d939da61b6f4f484d6f7caf6d617522934","unresolved":false,"context_lines":[{"line_number":77,"context_line":"* Improved utilization of a finite pool of resources, while ensuring there"},{"line_number":78,"context_line":"  is also fair sharing of the resources. As such, pre-emptable instances"},{"line_number":79,"context_line":"  and the reservation of resources are of particular interest."},{"line_number":80,"context_line":"* Better support for virtual organisations where authorization and resource"},{"line_number":81,"context_line":"  allocation is delegated to the manager of a virtual organisation, very"},{"line_number":82,"context_line":"  often this is the Principal Investigator who is responsible for ensuring"},{"line_number":83,"context_line":"  the resources granted to the virtual organisation are used responsibly."}],"source_content_type":"text/x-rst","patch_set":1,"id":"9f6a8fd7_0a3b07eb","line":80,"updated":"2018-05-02 09:23:20.000000000","message":"Is the VO/PI terminology understood widely enough to use here?","commit_id":"11be4be1aece08117ff3806155d033c5ea257daa"}]}