Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Exploring DAOS Interfaces and Performance

Nicolau Manubens, Johann Lombardi, Simon D. Smart, Emanuele Danovaro, Tiago Quintino, Dean Hildebrand, Adrian Jackson

TL;DR

DAOS provides a number of interfaces for applications to undertake I/O, ranging from a native object storage API to a DAOS FUSE module for seamless compatibility with existing applications using POSIX file system APIs.

Abstract

Distributed Asynchronous Object Store (DAOS) is a novel software-defined object store leveraging Non-Volatile Memory (NVM) devices, designed for high performance. It provides a number of interfaces for applications to undertake I/O, ranging from a native object storage API to a DAOS FUSE module for seamless compatibility with existing applications using POSIX file system APIs. In this paper we discuss these interfaces and the options they provide, exercise DAOS through them with various I/O benchmarks, and analyse the observed performance. We also briefly compare the performance with a distributed file system and another object storage system deployed on the same hardware, and showcase DAOS' potential and increased flexibility to support high-performance I/O.

Exploring DAOS Interfaces and Performance

TL;DR

DAOS provides a number of interfaces for applications to undertake I/O, ranging from a native object storage API to a DAOS FUSE module for seamless compatibility with existing applications using POSIX file system APIs.

Abstract

Distributed Asynchronous Object Store (DAOS) is a novel software-defined object store leveraging Non-Volatile Memory (NVM) devices, designed for high performance. It provides a number of interfaces for applications to undertake I/O, ranging from a native object storage API to a DAOS FUSE module for seamless compatibility with existing applications using POSIX file system APIs. In this paper we discuss these interfaces and the options they provide, exercise DAOS through them with various I/O benchmarks, and analyse the observed performance. We also briefly compare the performance with a distributed file system and another object storage system deployed on the same hardware, and showcase DAOS' potential and increased flexibility to support high-performance I/O.
Paper Structure (21 sections, 9 figures)

This paper contains 21 sections, 9 figures.

Table of Contents

  1. Introduction
  2. Methodology
  3. Benchmarks
  4. IORior
  5. HDF5hdf5
  6. Field I/Ofield-io
  7. fdb-hammerfdb-hammer
  8. Test system
  9. Results
  10. Hardware Bandwidth
  11. API performance comparison
  12. Scalability of DAOS and the APIs
  13. Performance of DAOS redundancy
  14. Comparison to POSIX file system performance
  15. Comparison to Ceph object store performance
  16. ...and 6 more sections

Figures (9)

  • Figure 1: Client node and process count optimisation results for IOR with the different DAOS APIs, against a 16-node DAOS instance.
  • Figure 2: Client node and process count optimisation results for IOR on DFUSE and DFUSE+IL with 1KiB I/O size, against a 16-node DAOS instance.
  • Figure 3: Client node and process count optimisation results for the different DAOS applications, against a 16-node DAOS instance.
  • Figure 4: Client node and process count optimisation results for IOR on libdaos and IOR/HDF5 on libdaos, against a 4-node DAOS instance.
  • Figure 5: Write scalability (top) and read scalability (bottom) of the DAOS APIs and applications, with no redundancy.
  • ...and 4 more figures