mirror of https://github.com/BOINC/boinc.git
- client: code shuffling
- scheduler: fix typo in msg svn path=/trunk/boinc/; revision=16750
This commit is contained in:
David Anderson
parent
209d519fc4
commit
2dc7056ee0
|
|
@ -10392,3 +10392,15 @@ David 26 Dec 2008
|
||||||
|
|
||||||
client/
|
client/
|
||||||
work_fetch.cpp
|
work_fetch.cpp
|
||||||
|
|
||||||
|
David 30 Dec 2008
|
||||||
|
- client: code shuffling
|
||||||
|
- scheduler: fix typo in msg
|
||||||
|
|
||||||
|
client/
|
||||||
|
cs_scheduler.cpp
|
||||||
|
makefile_sim
|
||||||
|
sim.h
|
||||||
|
work_fetch.cpp
|
||||||
|
sched/
|
||||||
|
sched_send.cpp
|
||||||
|
|
|
||||||
|
|
@ -62,6 +62,7 @@ using std::string;
|
||||||
//
|
//
|
||||||
#define REPORT_DEADLINE_CUSHION ((double)SECONDS_PER_DAY)
|
#define REPORT_DEADLINE_CUSHION ((double)SECONDS_PER_DAY)
|
||||||
|
|
||||||
|
#ifndef SIM
|
||||||
|
|
||||||
// Write a scheduler request to a disk file,
|
// Write a scheduler request to a disk file,
|
||||||
// to be sent to a scheduling server
|
// to be sent to a scheduling server
|
||||||
|
|
@ -895,4 +896,137 @@ int CLIENT_STATE::handle_scheduler_reply(
|
||||||
return 0;
|
return 0;
|
||||||
}
|
}
|
||||||
|
|
||||||
|
#endif // SIM
|
||||||
|
|
||||||
|
void CLIENT_STATE::check_project_timeout() {
|
||||||
|
unsigned int i;
|
||||||
|
for (i=0; i<projects.size(); i++) {
|
||||||
|
PROJECT* p = projects[i];
|
||||||
|
if (p->possibly_backed_off && now > p->min_rpc_time) {
|
||||||
|
p->possibly_backed_off = false;
|
||||||
|
request_work_fetch("Project backoff ended");
|
||||||
|
}
|
||||||
|
}
|
||||||
|
}
|
||||||
|
|
||||||
|
void PROJECT::set_min_rpc_time(double future_time, const char* reason) {
|
||||||
|
if (future_time <= min_rpc_time) return;
|
||||||
|
if (next_rpc_time && (future_time > next_rpc_time)) return;
|
||||||
|
min_rpc_time = future_time;
|
||||||
|
possibly_backed_off = true;
|
||||||
|
if (log_flags.sched_op_debug) {
|
||||||
|
msg_printf(this, MSG_INFO,
|
||||||
|
"[sched_op_debug] Deferring communication for %s",
|
||||||
|
timediff_format(min_rpc_time - gstate.now).c_str()
|
||||||
|
);
|
||||||
|
msg_printf(this, MSG_INFO, "[sched_op_debug] Reason: %s\n", reason);
|
||||||
|
}
|
||||||
|
}
|
||||||
|
|
||||||
|
// Return true if we should not contact the project yet.
|
||||||
|
//
|
||||||
|
bool PROJECT::waiting_until_min_rpc_time() {
|
||||||
|
return (min_rpc_time > gstate.now);
|
||||||
|
}
|
||||||
|
|
||||||
|
// find a project that needs to have its master file fetched
|
||||||
|
//
|
||||||
|
PROJECT* CLIENT_STATE::next_project_master_pending() {
|
||||||
|
unsigned int i;
|
||||||
|
PROJECT* p;
|
||||||
|
|
||||||
|
for (i=0; i<projects.size(); i++) {
|
||||||
|
p = projects[i];
|
||||||
|
if (p->waiting_until_min_rpc_time()) continue;
|
||||||
|
if (p->suspended_via_gui) continue;
|
||||||
|
if (p->master_url_fetch_pending) {
|
||||||
|
return p;
|
||||||
|
}
|
||||||
|
}
|
||||||
|
return 0;
|
||||||
|
}
|
||||||
|
|
||||||
|
// find a project for which a scheduler RPC is pending
|
||||||
|
//
|
||||||
|
PROJECT* CLIENT_STATE::next_project_sched_rpc_pending() {
|
||||||
|
unsigned int i;
|
||||||
|
PROJECT* p;
|
||||||
|
|
||||||
|
for (i=0; i<projects.size(); i++) {
|
||||||
|
p = projects[i];
|
||||||
|
|
||||||
|
// project request overrides backoff
|
||||||
|
//
|
||||||
|
if (p->next_rpc_time && p->next_rpc_time<now) {
|
||||||
|
p->sched_rpc_pending = RPC_REASON_PROJECT_REQ;
|
||||||
|
p->next_rpc_time = 0;
|
||||||
|
} else {
|
||||||
|
if (p->waiting_until_min_rpc_time()) continue;
|
||||||
|
}
|
||||||
|
// if (p->suspended_via_gui) continue;
|
||||||
|
// do the RPC even if suspended.
|
||||||
|
// This is critical for acct mgrs, to propagate new host CPIDs
|
||||||
|
//
|
||||||
|
if (p->sched_rpc_pending) {
|
||||||
|
return p;
|
||||||
|
}
|
||||||
|
}
|
||||||
|
return 0;
|
||||||
|
}
|
||||||
|
|
||||||
|
PROJECT* CLIENT_STATE::next_project_trickle_up_pending() {
|
||||||
|
unsigned int i;
|
||||||
|
PROJECT* p;
|
||||||
|
|
||||||
|
for (i=0; i<projects.size(); i++) {
|
||||||
|
p = projects[i];
|
||||||
|
if (p->waiting_until_min_rpc_time()) continue;
|
||||||
|
if (p->suspended_via_gui) continue;
|
||||||
|
if (p->trickle_up_pending) {
|
||||||
|
return p;
|
||||||
|
}
|
||||||
|
}
|
||||||
|
return 0;
|
||||||
|
}
|
||||||
|
|
||||||
|
// find a project with finished results that should be reported.
|
||||||
|
// This means:
|
||||||
|
// - we're not backing off contacting the project
|
||||||
|
// - the result is ready_to_report (compute done; files uploaded)
|
||||||
|
// - we're within a day of the report deadline,
|
||||||
|
// or at least a day has elapsed since the result was completed,
|
||||||
|
// or we have a sporadic connection
|
||||||
|
//
|
||||||
|
PROJECT* CLIENT_STATE::find_project_with_overdue_results() {
|
||||||
|
unsigned int i;
|
||||||
|
RESULT* r;
|
||||||
|
|
||||||
|
for (i=0; i<results.size(); i++) {
|
||||||
|
r = results[i];
|
||||||
|
if (!r->ready_to_report) continue;
|
||||||
|
|
||||||
|
PROJECT* p = r->project;
|
||||||
|
if (p->waiting_until_min_rpc_time()) continue;
|
||||||
|
if (p->suspended_via_gui) continue;
|
||||||
|
|
||||||
|
if (config.report_results_immediately) {
|
||||||
|
return p;
|
||||||
|
}
|
||||||
|
|
||||||
|
if (net_status.have_sporadic_connection) {
|
||||||
|
return p;
|
||||||
|
}
|
||||||
|
|
||||||
|
double cushion = std::max(REPORT_DEADLINE_CUSHION, work_buf_min());
|
||||||
|
if (gstate.now > r->report_deadline - cushion) {
|
||||||
|
return p;
|
||||||
|
}
|
||||||
|
|
||||||
|
if (gstate.now > r->completed_time + SECONDS_PER_DAY) {
|
||||||
|
return p;
|
||||||
|
}
|
||||||
|
}
|
||||||
|
return 0;
|
||||||
|
}
|
||||||
|
|
||||||
const char *BOINC_RCSID_d35a4a7711 = "$Id$";
|
const char *BOINC_RCSID_d35a4a7711 = "$Id$";
|
||||||
|
|
|
||||||
|
|
@ -6,13 +6,15 @@ OBJS = \
|
||||||
app.o \
|
app.o \
|
||||||
client_msgs.o \
|
client_msgs.o \
|
||||||
cs_apps.o \
|
cs_apps.o \
|
||||||
|
cs_scheduler.o \
|
||||||
cpu_sched.o \
|
cpu_sched.o \
|
||||||
log_flags.o \
|
log_flags.o \
|
||||||
|
rr_sim.o \
|
||||||
sim.o \
|
sim.o \
|
||||||
sim_util.o \
|
sim_util.o \
|
||||||
time_stats.o \
|
time_stats.o \
|
||||||
work_fetch.o \
|
work_fetch.o \
|
||||||
edf_sim.o \
|
../sched/edf_sim.o \
|
||||||
../lib/coproc.o \
|
../lib/coproc.o \
|
||||||
../lib/mfile.o \
|
../lib/mfile.o \
|
||||||
../lib/miofile.o \
|
../lib/miofile.o \
|
||||||
|
|
|
||||||
|
|
@ -306,4 +306,4 @@ extern bool work_fetch_old;
|
||||||
|
|
||||||
#define CPU_PESSIMISM_FACTOR 0.9
|
#define CPU_PESSIMISM_FACTOR 0.9
|
||||||
// assume actual CPU utilization will be this multiple
|
// assume actual CPU utilization will be this multiple
|
||||||
// of what we've actually measured recently
|
// of what we've actually measured recently
|
||||||
|
|
|
||||||
|
|
@ -56,14 +56,6 @@ using std::max;
|
||||||
using std::vector;
|
using std::vector;
|
||||||
using std::string;
|
using std::string;
|
||||||
|
|
||||||
// quantities like avg CPU time decay by a factor of e every week
|
|
||||||
//
|
|
||||||
#define EXP_DECAY_RATE (1./(SECONDS_PER_DAY*7))
|
|
||||||
|
|
||||||
// try to report results this much before their deadline
|
|
||||||
//
|
|
||||||
#define REPORT_DEADLINE_CUSHION ((double)SECONDS_PER_DAY)
|
|
||||||
|
|
||||||
static const char* urgency_name(int urgency) {
|
static const char* urgency_name(int urgency) {
|
||||||
switch(urgency) {
|
switch(urgency) {
|
||||||
case WORK_FETCH_DONT_NEED: return "Don't need";
|
case WORK_FETCH_DONT_NEED: return "Don't need";
|
||||||
|
|
@ -85,97 +77,6 @@ int CLIENT_STATE::proj_min_results(PROJECT* p, double subset_resource_share) {
|
||||||
return (int)(ceil(ncpus*p->resource_share/subset_resource_share));
|
return (int)(ceil(ncpus*p->resource_share/subset_resource_share));
|
||||||
}
|
}
|
||||||
|
|
||||||
void CLIENT_STATE::check_project_timeout() {
|
|
||||||
unsigned int i;
|
|
||||||
for (i=0; i<projects.size(); i++) {
|
|
||||||
PROJECT* p = projects[i];
|
|
||||||
if (p->possibly_backed_off && now > p->min_rpc_time) {
|
|
||||||
p->possibly_backed_off = false;
|
|
||||||
request_work_fetch("Project backoff ended");
|
|
||||||
}
|
|
||||||
}
|
|
||||||
}
|
|
||||||
|
|
||||||
void PROJECT::set_min_rpc_time(double future_time, const char* reason) {
|
|
||||||
if (future_time <= min_rpc_time) return;
|
|
||||||
if (next_rpc_time && (future_time > next_rpc_time)) return;
|
|
||||||
min_rpc_time = future_time;
|
|
||||||
possibly_backed_off = true;
|
|
||||||
if (log_flags.sched_op_debug) {
|
|
||||||
msg_printf(this, MSG_INFO,
|
|
||||||
"[sched_op_debug] Deferring communication for %s",
|
|
||||||
timediff_format(min_rpc_time - gstate.now).c_str()
|
|
||||||
);
|
|
||||||
msg_printf(this, MSG_INFO, "[sched_op_debug] Reason: %s\n", reason);
|
|
||||||
}
|
|
||||||
}
|
|
||||||
|
|
||||||
// Return true if we should not contact the project yet.
|
|
||||||
//
|
|
||||||
bool PROJECT::waiting_until_min_rpc_time() {
|
|
||||||
return (min_rpc_time > gstate.now);
|
|
||||||
}
|
|
||||||
|
|
||||||
// find a project that needs to have its master file fetched
|
|
||||||
//
|
|
||||||
PROJECT* CLIENT_STATE::next_project_master_pending() {
|
|
||||||
unsigned int i;
|
|
||||||
PROJECT* p;
|
|
||||||
|
|
||||||
for (i=0; i<projects.size(); i++) {
|
|
||||||
p = projects[i];
|
|
||||||
if (p->waiting_until_min_rpc_time()) continue;
|
|
||||||
if (p->suspended_via_gui) continue;
|
|
||||||
if (p->master_url_fetch_pending) {
|
|
||||||
return p;
|
|
||||||
}
|
|
||||||
}
|
|
||||||
return 0;
|
|
||||||
}
|
|
||||||
|
|
||||||
// find a project for which a scheduler RPC is pending
|
|
||||||
//
|
|
||||||
PROJECT* CLIENT_STATE::next_project_sched_rpc_pending() {
|
|
||||||
unsigned int i;
|
|
||||||
PROJECT* p;
|
|
||||||
|
|
||||||
for (i=0; i<projects.size(); i++) {
|
|
||||||
p = projects[i];
|
|
||||||
|
|
||||||
// project request overrides backoff
|
|
||||||
//
|
|
||||||
if (p->next_rpc_time && p->next_rpc_time<now) {
|
|
||||||
p->sched_rpc_pending = RPC_REASON_PROJECT_REQ;
|
|
||||||
p->next_rpc_time = 0;
|
|
||||||
} else {
|
|
||||||
if (p->waiting_until_min_rpc_time()) continue;
|
|
||||||
}
|
|
||||||
// if (p->suspended_via_gui) continue;
|
|
||||||
// do the RPC even if suspended.
|
|
||||||
// This is critical for acct mgrs, to propagate new host CPIDs
|
|
||||||
//
|
|
||||||
if (p->sched_rpc_pending) {
|
|
||||||
return p;
|
|
||||||
}
|
|
||||||
}
|
|
||||||
return 0;
|
|
||||||
}
|
|
||||||
|
|
||||||
PROJECT* CLIENT_STATE::next_project_trickle_up_pending() {
|
|
||||||
unsigned int i;
|
|
||||||
PROJECT* p;
|
|
||||||
|
|
||||||
for (i=0; i<projects.size(); i++) {
|
|
||||||
p = projects[i];
|
|
||||||
if (p->waiting_until_min_rpc_time()) continue;
|
|
||||||
if (p->suspended_via_gui) continue;
|
|
||||||
if (p->trickle_up_pending) {
|
|
||||||
return p;
|
|
||||||
}
|
|
||||||
}
|
|
||||||
return 0;
|
|
||||||
}
|
|
||||||
|
|
||||||
// Return the best project to fetch work from, NULL if none
|
// Return the best project to fetch work from, NULL if none
|
||||||
//
|
//
|
||||||
// Pick the one with largest (long term debt - amount of current work)
|
// Pick the one with largest (long term debt - amount of current work)
|
||||||
|
|
@ -237,46 +138,6 @@ PROJECT* CLIENT_STATE::next_project_need_work() {
|
||||||
return p_prospect;
|
return p_prospect;
|
||||||
}
|
}
|
||||||
|
|
||||||
// find a project with finished results that should be reported.
|
|
||||||
// This means:
|
|
||||||
// - we're not backing off contacting the project
|
|
||||||
// - the result is ready_to_report (compute done; files uploaded)
|
|
||||||
// - we're within a day of the report deadline,
|
|
||||||
// or at least a day has elapsed since the result was completed,
|
|
||||||
// or we have a sporadic connection
|
|
||||||
//
|
|
||||||
PROJECT* CLIENT_STATE::find_project_with_overdue_results() {
|
|
||||||
unsigned int i;
|
|
||||||
RESULT* r;
|
|
||||||
|
|
||||||
for (i=0; i<results.size(); i++) {
|
|
||||||
r = results[i];
|
|
||||||
if (!r->ready_to_report) continue;
|
|
||||||
|
|
||||||
PROJECT* p = r->project;
|
|
||||||
if (p->waiting_until_min_rpc_time()) continue;
|
|
||||||
if (p->suspended_via_gui) continue;
|
|
||||||
|
|
||||||
if (config.report_results_immediately) {
|
|
||||||
return p;
|
|
||||||
}
|
|
||||||
|
|
||||||
if (net_status.have_sporadic_connection) {
|
|
||||||
return p;
|
|
||||||
}
|
|
||||||
|
|
||||||
double cushion = std::max(REPORT_DEADLINE_CUSHION, work_buf_min());
|
|
||||||
if (gstate.now > r->report_deadline - cushion) {
|
|
||||||
return p;
|
|
||||||
}
|
|
||||||
|
|
||||||
if (gstate.now > r->completed_time + SECONDS_PER_DAY) {
|
|
||||||
return p;
|
|
||||||
}
|
|
||||||
}
|
|
||||||
return 0;
|
|
||||||
}
|
|
||||||
|
|
||||||
// the fraction of time a given CPU is working for BOINC
|
// the fraction of time a given CPU is working for BOINC
|
||||||
//
|
//
|
||||||
double CLIENT_STATE::overall_cpu_frac() {
|
double CLIENT_STATE::overall_cpu_frac() {
|
||||||
|
|
|
||||||
|
|
@ -880,7 +880,7 @@ bool work_needed(bool locality_sched) {
|
||||||
if (g_wreq->nresults_on_host >= config.max_wus_in_progress*ncpus) {
|
if (g_wreq->nresults_on_host >= config.max_wus_in_progress*ncpus) {
|
||||||
if (config.debug_send) {
|
if (config.debug_send) {
|
||||||
log_messages.printf(MSG_DEBUG,
|
log_messages.printf(MSG_DEBUG,
|
||||||
"in-progress job limit exceeded; %d > %d*%d\n",
|
"in-progress job limit exceeded; %d >= %d*%d\n",
|
||||||
g_wreq->nresults_on_host, config.max_wus_in_progress, ncpus
|
g_wreq->nresults_on_host, config.max_wus_in_progress, ncpus
|
||||||
);
|
);
|
||||||
}
|
}
|
||||||
|
|
|
||||||
Loading…
Reference in New Issue