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.

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

FEAT-#285: Implement shared object storage for MPI backend #286

Merged
merged 46 commits into from
Sep 25, 2023

Conversation

Retribution98
Copy link
Collaborator

@Retribution98 Retribution98 commented Apr 25, 2023

What do these changes do?

Implementing shared storage to share data between workers instead of directly sending data.

Subtasks to do:

  • Add shared memory allocation
  • Add shared memory allocation on cluster
  • Add an additional communication message type to send information about shared data
  • Add procedure for reserving space in shared memory (Option 1 via Monitor worker)
  • Add procedure for reserving space in shared memory (Option 2 via extra shared space)
  • Add writing serialized data to shared memory and reading it from there
  • Add environment variable for shared memory usage threshold
  • Add shared data communication between different hosts in a cluster
  • Add data cleanup in shared memory
  • Check performance boost
  • Fix slowdown on high CPU count (there are two thoughts about the issue: 1) the previous approach is more beneficial in certain cases, 2) high-memory consumption)

  • first commit message and PR title follow format outlined here

    NOTE: If you edit the PR title to match this format, you need to add another commit (even if it's empty) or amend your last commit for the CI job that checks the PR title to pick up the new PR title.

  • passes flake8 .
  • passes black .
  • signed commit with git commit -s
  • Resolves [MPI] Implement shared object storage per node #285
  • tests added and passing
  • module layout described at docs/developer/architecture.rst is up-to-date

docs/developer/architecture.rst Outdated Show resolved Hide resolved
docs/developer/architecture.rst Outdated Show resolved Hide resolved
docs/flow/unidist/core/backends/mpi/core/object_store.rst Outdated Show resolved Hide resolved
unidist/config/backends/mpi/envvars.py Outdated Show resolved Hide resolved
unidist/core/backends/mpi/core/controller/api.py Outdated Show resolved Hide resolved
unidist/core/backends/mpi/core/object_store.py Outdated Show resolved Hide resolved
docs/flow/unidist/core/backends/mpi/core/object_store.rst Outdated Show resolved Hide resolved
unidist/core/backends/mpi/core/monitor/shared_store.py Outdated Show resolved Hide resolved
unidist/core/backends/mpi/core/object_store.py Outdated Show resolved Hide resolved
@Retribution98 Retribution98 force-pushed the FEAT-285 branch 3 times, most recently from 5c6a153 to 3a0598e Compare June 19, 2023 15:52
@Retribution98 Retribution98 force-pushed the FEAT-285 branch 3 times, most recently from cfef60d to 5010735 Compare June 26, 2023 16:16
@Retribution98 Retribution98 force-pushed the FEAT-285 branch 3 times, most recently from 76bd87c to de0e548 Compare July 11, 2023 10:53
@Retribution98 Retribution98 force-pushed the FEAT-285 branch 2 times, most recently from 52cae84 to 2c2cd45 Compare July 18, 2023 16:54
@Retribution98 Retribution98 force-pushed the FEAT-285 branch 2 times, most recently from 4814658 to d7c7f74 Compare July 21, 2023 14:12
setup.py Show resolved Hide resolved
unidist/core/backends/mpi/core/communication.py Outdated Show resolved Hide resolved
unidist/core/backends/mpi/core/communication.py Outdated Show resolved Hide resolved
unidist/core/backends/mpi/core/shared_object_store.py Outdated Show resolved Hide resolved
unidist/core/backends/mpi/core/shared_object_store.py Outdated Show resolved Hide resolved
unidist/core/backends/mpi/core/shared_object_store.py Outdated Show resolved Hide resolved
unidist/core/backends/mpi/core/shared_object_store.py Outdated Show resolved Hide resolved
unidist/core/backends/mpi/core/shared_object_store.py Outdated Show resolved Hide resolved
unidist/core/backends/mpi/utils.py Outdated Show resolved Hide resolved
unidist/ext_modules/memory/README.md Outdated Show resolved Hide resolved
unidist/core/backends/mpi/core/common.py Outdated Show resolved Hide resolved
unidist/core/backends/mpi/core/common.py Outdated Show resolved Hide resolved
unidist/core/backends/mpi/core/common.py Outdated Show resolved Hide resolved
unidist/ext_modules/memory/cmemory.pyx Outdated Show resolved Hide resolved
unidist/ext_modules/memory/cmemory.pyx Outdated Show resolved Hide resolved
unidist/ext_modules/memory/memory.cpp Outdated Show resolved Hide resolved
unidist/ext_modules/memory/memory.h Outdated Show resolved Hide resolved
unidist/ext_modules/memory/memory.pxd Outdated Show resolved Hide resolved
@YarShev YarShev changed the title FEAT-#285: Implement shared storage FEAT-#285: Implement shared object storage for MPI backend Sep 21, 2023
setup.py Outdated Show resolved Hide resolved
Retribution98 and others added 4 commits September 22, 2023 17:37
Signed-off-by: Igoshev, Iaroslav <[email protected]>
Signed-off-by: Igoshev, Iaroslav <[email protected]>
Signed-off-by: Igoshev, Iaroslav <[email protected]>
Signed-off-by: Igoshev, Iaroslav <[email protected]>
unidist/test/utils.py Outdated Show resolved Hide resolved
YarShev and others added 5 commits September 23, 2023 21:36
Signed-off-by: Igoshev, Iaroslav <[email protected]>
Signed-off-by: Igoshev, Iaroslav <[email protected]>
Signed-off-by: Igoshev, Iaroslav <[email protected]>
Signed-off-by: Igoshev, Iaroslav <[email protected]>
Copy link
Collaborator

@YarShev YarShev left a comment

Choose a reason for hiding this comment

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

@Retribution98, awesome changes! Now the MPI backend supports the shared object storage. Keep it up. Thanks!

@YarShev YarShev merged commit c5c10ab into modin-project:master Sep 25, 2023
23 checks passed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

Successfully merging this pull request may close these issues.

[MPI] Implement shared object storage per node
3 participants