Use SMC-R (kernel-mode) or NetACC (user-mode) to adapt applications to eRDMA with minimal or no code changes.
Use cases
eRDMA supports Redis, Spark, HPC, and Kafka scenarios. Transparent RDMA adaptation requires no code changes — only recompilation. Non-transparent adaptation requires code modifications.
-
Redis applications
Recommended: use SMC-R for transparent adaptation. Deploy SMC-R and Redis with eRDMA.
-
HPC applications
eRDMA enables multi-node parallel jobs in E-HPC NEXT clusters, such as climate meteorology, industrial simulation, and molecular dynamics, to achieve network performance comparable to on-premises clusters. Deploy an eRDMA-enabled E-HPC cluster.
-
Spark applications
Use jVerbs to modify Spark communication code. Deploy Spark cluster with eRDMA.
-
Kafka clusters
Deploy a Kafka cluster on eRDMA-capable ECS instances to accelerate inter-node data transfers. Deploy a high-network-performance Kafka cluster on eRDMA-enhanced instances.
Adaptation solutions
Kernel-mode adaptation solution: SMC-R
IBM open-sourced SMC-R for Linux 4.11 in 2017 (RFC 7609). Alibaba Cloud Linux 3 enables SMC-R in the cloud with eRDMA. SMC-R transparently replaces TCP without functionality loss, providing hardware-offloaded, high-performance networking.
User-mode adaptation solution: NetACC
NetACC is a user-mode network acceleration library. Load it via the LD_PRELOAD environment variable — no code changes required. NetACC leverages eRDMA benefits, such as low latency, high throughput, kernel bypass, and protocol stack offload, to accelerate TCP connection establishment through compatible socket interfaces.