Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

[GCS]Add gcs resource scheduler #13072

Merged
merged 32 commits into from
Jan 14, 2021
Merged

[GCS]Add gcs resource scheduler #13072

Show file tree
Hide file tree
Changes from 1 commit
Commits
Show all changes
32 commits
Select commit Hold shift + click to select a range
bd47564
add part code
Dec 10, 2020
b78c894
rebase master
Dec 12, 2020
4170043
add part code
Dec 10, 2020
8578907
add part code
Dec 11, 2020
58b5d85
fix code style
Dec 13, 2020
4897d7f
fix ut bug
Dec 13, 2020
c6333ce
add part code
Dec 9, 2020
600bd43
add part code
Dec 10, 2020
3a788e6
add part code
Dec 10, 2020
b07f369
add part code
Dec 14, 2020
003d6e2
add part code
Dec 14, 2020
d68e9df
rebase master
Dec 24, 2020
6cec27c
add part code
Dec 27, 2020
910d186
rebase master
Dec 27, 2020
b78a1d2
add part code
Dec 27, 2020
d5bc7ff
add part code
Dec 28, 2020
7dd33b9
add part code
Dec 28, 2020
71a8547
add part code
Dec 29, 2020
fcf4bab
add part code
Dec 29, 2020
1a003bc
add part code
Dec 30, 2020
368dab5
add part code
Dec 30, 2020
c13ebfe
add part code
Dec 30, 2020
6d77661
add part code
Dec 30, 2020
ff5bf3b
add part code
Dec 30, 2020
0cab178
fix review comments
Jan 5, 2021
9101161
fix review comments
Jan 7, 2021
3ce8e41
fix review comment
Jan 7, 2021
4d1bc9b
fix review comments
Jan 7, 2021
ec7667e
fix review comment
Jan 7, 2021
351d4ec
add testcase
Jan 7, 2021
d2f9c85
fix review comments
Jan 7, 2021
f58c95d
fix review comment
Jan 8, 2021
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
Prev Previous commit
Next Next commit
add part code
  • Loading branch information
灵洵 committed Dec 27, 2020
commit b07f369fad09324d73836ee5ba12c3d883db5494
39 changes: 28 additions & 11 deletions src/ray/gcs/gcs_server/gcs_resource_scheduler.cc
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -53,26 +53,39 @@ std::vector<NodeID> GcsResourceScheduler::Schedule(
const auto &cluster_resources = gcs_resource_manager_.GetClusterResources();

// Filter candidate nodes.
absl::flat_hash_set<NodeID> candidate_nodes = Filter(cluster_resources, node_filter_func);
absl::flat_hash_set<NodeID> candidate_nodes =
FilterCandidateNodes(cluster_resources, node_filter_func);
if (candidate_nodes.size() < required_resources.size()) {
return {};
}

// TODO(ffbin): Filter required resources.
// First schedule scarce resources (such as GPU) and large capacity resources to improve
// the scheduling success rate.
const auto &to_schedule_resources = SortRequiredResources(required_resources);

// Scorer.
// std::vector<NodeScore> node_score_list;
// for (const auto &iter : cluster_resources) {
// double node_score = node_scorer_->MakeGrade(required_resources, iter->second);
// node_score_list.emplace_back(NodeScore(iter->first, node_score));
// }
// Score and rank nodes.
switch (policy.type_) {
case SPREAD:
break;
case STRICT_SPREAD:
break;
case PACK:
break;
case STRICT_PACK:
break;
default:
break;
}

for (const auto &resource : to_schedule_resources) {

}

// TODO(ffbin): Rank.
std::vector<NodeID> result;
return result;
}

absl::flat_hash_set<NodeID> GcsResourceScheduler::Filter(
absl::flat_hash_set<NodeID> GcsResourceScheduler::FilterCandidateNodes(
const absl::flat_hash_map<NodeID, ResourceSet> &cluster_resources,
std::function<bool(const NodeID &)> node_filter_func) {
absl::flat_hash_set<NodeID> result;
Expand All @@ -84,7 +97,11 @@ absl::flat_hash_set<NodeID> GcsResourceScheduler::Filter(
}
return result;
}
//////////////////////////////////// End of GcsResourceScheduler ////////////////////////////////

std::vector<ResourceSet> GcsResourceScheduler::SortRequiredResources(
const std::vector<ResourceSet> &required_resources) {}

/////////////////////////////// End of GcsResourceScheduler ///////////////////////////

} // namespace gcs
} // namespace ray
15 changes: 10 additions & 5 deletions src/ray/gcs/gcs_server/gcs_resource_scheduler.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -80,18 +80,23 @@ class LeastResourceScorer : public NodeScorer {
class GcsResourceScheduler {
public:
GcsResourceScheduler(GcsResourceManager &gcs_resource_manager)
: gcs_resource_manager_(gcs_resource_manager), node_scorer_(new LeastResourceScorer()) {}
: gcs_resource_manager_(gcs_resource_manager),
node_scorer_(new LeastResourceScorer()) {}
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

if NodeScorer is Pluggable, I think we should add a ray config here?

Copy link
Contributor Author

@ffbin ffbin Jan 5, 2021 •
edited
Loading

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Good idea, but there is only one implementation at the moment, so i prefer to add configuration after having other implementations, thx.

Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I guess we'll want more dynamic configuration than ray config for this potentially right?

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Yeah, i find it is hard to choose a specific scorer type with ray config.


virtual ~GcsResourceScheduler() = default;

std::vector<NodeID> Schedule(
std::vector<ResourceSet> required_resources, SchedulingPolicy policy,
const std::function<bool(const NodeID &)> &node_filter_func);
const std::vector<ResourceSet> &required_resources, SchedulingPolicy policy,
const std::function<bool(const NodeID &)> &node_filter_func,
const std::function<std::vector<NodeID>(const std::vector<NodeScore> &)> &node_rank_func);

private:
absl::flat_hash_set<NodeID> Filter(
absl::flat_hash_set<NodeID> FilterCandidateNodes(
const absl::flat_hash_map<NodeID, ResourceSet> &cluster_resources,
std::function<bool(const NodeID &)> node_filter_func);
const std::function<bool(const NodeID &)> &node_filter_func);

std::vector<ResourceSet> SortRequiredResources(
const std::vector<ResourceSet> &required_resources);

/// Reference of GcsResourceManager.
GcsResourceManager &gcs_resource_manager_;
Expand Down