The BOINC Client Emulator

The BOINC Client Emulator (BCE) emulates a BOINC client attached to one or more projects. It predicts, in a few seconds, what the latest BOINC client will do over a period of days or months. By reporting problem situations to BOINC developers, you can help us fix bugs and improve performance.

Scenarios

The inputs to BCE, called scenarios, describe a particular computer and the project to which it's attached. A scenario consists of: You create a scenario by uploading these files using the Create a scenario button below. You can use the files from a running BOINC client to emulate that client.

You can modify these files, or create new ones, to study hypothetical scenarios (e.g. hosts with a large number of CPUs, hosts attached to a large number of projects, projects with very short or long jobs, and so on). See The BCE documentation for details.

Create a scenario

Simulations

You can run simulations based on existing scenarios (including scenarios created by other people). The parameters of a simulation include The outputs of a simulation include

Comments and reports

When you examine the results of a simulation, you may find places where the BOINC client made bad scheduling or work-fetch decisions. Or you find may places where the simulator doesn't seem to be working correctly. In such cases, please add a comment to the simulation, indicating the nature of the problem and the simulation time when it occurred.

Please post to the boinc_dev email list if


Existing scenarios

ID
Click to see simulations
Who When # Simulations Description
204 N0vas 27 Aug 2023 1 Yoyo ECM OOM TEST
203 Lucas Dobre 15 Jul 2023 2 As per https://github.com/BOINC/boinc/issues/5306
202 qiongmei521 5 Jul 2023 0 not use all cpus
201 Richard Haselgrove 28 May 2023 2 Investigating concurrent running of GPU, MT, and CPU tasks. Using the most recent relevant GitHub artifact under Linux (#5251), CPU over-commit is seen, as reported in issue #5254.
200 AnandBhat 15 May 2023 1 Test
199 Peter Hucker of the Scottish Boinc Team 28 Apr 2023 1 Primegrid excessive download
198 elevankoff 20 Apr 2023 1 kkk
197 kotenok2000 12 Oct 2022 1 Hangs on connecting to manager
196 kotenok2000 24 Aug 2022 1 When i disable leave non-gpu tasks in memory and suspend computations apps stay in memory.
195 taranis 5 Aug 2022 1 Only about 25% of CPU capacity being utilised. Scheduler debug log reporting all CPUs in use and skipping tasks.
194 skydivingnerd 8 Mar 2022 1 Amaranth work fetch test with Rosetta and LHC
193 skydivingnerd 8 Mar 2022 2 Joruth host for LHC@home project
192 Mad_Max 1 Dec 2021 7 Max concurrent bug test
191 Mad_Max 30 Nov 2021 1 Max concurrent bug test
190 Mad_Max 30 Nov 2021 1 Max concurrent bug test
189 ZeNo 26 Oct 2021 5 ros/ein kombi
188 HeatForScience 28 Jan 2021 2 BOINC Client 7.16.11 (And 7.16.14 on OS X) schedule too many cores (Example has tasks (1 GPU + 1CPU and 12 CPU Tasks) on 12 core system. On systems with more than one GPU, this intermittently results in a GPU task being skipped with an error like: "1/27/2021 4:53:58 PM | collatz | [cpu_sched_debug] skipping GPU job collatz_sieve_4404021a-de64-41b2-bbb3-0b8228e66814_0; CPU committed " Workaround that mostly works in my case: Add app_config.xml with project_max_concurrent to limit CPU tasks for any project enough let the GPU tasks get scheduled first. The CPU tasks then typically end up getting skipped instead. (I add project_max_concurrent to Rosetta and world community grid projects, but this isn't a perfect workaround).
187 Richard Haselgrove 10 Jan 2021 1 Minimal scenario (one project, no discrete GPU, intel GPU not used, no app_config file) for investigating MT work fetch.
186 Richard Haselgrove 2 Dec 2020 2 For further investigation of #4117.
185 Richard Haselgrove 28 Nov 2020 0 After test run locally - client actually overfetched.
184 Richard Haselgrove 3 Dec 2020 2 Starting state for work fetch bug exploration.
183 Jord 20 Jul 2020 1 general test of rec_half_life_days set to 1
182 Jord 20 Jul 2020 2 rec_half_life = 1 test
181 Jord 20 Jul 2020 2 general rec_half_life_days = 1 test
180 Jord 20 Jul 2020 0 General test with rec_half_life = 1
179 Richard Haselgrove 22 May 2020 6 Testing: GPU exclusion leaves GPU starved
178 NataliaNik 6 May 2020 1 Client with 3 active projects
177 Paganini 19 Feb 2020 2 Test scenario
176 Sabrina Tarson 25 Jan 2020 1 Testing
175 jjv1001 2 Nov 2019 1 Multi-GPU scheduling
174 Keith Myers 21 Apr 2019 1 For #3076. Haven't seen anything out of the ordinary to report. max_concurrent for all projects obeyed
173 Richard Haselgrove 17 Apr 2019 2 For #3076. NumberFields has app_config with max_concurrent. Has no work cached, can fetch for CPU, but isn't.
172 Richard Haselgrove 5 Apr 2019 1 Test for #3076 - work buffering with max concurrent. Testing CPU work fetch. Only active CPU project is NumberFields. Machine was running on 3 CPU cores - fourth core used for SETI GPU support. Reduced CPU max_concurrent from 3 to 2. Machine has fetched one new task, "1.00 seconds; 0.40 devices" every minute for the last 20 minutes. Log to follow @GitHub
171 Richard Haselgrove 22 Mar 2019 1 For work fetch issue #3065 To explore REC calculation for different projects run on devices of different speeds.
170 additude 24 Feb 2019 3 Scheduler interaction between long duration short due date tasks and short duration long due date tasks
169 David Anderson 12 Jan 2019 2 S@h and E@h getting too many jobs?
168 Jacob Klein 4 Feb 2019 5 Revisiting GPU Exclusions - Work Fetch incorrectly detecting idle GPUs as not idle
167 Richard Haselgrove 31 Dec 2018 3 Another test of #2918. This is showing 31/12/2018 19:33:55 | NumberFields@home | [work_fetch] REC 2638.046 prio -1.000 can't request work: at max_concurrent limit when project has less work cached than current request level.
166 Richard Haselgrove 26 Dec 2018 2 Repeat of #165, using second app_config
165 Richard Haselgrove 21 Dec 2018 1 Testing PR #2918. Requires second app_config.xml: 1 for http://numberfields.asu.edu/NumberFields/
164 Keith Myers 21 Dec 2018 2 all cpu tasks forced to waiting to run
163 Keith Myers 19 Dec 2018 2 No cpu tasks are able to run with both and 16 in use
162 David Anderson 17 Dec 2018 2 test app config
161 Richard Haselgrove 16 Dec 2018 1 Second test for Keith Myers, #1677, with app_config for SETI.
160 Richard Haselgrove 15 Dec 2018 1 For Keith Myers, #1677, CS reduced
159 Jacob Klein 18 Nov 2018 2 Not asking for backup CPU work with less than 3 mins left and GPUs suspended - Try 2
158 Jacob Klein 17 Nov 2018 2 Not asking for backup CPU work with less than 3 mins left and GPUs suspended - Try 1
157 Peter Humphrey 25 Jun 2018 1 First run of client emulator.
156 Hibou 15 Dec 2017 1 Test Adrie
155 Toby Broom 18 Sep 2017 6 24 Job Limit
154 Toby Broom 18 Sep 2017 2 Hung VM
153 Dennis Regan 29 Aug 2017 3 Gorkskie
152 AndCycle 11 Aug 2017 1 test https://boinc.berkeley.edu/dev/forum_thread.php?id=11772
151 Darrell 24 May 2017 1 Boinc is running thee cpu tasks from Seti_beta and one gpu task from various projects. With compute preference set to 84% five cores should be use, but boinc is only using three.
150 Yair Givoni 7 Mar 2017 1 Fetching GPU WUs
149 ??? 4 Feb 2017 1 Check scheduler policy using in app_config.xml to limit cores usage by mt app
148 Jord 17 Jun 2016 1 Running a new entry to see if the simulator will say tasks are in deadline trouble vs 7.6.33 claiming that.
147 Jord 16 Jun 2016 3 Once more to run sims of rr_sim, against 7.6.33
146 Jord 16 Jun 2016 5 Testing rr_sim
145 Kiska 22 Feb 2016 1 Exploring why client grabs more work then it can process.
144 Kiska 21 Feb 2016 2 Checking the deadline, if 100 Generalized Fermat tasks will complete on time. Scheduler decided to grab 100 of these.
143 Kiska 21 Feb 2016 3 Work fetch debug flags not shown in event log
142 ??? 6 Feb 2016 2 mw simulated credit
141 Richard Haselgrove 6 Feb 2016 1 To explore why CPU and intel_gpu tasks are not scheduled, while an NVidia GPU exclusion is active.
140 Jacob Klein 7 Nov 2015 1 7.6.14 - Erroneous Backup Project requests - Try 1
139 Richard Haselgrove 14 Aug 2015 1 Task starvation with single project, multiple GPUs
138 ??? 25 Mar 2015 3 milkyway only uses 1 core leaving 7 idle on BOINC 7.2.42 & Linux version 3.10.0-123.13.2.el7.x86_64
137 Jacob Klein 17 Feb 2015 1 BOINC 7.4.36 not properly asking GPUGrid for work, to fill idle NVIDIA device 0.
136 Jon Fox 14 Jan 2015 2 7.4.36 not properly restricting number of wu based on app_config.xml for Collatz (Solo_Collatz)
135 Billy 14 Jan 2015 1 iMac with 4 CPU's and 1 ATI GPU connected to several projects with many CPU tasks and GPU tasks available
134 Jacob Klein 16 Dec 2014 1 7.4.36 - Idle GPU (device 0) while upload is in progress
133 ??? 7 Dec 2014 0 GPUGRID will not fetch tasks
132 Jacob Klein 25 Nov 2014 4 New scenario to test 1024 asic_miner devices :)
131 Jacob Klein 25 Nov 2014 2 New scenario to test 65 asic_miner devices
130 Jacob Klein 25 Nov 2014 2 New scenario to test 64 asic_miner devices
129 Jacob Klein 25 Nov 2014 2 New scenario to test 63 asic_miner devices (2nd try)
128 Jacob Klein 25 Nov 2014 1 New scenario to test 63 asic_miner devices
127 Jacob Klein 20 Nov 2014 2 BOINC 7.4.27 - Work Fetch erroneously says "job cache full" and doesn't request work, even though some of the 32 miner_asic instances are idle.
126 Jacob Klein 15 Nov 2014 8 7.4.27 bug - Work Fetch vs. 0 resource share - Not properly requesting work from all resource types.
125 Jord 11 Nov 2014 1 Trying a scenario where BOINC runs one project on a GPU, 2 projects on the CPU. One of the projects on the CPU has no work, the other project at RS 0 should ask for work, but doesn't. Seti RS800: GPU. Pirates RS800: CPU. WCG RS0: CPU.
124 MarkJ 18 Oct 2014 3 7.4.22 reports results immediately when it shouldn't
123 Jacob Klein 11 Oct 2014 2 7.4.22 Client Scheduler -- Not starting my MT job, leaving CPUs idle
122 Billy 25 Sep 2014 1 7.4.22 Oproject nnt
121 Billy 25 Sep 2014 1 7.4.22
120 Richard Haselgrove 3 Jul 2014 3 v7.3.19 - exploring why Albert@home isn't running NVIDIA tasks in FIFO order (Perseus Arm preferred above Arecibo GPU)
119 rebirther 21 Jun 2014 2 7.4.2 - Bitcoin Utopia NCI app is enable to get work, Primegrid is running on 6 cores with CPU + one other WUprop NCI WU. Always no work fetch with a low work cache size for Bitcoin Utopia NCI app. Mostly no work fetch with a higher work cache size for NCI.
118 Jacob Klein 14 Jun 2014 1 7.4.0 Work Fetch - Fetches GPU work from 0-resource-share projects even when no GPUs are idle
117 Jacob Klein 11 Jun 2014 1 GPUs idle, but BOINC 7.3.19 not performing work-fetch. Was expecting POEM@Home to be asked for work.
116 Jacob Klein 20 May 2014 5 7.3.19 Work Fetch is not fetching a task for the idle GPU (GTS 240). I would have expected it to fetch from either Albert/Einstein/SETI/SETIBETA. Could the "Ready to report" Einstein task be causing a problem?
115 Darrell 25 Apr 2014 1 Run with Boinc showing two opencl tasks running on an HD3000 and two running on an HD5850
114 William 5 Mar 2014 1 max_concurrent workfetch problems
113 Jacob Klein 9 Dec 2013 11 7.2.34 - MT tasks not scheduled, and work fetch leaves idle CPUs
112 Jacob Klein 25 Jun 2013 3 7.1.18 - NCI project has been marked as No New Tasks, yet work fetch requests/gets a task.
111 MarkJ 22 Jun 2013 1 Work fetch starvation
110 Jacob Klein 18 Jun 2013 5 7.1.17 new idle-GPU-work-fetch code skips certain projects when asking for GPU work
109 Jacob Klein 17 Jun 2013 2 Attempt 2 (added POEM available section in XML) to recreate issue where NVIDIA GPU is idle, and BOINC work fetch skips over 0-priority projects when asking projects for NVIDIA
108 Jacob Klein 17 Jun 2013 1 Attempt to recreate issue where NVIDIA GPU is idle, and BOINC work fetch skips over 0-priority projects when asking projects for NVIDIA
107 Jacob Klein 15 Jun 2013 2 7.1.17 Work Fetch GPU Exclusion bug - Device 0 Idle, should request NVIDIA work from POEM, but doesn't!
106 Richard Haselgrove 9 Jun 2013 2 Interested to see what I can expect from the MilkyWay MT task I have downloaded, with a 12 day deadline and a 4 YEAR estimated runtime. NB I have both app_info.xml and app_config.xml files active in the real world, so this may not proceed exactly as expected.
105 Toralf Foerster 30 May 2013 0 pairing of jobs
104 Jacob Klein 29 May 2013 7 BOINC v7.0.2 - GPU Idle - Work fetch not asking for work.
103 Jacob Klein 20 May 2013 1 7.1.1 Additional Work Fetch problem (SETI suspended, POEM suspended, expected GPUGrid work)
102 Jacob Klein 16 May 2013 4 7.1.1 Work Fetch problem: Low buffer settings + GPU Exclusions = GPU left idle (modified client_state.xml so POEM does not get work)
101 Jacob Klein 16 May 2013 2 7.1.1 Work Fetch problem: Low buffer settings + GPU Exclusions = GPU left idle
100 Augustine 16 May 2013 1 #8362
69 Beyond 27 Feb 2013 11 Client Ca__o with exclude statement & app_info
68 MarkJ 27 Feb 2013 9 not honour RS 7.0.52
67 Darrell 16 Feb 2013 7 Sim with ati changed to ATI in config
66 Darrell 15 Feb 2013 1 Repeat of run 65, but starting with PrimeGrid and Collatz
65 Darrell 15 Feb 2013 2 1 gpu opencl capabable, 1 gpu not. exclusions not holding test
64 Toralf Foerster 8 Feb 2013 7 Gentoo n22 just a test
63 Jacob Klein 7 Feb 2013 2 BOINC v7.0.47 - Task Scheduler - Starving GPU Device 0
62 -ShEm- 7 Feb 2013 2 Checking HP
61 Jacob Klein 19 Jan 2013 8 Boinc v7.0.44 GPU Starvation
60 Jacob Klein 9 Jan 2013 1 7.0.42 CPU Starvation (possibly due to not having enough memory to start existing tasks, but still should ask projects for more work)
59 Jacob Klein 8 Jan 2013 1 7.0.42 CPU starvation -- possibly due to having several tasks of a suspended project
58 Jacob Klein 19 Oct 2012 10 BOINC v7.0.36 not fetching work for 2 idle GPUs that have some configurations set
57 ??? 12 Sep 2012 10 check
56 GPV67 6 Jul 2012 1 Mac Pro 2010
55 John McLeod VII 29 Mar 2012 3 CPDN getting run too frequently.
54 wiyosaya 25 Feb 2012 3 GPUGRID, Milkyway, WCG. BOINC ignoring fractional additional work buffer
53 arkayn 11 Feb 2012 2 Backup projects
52 MarkJ 3 Feb 2012 1 7.0.14 GPU work fetch
51 Jord 21 Jan 2012 2 Testing work fetch, on CI 0.1 + AD 3.5, CPU + GPU.
50 arkayn 20 Jan 2012 2 BOINC 7.0.11 requesting work from backup projects when it does not need too.
49 MarkJ 14 Jan 2012 1 7.0.8 (lack of) work fetch
48 Tapio 1 Jan 2012 2 Test
47 MarkJ 31 Dec 2011 1 Net work suspended (after 8 hours) WFD
46 MarkJ 30 Dec 2011 1 Network suspend (start) work fetch
45 MarkJ 27 Dec 2011 1 MarkJ's 7.0.3 #1
44 Jacob Klein 16 Dec 2011 3 7.0.3 Bug - GPU Starvation
43 Billy 5 Dec 2011 1 7.02 Mac 10.4 deadlines
42 Jord 1 Dec 2011 3 Oops, should pause key tasks to be tested.
41 Jord 1 Dec 2011 1 Once more, more vigorous.
40 Jacob Klein 1 Dec 2011 12 6.13.12 Bug - Scheduler not obeying REC and requesting tasks for only 1 project (World Community Grid)
19 Jord 18 Oct 2011 5 Running 6.13.8 to check work fetch problems.
18 Jord 11 Oct 2011 1 Testing possible Seti work overfetch.
17 Jord 11 Oct 2011 1 Testing possible Seti work overfetch.
16 Jimmy G (BA) 30 Sep 2011 3 iMac 12,2 OSX 10.6.8 BOINC Mgr 6.12.35 (86) -Investigate WU mischief's
15 David Anderson 26 Sep 2011 13 GPU starvation, from Jacob Klein
14 John McLeod VII 19 Aug 2011 11 Current state multi CPU with long tasks.
13 Jacob Klein 23 Jul 2011 3 Want to see what 6.13.1 scheduler has in plans for my jobs
12 Richard Haselgrove 29 Jun 2011 2 CPDN showing unexpected 'high priority' (53 day estimate, 90 day deadline)
11 ??? 7 Jun 2011 1 my file
10 Richard Haselgrove 5 Jun 2011 1 Exploring problem with app_version FLOPS estimates and 32b/64b choices with AQUA test app.
6 Richard Haselgrove 19 Apr 2011 3 Work fetch estimation and logging - MT apps
5 arkayn 8 Apr 2011 1 iMac T7200
4 Billy 6 Apr 2011 3 OSX 10.4
1 GPV67 6 Apr 2011 8 GPV MacPro v0


Copyright © 2024 University of California.
Permission is granted to copy, distribute and/or modify this document under the terms of the GNU Free Documentation License, Version 1.2 or any later version published by the Free Software Foundation.