Changes between Initial Version and Version 1 of ClientSchedOld


Ignore:
Timestamp:
11/21/11 00:20:45 (6 years ago)
Author:
davea
Comment:

--

Legend:

Unmodified
Added
Removed
Modified
  • ClientSchedOld

    v1 v1  
     1= Client scheduling policies =
     2
     3NOTE: this document is outdated because of development in the 6.4 client
     4to handle GPU and multithread apps.
     5Changes include:
     6
     7 1. app versions now include avg_ncpus, coprocessor usage, and
     8 a FLOPS estimate (this defaults to the CPU benchmark,
     9 but for GPU and multithread apps it will be different).
     10 This info is sent from the server.
     11 1. Estimating the duration of unstarted jobs:
     12 jobs are now associated with specific app versions.
     13 The estimated duration of an unstarted job is the WU's
     14 FLOP estimate divided by the app version's FLOPS,
     15 scaled by the duration correction factor.
     16 1. Duration correction factor: this is now based on elapsed time
     17 (i.e. wall time during which the job has been running) rather than CPU time.
     18 Yes, this is affected by non-BOINC CPU load; that's as it should be.
     19 1. "CPU efficiency" is no longer maintained; it's subsumed in DCF.
     20 1. Estimating the duration of running jobs:
     21 this is a weighted average of static and dynamic estimates.
     22 The dynamic estimate is now based on elapsed time rather than CPU time.
     23 So if a GPU job has been running for 5 min, is 25% done,
     24 and has used 1 min of CPU, its dynamic estimate is 20 min (not 4 min).
     25 1. Round-robin simulation: this was modified to reflect multi-thread
     26 and coproc apps (e.g., if the host has 1 GPU, only one coproc app
     27 can run at a time).
     28 If CPUs are idle because coprocs are in use,
     29 don't count it towards CPU shortfall.
     30 1. scheduler_cpus() and enforce_schedule() take coprocs and
     31 avg_ncpus into account.  They try to keep GPUs busy if possible.
     32
     33---------------
     34
     35This document describes three related parts of the BOINC core client:
     36
     37 '''CPU scheduling policy'''::
     38        Of the results that are runnable, which ones to execute? BOINC will generally execute NCPUS results at once, where NCPUS is the minimum of the physical number of CPUs (counting hyperthreading) and the user's 'max_cpus' general preference.
     39 '''CPU scheduling enforcement'''::
     40        When to actually enforce the schedule (i.e. by preempting and starting tasks)? Sometimes it's preferable to delay the preemption of an application until it checkpoints.
     41 '''Work-fetch policy'''::
     42        When should the core client ask a project for more work, which project should it ask, and how much work should it ask for?
     43
     44The goals of these policies are (in descending priority):
     45
     46 1. Results should be completed and reported by their deadline (because results reported after their deadline may not have any value to the project and may not be granted credit).
     47 1. NCPUS processors should be kept busy.
     48 1. At any given point, a computer should have enough work so that NCPUS processors will be busy for at least min_queue days (min_queue is a user preference).
     49 1. Project resource shares should be honored over the long term.
     50 1. If a computer is attached to multiple projects, execution should rotate among projects on a frequent basis (as defined by the user's 'CPU scheduling period' preference).
     51 1. Execution should not switch between projects much more frequently than the scheduling period, Otherwise, if the 'remove processes from memory' preference is set, and some applications take a long time to resume from a checkpoint, lot of CPU time will be wasted.
     52
     53In previous versions of BOINC, the core client attempted to maintain at least one result for each attached project, and would do weighted round-robin CPU scheduling among all projects. In some scenarios (any combination of slow computer, lots of projects, and tight deadlines) a computer could miss the deadlines of all its results. The new policies solve this problem as follows:
     54
     55 * Work fetch is limited to ensure that deadlines can be met. A computer attached to 10 projects might have work for only a few (perhaps only one) at a given time.
     56 * If deadlines are threatened, the CPU scheduling policy optimizes the likelihood of meeting deadlines, at the expense of variety.
     57
     58== Concepts and terms ==
     59
     60=== Wall CPU time ===
     61'''Wall CPU time''' is the amount of wall-clock time a process has been runnable at the OS level. The actual CPU time may be less than this, e.g. if the process does a lot of paging, or if other (non-BOINC) processing jobs run at the same time.
     62
     63BOINC uses wall CPU time as the measure of CPU resource usage. Wall CPU time is more fair than actual CPU time in the case of paging apps. In addition, the measurement of actual CPU time depends on apps to report it correctly, and they may not do this.
     64
     65=== Normalized CPU time === #NormalizedCPUTime
     66The '''normalized CPU time''' of a result is an estimate of the wall time it will take to complete, taking into account
     67
     68 * the fraction of time BOINC runs ('on-fraction')
     69 * the fraction of time computation is enabled ('active-fraction')
     70 * CPU efficiency (the ratio of actual CPU to wall CPU)
     71
     72but ''Not'' taking into account the Resource Share of Projects.
     73
     74=== Project-normalized CPU time ===
     75
     76The '''project-normalized CPU time''' of a result is an estimate of the wall time it will take to complete, taking into account the above factors plus the project's resource share relative to other potentially runnable projects.
     77
     78The 'work_req' element of a scheduler RPC request is in units of project-normalized CPU time. In deciding how much work to send, the scheduler must take into account the project's resource share fraction, and the host's on-fraction and active-fraction.
     79
     80For example, suppose a host has 1 GFLOP/sec CPUs, the project's resource share fraction is 0.5, the host's on-fraction is 0.8 and the host's active-fraction is 0.9. Then the expected processing rate per CPU is
     81
     82{{{
     83(1 GFLOP/sec)*0.5*0.8*0.9 = 0.36 GFLOP/sec
     84}}}
     85
     86If the host requests 1000 project-normalized CPU seconds of work, the scheduler should send it at least 360 GFLOPs of work.
     87
     88=== Result states ===
     89
     90R is '''runnable''' if
     91 * Neither R nor R.project is suspended, and
     92 * R's input files have been downloaded, and
     93 * R hasn't finished computing
     94
     95R is '''nearly runnable''' if
     96 * Neither R nor R.project is suspended, and
     97 * None of R's input files is in a 'download deferred' state.
     98 * R hasn't finished computing
     99
     100=== Project states ===
     101
     102P is '''runnable''' if
     103 * P has at least one runnable result (this implies that P is not suspended).
     104
     105P is '''downloading''' if
     106 * P is not suspended, and
     107 * P has at least one result whose files are being downloaded and none of the downloads is deferred.
     108
     109P is '''fetchable''' (i.e. the work-fetch policy allows work to be fetched from it) if
     110 * P is not suspended, and
     111 * P is not deferred (i.e. its minimum RPC time is in the past), and
     112 * P's no-new-work flag is not set, and
     113 * P is not overworked (see definition below), and
     114 * a fetch of P's master file is not pending
     115
     116P is '''latency-limited''' if
     117 * The client's last scheduler RPC to P returned a 'no work because of deadlines' flag, and
     118 * the RPC reply's delay request has not yet elapsed.
     119
     120This means that P has work available, but didn't send any because the work's deadlines couldn't be met given the existing work queue. P is '''potentially runnable''' if
     121
     122 * P is either runnable, downloading, fetchable, overworked, or latency-limited.
     123
     124This means that, to the best of the client's knowledge, it could do work for P if it wanted to.
     125
     126=== Debt ===
     127
     128Intuitively, a project's 'debt' is how much work is owed to it, relative to other projects. BOINC uses two types of debt; each is defined for a set S of projects. In each case, the debt is recalculated periodically as follows:
     129 * A = the wall CPU time used by projects in S during this period
     130 * R = sum of resource shares of projects in S
     131 * For each project P in S:
     132  * F = P.resource_share / R (i.e., P's fractional resource share)
     133  * W = A*F (i.e., how much wall CPU time P should have gotten)
     134  * P.debt += W - P.wall_cpu_time (i.e. what P should have gotten            minus what it got).
     135 * P.debt is normalized so that the mean or minimum is zero.
     136
     137'''Short-term debt''' is used by the CPU scheduler. It is adjusted over the set of runnable projects. It is normalized so that minimum short-term debt is zero, and maximum short-term debt is no greater than 86,400 (i.e. one day).
     138
     139'''Long-term debt''' is used by the work-fetch policy. It is defined for all projects, and adjusted over the set of potentially runnable projects. It is normalized so that average long-term debt, over all project, is zero.
     140
     141== Round-robin simulation ==
     142
     143The CPU scheduling and work fetch policies use the results of a simulation of weighted round-robin scheduling applied to the set of nearly runnable results. The simulation takes into account on-fraction and active-fraction. It produces the following outputs:
     144
     145 * deadline_missed(R): whether result R misses its deadline.
     146 * deadlines_missed(P): the number of results R of P for which deadline_missed(R).
     147 * total_shortfall: the additional normalized CPU time needed to keep all CPUs busy for the next min_queue seconds.
     148 * shortfall(P): the additional normalized CPU time needed for project P to keep it from running out of work in the next min_queue seconds.
     149
     150In the example below, projects A and B have resource shares 2 and 1 respectively. A has results A1 and A2, and B has result B1. The computer has two CPUs. From time 0 to 4 all three results run with equal weighting. At time 4 result A2 finishes. From time 4 to 8, project A gets only a 0.5 share because it has only one result. At time 8, result A1 finishes.
     151
     152In this case, shortfall(A) is 4, shortfall(B``) is 0, and total_shortfall is 2.
     153
     154[[Image(http://boinc.berkeley.edu/rr_sim.png)]]
     155
     156== CPU scheduling policy ==
     157
     158The CPU scheduler uses an earliest-deadline-first (EDF) policy for results that are in danger of missing their deadline, and weighted round-robin among other projects if additional CPUs exist. This allows the client to meet deadlines that would otherwise be missed, while honoring resource shares over the long term. The scheduling policy is:
     159
     160 1. Set the 'anticipated debt' of each project to its short-term debt
     161 1. Let P be the project with the earliest-deadline runnable result among projects with deadlines_missed(P)>0. Let R be P's earliest-deadline runnable result not scheduled yet. Tiebreaker: least index in result array.
     162 1. If such an R exists, schedule R, decrement P's anticipated debt, and decrement deadlines_missed(P).
     163 1. If there are more CPUs, and projects with deadlines_missed(P)>0, go to 1.
     164 1. If all CPUs are scheduled, stop.
     165 1. If there is a result R that is currently running, and has been running for less than the CPU scheduling period, schedule R and go to 5.
     166 1. Find the project P with the greatest anticipated debt, select one of P's runnable results (picking one that is already running, if possible, else the one received first from the project) and schedule that result.
     167 1. Decrement P's anticipated debt by the 'expected payoff' (the scheduling period divided by NCPUS).
     168 1. Go to 5.
     169
     170The CPU scheduler runs when a result is completed, when the end of the user-specified scheduling period is reached, when new results become runnable, or when the user performs a UI interaction (e.g. suspending or resuming a project or result).
     171
     172
     173== CPU schedule enforcement ==
     174
     175The CPU scheduler decides what results should run, but it doesn't enforce this decision. This enforcement is done by a separate '''scheduler enforcement function''', which is called by the CPU scheduler at its conclusion. Let X be the set of scheduled results that are not currently running, let Y be the set of running results that are not scheduled, and let T be the time the scheduler last ran. The enforcement policy is as follows:
     176
     177 1. If deadline_missed(R) for some R in X, then preempt a result in Y, and run R (preempt the result with the least CPU wall time since checkpoint). Repeat as needed.
     178 1. If there is a result R in Y that checkpointed more recently than T, then preempt R and run a result in X.
     179
     180== Work-fetch policy ==
     181A project P is '''overworked''' if
     182
     183 * P.long_term_debt < -sched_period
     184
     185This condition occurs if P's results run in EDF mode (and in extreme cases, when a project with large negative LTD is detached). The work-fetch policy avoids getting work from overworked projects. This prevents a situation where a project with short deadlines gets more than its share of CPU time.
     186
     187The work-fetch policy uses the functions
     188
     189{{{
     190frs(project P)
     191}}}
     192
     193P's fractional resource share among fetchable projects.
     194
     195The work-fetch policy function is called every few minutes (or as needed) by the scheduler RPC polling function. It sets the variable '''P.work_request_size''' for each project P, which is the number of seconds of work to request if we do a scheduler RPC to P. This is computed as follows:
     196
     197{{{
     198for each project P
     199    if P is suspended, deferred, overworked, or no-new-work
     200        P.work_request_size = 0
     201    else
     202        P.work_request_size = shortfall(P)
     203
     204if total_shortfall > 0
     205    if P.work_request_size==0 for all P
     206        for each project P
     207            if P is suspended, deferred, overworked, or no-new-work
     208                continue
     209            P.work_request_size = 1
     210
     211    if P.work_request_size==0 for all P
     212        for each project P
     213            if P is suspended, deferred, or no-new-work
     214                continue
     215            P.work_request_size = 1
     216
     217    if P.work_request_size>0 for some P
     218        Normalize P.work_request_size so that they sum to total_shortfall
     219        and are proportional to P.resource_share
     220}}}
     221
     222For non-CPU-intensive projects, P.work_request_size is set to 1 if P has no nearly-runnable result, otherwise 0.
     223
     224The scheduler RPC mechanism may select a project to contact because of a user request, an outstanding trickle-up message, or a result that is overdue for reporting. If it does so, it will also request work from that project. Otherwise, the RPC mechanism chooses the project P for which
     225
     226{{{
     227P.work_request_size>0 and
     228P.long_term_debt + shortfall(P) is greatest
     229}}}
     230
     231and requests work from that project. Note: P.work_request_size is in units of normalized CPU time, so the actual work request (which is in units of project-normalized CPU time) is P.work_request_size divided by P's resource share fraction relative to potentially runnable projects.
     232
     233----
     234
     235== Scheduler work-send policy ==
     236
     237NOTE: the following has not been implemented, and is independent of the above policies.
     238
     239The scheduler should avoid sending results whose deadlines are likely to be missed, or which are likely to cause existing results to miss their deadlines. This will be accomplished as follows:
     240 * Scheduler requests includes connection period, list of queued result (with estimated time remaining and deadline) and project resource fractions.
     241 * The scheduler won't send results whose deadlines are less than now + min_queue.
     242 * The scheduler does an EDF simulation of the initial workload to determine by how much each result misses its deadline. For each result R being considered for sending, the scheduler does an EDF simulation. If R meets its deadline and no result misses its deadline by more than it did previously, R is sent.
     243 * If the scheduler has work but doesn't send any because of deadline misses, it returns a 'no work because of deadlines' flag. If the last RPC to a project returned this flag, it is marked as latency-limited and accumulates LTD.
     244
     245----
     246
     247== Describing scenarios ==
     248
     249We encourage the use of the following notation for describing scheduling scenarios (times are given in hours):
     250
     251P(C, D, R)
     252
     253This describes a project with
     254
     255  * C = CPU time per task
     256  * D = delay bound
     257  * R = fractional resource share
     258
     259A scenario is described by a list of project, plus the following optional parameters:
     260  * NCPUS: number of CPUS (default 1)
     261  * min_queue
     262  * leave_in_memory
     263  * cpu_scheduling_period
     264
     265An example scenario description is:
     266{{{
     267P1(1000, 2000, .5)
     268P2(1, 10, .5)
     269NCPUS=4
     270}}}
     271
     272== Scenarios ==
     273
     274=== Scenario 1 ===
     275
     276{{{
     277P1(0.1, 1, .5)
     278P2(1, 24, .25)
     279P3(1, 24, .25)
     280NCPUS = 2
     281leave_in_memory = false
     282cpu_scheduling_period = 1
     283}}}
     284
     285Typically one CPU will process 6-minute tasks for P1, and the other CPU will alternate between P2 and P3. It's critical that the scheduler run each task of P2 and P3 for the full CPU scheduling period. If we went strictly by debt, we'd end up switching between them every 6 minutes, and both P2 and P3 would have to resume from a checkpoint each time. For some apps (e.g. Einstein@home) resuming from a checking takes several minutes. So we'd end up wasting most of the time on one CPU.