Shifting to a new scale
Shifting to a new scale
Exascale entails an explosion of performance, of the number of nodes/cores, of data volume and data movement. At such a scale, optimizing the network that is the backbone of the system becomes a major contributor to global performance.
The interconnect is going to be a key enabling technology for exascale systems. This is why one of the cornerstones of Atos’ eXascale program is the development of our own new-generation interconnect.
The BullSequana eXascale Interconnect V2 or BXI V2 introduces a paradigm shift in terms of performance, scalability, efficiency, reliability and quality of service for extreme workloads.
BXI V2 components
The BXI V2 fabric relies on two types of ASICs as its building blocks, a Network Interface Controller (NIC) and a switch, and comes with its complete software suite. BXI V2 switches are managed through a distributed and out-of-band fabric management suite allowing to scale up to 64K nodes. Out-of-band management eliminates any interference of the management traffic with the applications.
Interconnect for hybrid platforms
While BXI V1 was dedicated to BullSequana X1000 systems, BXI V2 is available on all BullSequana X platforms: it integrates the latest industry leading technologies within all BullSequana X servers, from BullSequana X400 mid-range series to BullSequana XH2000 high-end systems. Moreover, BXI V2 Network Interface Controller supports AMD, and ARM compute nodes, in addition to Intel ones.
The highest level of communication performance for HPC applications
BXI V2’s main feature is full hardware offload of communications. Indeed, by offloading to the hardware components, BXI V2 communications do not interfere with computations.
Offload boosts the communication performance of HPC applications, enabling CPUs to be fully dedicated to computational tasks while communications are independently managed by BXI V2.
As a result, BXI V2 can deliver high communication performance considering bandwidth, latencies, or message rates.
Advanced Fabric Management (AFM) is responsible for monitoring and routing a BXI V2 fabric while IBMS handles Device Management and Performance Management tasks.
With a new patented routing algorithm, AFM boosts the performance of BXI V2. BXI V2 allows a wide range of topologies: Generalized Fat-Tree, and DragonFly +.
AFM and IBMS software are fully compliant within SuperComputer Suite, a scalable, and robust software suite that meets the requirements including the need for enhanced security of even the most challenging high-performance computing environments.
A wide ecosystem available for application environment
The application environment of BXI has been significantly developed between V1 and V2. BXI V2 now enables Intel MPI library support thanks to an OFI provider.
Moreover, BXI V2 brings new features such as:
- GPU direct
- Virtualization (up to 16VMs)