[hpc-announce] [CFP] HCDS 2026 – 5th Workshop on Heterogeneous Composable and Disaggregated Systems (co-located with ASPLOS 2026)

[Ren, Jie]

===============================================================
5th Workshop on Heterogeneous Composable and Disaggregated Systems (HCDS 2026)
March 22, 2026
Pittsburgh, USA

Co-located with ASPLOS 2026

https://urldefense.us/v3/__https://hcds-workshop.github.io/edition/2026/__;!!G_uCfscf7eWS!cE7JgyunjE_PEvDnkWnWdwCTUZVx-ugTDtfW2of691i3XBSQAybAlBVmgLxbB25BixyqDFsz9el_tDMe0pyZ2g$ 
===============================================================

Heterogeneous and Composable Disaggregated Systems (HCDS) are an emerging way to build computing platforms from pools of CPUs, memory, accelerators, and other devices that can be dynamically composed to match application needs. With commodity interconnects such as Compute Express Link (CXL), these systems can increase effective capacity, improve utilization, and reduce total cost of ownership.

The HCDS workshop provides a forum for researchers and practitioners from computer architecture, operating systems, runtime systems, data centers, and high-performance computing (HPC) to present and discuss new ideas in composable and disaggregated systems, from hardware and interconnects up through system software, programming models, and applications.

HCDS 2026 is soliciting original research papers on all aspects of heterogeneous, composable, and disaggregated systems.

Topics of interest include, but are not limited to:

Hardware and Prototyping
- Architectures for composable and disaggregated systems (e.g., CXL-based designs)
- Prototypes and proof-of-concept systems
- Interconnect technologies for disaggregation
- Memory pooling and memory disaggregation

Modeling and Evaluation
- Simulation frameworks for heterogeneous composable systems
- Characterization of composable systems (performance, energy, reliability)
- Algorithms and performance models for managing HCDS resources

System Software, Programming Models, and Tools
- Operating-system support for HCDS (e.g., memory profiling, page migration, allocation)
- Control-plane software and runtime systems for composing and managing resources
- Programming models and abstractions for composable memory and accelerators
- Analysis and profiling tools for composable/disaggregated systems
- Virtualization for composable heterogeneous platforms

Applications and Use Cases
- Deep learning and large language model workloads (training, inference, serving, fine-tuning)
- Agentic AI systems and data-intensive analytics
- Data input pipelines, streaming workloads, and scientific/HPC applications
- Disaggregated and distributed AI system design (e.g., distributed inference)
- CXL-based memory sharing and pooling for AI
- Co-design of AI models and disaggregated system architectures

Submission Guidelines

- Papers will be reviewed using a double-blind process. Submissions must not include author names or other identifying information.
- Manuscripts should be 4–6 pages in length, including figures and tables but excluding references.
- It is expected that at least one author of each accepted paper will register for and present the work at the workshop.

Accepted papers may be published in the ACM Digital Library under the ACM International Conference Proceedings Series (ICPS) open-access model. Authors who prefer not to appear in the official ACM proceedings (e.g., to enable extended submissions elsewhere or due to APC constraints) may opt out and have their accepted paper appear only on the workshop website; details are on the workshop page.

Important Deadlines (AoE)

- Paper submission deadline: January 23, 2026
- Author notification:       February 13, 2026
- Camera-ready deadline:     February 20, 2026
- Workshop:                  March 22, 2026

Submission Link
Please submit your papers at: https://urldefense.us/v3/__https://hcds26.hotcrp.com__;!!G_uCfscf7eWS!cE7JgyunjE_PEvDnkWnWdwCTUZVx-ugTDtfW2of691i3XBSQAybAlBVmgLxbB25BixyqDFsz9el_tDMxqF55HA$ 

Contact
If you have any questions, please contact: hcdsworkshop at gmail.com