| 1 | [[PageOutline]] |
| 2 | = A web-based system for LAMMPS jobs = |
| 3 | |
| 4 | This document describes a system that allows scientists |
| 5 | to submit and monitor groups of LAMMPS jobs using BOINC. |
| 6 | The system has the following properties: |
| 7 | |
| 8 | * Users (i.e. scientists) interact entirely through a web-based interface. |
| 9 | They don't need to log into the project server, |
| 10 | and they don't need to know anything about BOINC. |
| 11 | * Users are authenticated by BOINC project accounts. |
| 12 | They do not need login accounts on the project server. |
| 13 | * Users can submit parameter sweeps consisting of thousands of jobs |
| 14 | as easily as submitting a single job. |
| 15 | * Users can get an estimate for the completion of a group of jobs |
| 16 | priori to submitting it, |
| 17 | and can get updated completion estimates as the batch is processed. |
| 18 | |
| 19 | This system was developed for researchers at Tsinghua University. |
| 20 | It can be modified to meet the needs of other projects using LAMMPS, |
| 21 | and many parts of it can be used to build similar systems |
| 22 | for other applications. |
| 23 | |
| 24 | == Authentication, access control, and quotas == |
| 25 | |
| 26 | Users (that is, job submitters) must create an account on the BOINC project; |
| 27 | this is done using a form on the project web site. |
| 28 | A project administrator |
| 29 | must then grant the user the right to submit jobs for LAMMPS |
| 30 | (and potentially for other applications). |
| 31 | Optionally, a designated user may be given the ability to |
| 32 | grant access rights to other users. |
| 33 | Each user has an associated "quota" that determines their |
| 34 | share of processing power. |
| 35 | |
| 36 | == Per-user file sandbox == |
| 37 | |
| 38 | LAMMPS input files can be large, |
| 39 | and it would be inconvenient to upload these files each time jobs are submitted. |
| 40 | Instead, we allow users to maintain a set of files on the project server; |
| 41 | this is called the user's "file sandbox". |
| 42 | |
| 43 | Using a web interface, users can |
| 44 | |
| 45 | * upload files from PC to sandbox |
| 46 | * view the files in their sandbox, including size and MD5. |
| 47 | * download files from sandbox to PC |
| 48 | * delete files from the sandbox |
| 49 | |
| 50 | Files in the sandbox can be modified, |
| 51 | and all old versions are retained on the server. |
| 52 | When a batch of jobs is submitted, |
| 53 | it uses the input file versions at the moment of submission, |
| 54 | even if the files are then modified while the batch is in progress. |
| 55 | |
| 56 | == LAMMPS job submission == |
| 57 | |
| 58 | Batches of LAMMPS jobs can be submitted using a web interface. |
| 59 | This process has two steps. |
| 60 | First, the user fills out a form specifying the following files, |
| 61 | which must be in the sandbox: |
| 62 | |
| 63 | * The atomic structure file |
| 64 | * The LAMMPS command script |
| 65 | * A zipped file containing the potential files needed for the simulation |
| 66 | * A file containing command lines to be passed to LAMMPS. |
| 67 | One job will be created for each line of this file. |
| 68 | |
| 69 | The user clicks the "Prepare" button on this form. |
| 70 | This validates the input files and estimates the resource requirements of the batch. |
| 71 | If there is an error in the input files, |
| 72 | the user sees the corresponding LAMMPS error messages. |
| 73 | Otherwise, they are shown an estimated completion time for the batch, |
| 74 | and an estimate of its disk usage both on the server and on volunteer computers. |
| 75 | If either of these is excessive, the user may opt to not submit the batch. |
| 76 | Otherwise, they submit the batch by clicking the "Submit" button. |
| 77 | |
| 78 | The input validation and runtime estimation is done by running |
| 79 | LAMMPS on the project server, |
| 80 | checking the output for error messages, |
| 81 | aborting it after a few time steps, |
| 82 | and measuring the average CPU time per time step. |
| 83 | From this, the FLOPS requirements of each job is estimated, |
| 84 | and (based on the performance statistics of the volunteer host population) |
| 85 | the completion time of the batch is estimated. |
| 86 | |
| 87 | == Batch monitoring, control, and output retrieval == |
| 88 | |
| 89 | Users can monitor and control batches through a web interface. |
| 90 | While a batch is in progress, the user can see its fraction done |
| 91 | and an updated estimate of its completion time. |
| 92 | In addition, the user can see the status of each of its component jobs: |
| 93 | unsent, in progress, failed, or completed. |
| 94 | When a job is completed the user can download its output files. |
| 95 | |
| 96 | A batch can be aborted at any time |
| 97 | (for example, because outputs of completed jobs are seen to be erroneous, |
| 98 | or because many jobs are failing). |
| 99 | If this is done, no further jobs from that batch will be issued. |
| 100 | |
| 101 | The user can, at any time, download a zipped file of all the output files |
| 102 | of all completed jobs in the batch. |
| 103 | |
| 104 | After a batch is completed or aborted, and all desired output files |
| 105 | have been downloaded, the user can "retire" the batch. |
| 106 | This causes its output files and database records to be deleted from the server. |