Skip to content
Book Open access

Shadix: A Distributed Message Broker Leveraging Shared Disaggregated CXL Memory

Sep 2026 · Proceedings of the 4th Workshop on Disruptive Memory Systems · pp. 64-72 · 0 citations · 17 references

TL;DR

Shadix, the first message broker that leverages CXL shared memory, is presented and it is shown that Shadix offers competitive performance with single-digit μs latency and saturates the throughput of the CXL hardware.

Abstract

Message brokers are a core component of many data processing systems. They distribute messages from a set of producers to a set of consumers. Because of their central role, they need to offer low latency and high throughput. One approach to improve message broker performance is to specialize the implementation for modern hardware technologies. Compute Express Link (CXL) is a new interconnect technology based on PCI Express for cache-coherent access to remote memory. Multiple servers can use CXL to interact with shared resources and communicate over shared memory. CXL communication has different performance characteristics compared to technologies commonly used for communication with message brokers, like TCP. We present Shadix, the first message broker that leverages CXL shared memory. We compare throughput and round-trip time of our message broker with Valkey, NATS, and Apache Pulsar, three state-of-the-art message brokers. Shadix offers competitive performance with single-digit μs latency and saturates the throughput of the CXL hardware.

Read PDF

Similar papers

Book Open access Sep 2026

Borges: A Low-Latency Distributed Shared Log on a CXL Memory/SSD Hybrid

Borges is presented, to the authors' knowledge, the first fault-tolerant shared log whose data and ordering state reside in a DRAM/SSD hybrid device accessed with Compute Express Link (CXL).

Hao-Wei Chen, Yi-Ming Xiang, Zhi-Peng Jia et al. · 0 citations

T !" : E ! icient Drop-in Networking for Database Systems

This work presents Tux, a kernel-bypass networking stack designed for database systems that achieves high performance without giving up compatibility and robustness, and implements T !" in a library called T !".

Xin-Jing Zhou, Viktor Leis, Xiang-Yao Yu et al. · 0 citations
Review Open access Sep 2026

Inter-Process Communication (IPC) Mechanisms for Client–Server Communication: Architectures, Performance Trade-Offs, and Emerging Paradigms

Client-server systems are built on inter-process communication (IPC) and, as a direct consequence of this technology, on performance, scalability, and security frameworks in modern computing environments, especially in cloud-native and microservices systems. This paper critically analyzes the IPC mechanisms employed...

Benson Ikechukwu Igbokwe · 0 citations
Conference Sep 2026

CoCoA: An in-CXL Switch Collective Communication Accelerator

The Message Passing Interface (MPI) is widely adopted in high-performance computing (HPC) systems to support scalable inter-process communication. While HPC systems are theoretically expected to offer high performance, they often fall short in practice due to the overhead from MPI. In particular, MPI collective communi...

Seon Young Kim, Hooyoung Ahn, Jaewon Lee et al. · 0 citations

S !"#$ : A Scalable and Resize-optimized Hash Index on Disaggregated Memory

A novel architecture called S !"#$, designed to enhance the performance of hash indexes in disaggregated memory, is introduced and the results show that S !"#$ outperforms state-of-the-art DM-optimized hash indexes by at most 6.7 → (RACE), 3.6 → (SepHash), and 1.8 → (Outback) in YCSB workloads, respectively.

Han-Tian Zha, Teng Ma, Bao-Tong Lu et al. · 0 citations
Conference Open access Sep 2026

Rethinking Near-Data Processing Execution: Enabling Device Collaboration with CXL Peer-To-Peer Communication

NDP-CP's task handoff between peers reduces latency by 12-46% over prior systems relying on host-mediated coordination, while deferring memory operations in NDP-CP leads to a 21-41% latency reduction over modern caching mechanisms for NUMA memory management without explicit data replication.

Shao-Peng Yang, Sandy Lin, Gong-Jin Sun et al. · 0 citations

We use cookies to run the site and, with your consent, for analytics and to show ads. See our Cookie Policy.