Home / Applications / AI/HPC Data Center & Cloud
Applications

AI/HPC Data Center & Cloud

OCS-based optical switching for reconfigurable, high-radix AI cluster networks.

Discuss your fabric View OCS models

The Challenge

AI clusters are outgrowing the electrical fabric: cabling, reconfiguration and power become the limiting factors long before the GPUs do.

AI Cluster Scale

GPU node counts keep climbing, and every added electrical spine layer increases switch power, latency and cost. Expanding an oversubscribed fat-tree also brings more cabling and configuration work.

Reconfiguration

Training, fine-tuning and inference each place different demands on the fabric. A fixed electrical topology cannot adapt — operators either strand bandwidth or overbuild for the worst case.

Cabling Complexity

High-radix electrical spines translate into thousands of fiber links to install, document and re-patch by hand. At cluster scale, manual re-cabling is slow, error-prone and hard to scale.

How We Fit In

An optical circuit switch replaces part of the electrical spine with a transparent, software-defined optical layer.

Dragonfly topology with OCS at the spine

OCS at the spine of a Dragonfly fabric — a full mesh with N+1 redundancy.

Where the OCS lives — the data-center fabric. Large AI/HPC clusters still rely heavily on electrical switch fabrics, where spine stages normally involve optical-to-electrical-to-optical conversion and the associated power and configuration overhead. A MEMS-based OCS from Triple-Stone replaces selected electrical spine stages with a transparent optical layer: micro-mirrors steer light from each incoming fiber directly to an outgoing fiber, creating a direct end-to-end optical path with no OEO conversion and no store-and-forward.

In a Dragonfly topology the OCS sits at the spine, forming a full mesh with N+1 redundancy so pods join or leave the fabric on demand. Because the optical path is transparent to protocol and data rate within the device’s operating bandwidth, one physical fabric can serve today's 400G/800G transceivers and later upgrades — and can be re-provisioned in milliseconds instead of re-patching fiber in the colo.

OCS link with circulator separating TX and RX on one fiber

One fiber, both directions — TX/RX separated by a circulator.

Single-fiber bidirectional links — fewer fibers and ports. An OCS link should use as little fiber as possible. Our planar and arrayed circulators let a single fiber carry both directions: the transmit and receive signals are routed through different ports inside the device. High isolation helps limit crosstalk, while low insertion loss preserves the link budget and reduces the number of fibers and OCS ports required.

From discrete 3-port devices to integrated arrays, we cover the full range of circulators an OCS link needs. The 8-in-1 arrayed circulator packs eight bidirectional paths into one high-density package for high-density OCS interfaces. At the high-port-count optical interface of an OCS, a 2D optical fiber collimator array couples multiple fibers into a compact array to meet high-density integration needs.

Multi-plane fabrics and the Shuffle Box. Beyond OCS, Triple-Stone also supplies Shuffle Boxes and flexible boards for multi-plane AI networks. Multi-plane networking assigns the different optical channels of the same server to separate, independent network planes, so a cluster no longer scales only by adding switching tiers. At the physical layer, the Shuffle Box organizes and re-maps fibers according to a predefined mapping, so servers can connect to several planes through the predefined mapping, larger GPU clusters can be supported, and training and inference traffic gets multiple transmission paths. A pre-designed fiber mapping also reduces manual patching and the risk of cabling errors and simplifies later expansion and maintenance. Traffic scheduling and failover are handled by the network interface cards and the switching fabric; when part of a path fails, the remaining paths keep carrying traffic, improving the reliability of training and inference networks.

MCF FIFO — denser optical interconnects for AI clusters. As AI data centers accommodate growing numbers of optical channels, space inside equipment becomes increasingly constrained, and cable management becomes increasingly difficult. Our MCF FIFO (fan-in / fan-out) connects four single-mode fibers to the four cores of one multi-core fiber, using space-division multiplexing to carry four optical channels over a single fiber. Its miniature, low-crosstalk design saves cabling space in high-density optical modules and inside equipment, while simplifying connections and cable management for AI cluster expansion.

OCS used as test equipment: any instrument routed to any DUT, non-blocking

One OCS shared by every instrument — any-to-any, non-blocking, routed in software.

OCS as test equipment — one switch, shared by every instrument. In a lab or on a production line, devices under test (DUTs) and the instruments that measure them — light sources, BERTs, power meters, or other instruments — are normally re-cabled by hand every time the test plan changes. Put a non-blocking OCS in between, and any instrument can be software-routed to any DUT on the opposite side: one OCS shared by every test station.

Instrument ports and DUT ports line up port for port, and the crossing paths are built and torn down in software, so the test topology is configured once and re-used across many devices. It is the same optical-switching core that drives a data-center fabric — here it switches test resources instead. Our OCS is developed fully in-house, including the core 2D FAU collimator arrays, so the any-to-any paths hold stable insertion loss and the system stays reliable in long-term test applications.

Products for This Application

300 × 300 OCS

MEMS-based, non-blocking optical circuit switching for AI-scale data center interconnects — in mass production, fully software-controlled via NETCONF and a REST API.

View details →

8-in-1 Circulator

Arrayed circulator integrating 8 independent circulators for dense, bidirectional OCS links.

View details →

Fiber Shuffle

Fiber shuffle flexible boards re-map high-fiber-count links between OCS and line cards.

View details →

MCF FIFO

Fan-in/fan-out between four single-mode fibers and one four-core fiber for channel management in high-density optical modules and equipment.

View details →
1,500+Employees
61,000 m²Floor Space
4Manufacturing Facilities
500+Customers
3R&D Centers
62Patents

Relevant products are qualified under Telcordia GR-1221 / GR-1073 reliability programs.

Building an AI fabric?
Tell us your port count and timeline.

Evaluation units ship worldwide. Custom wavelength range, connector type and chassis options available.

sales@triple-stone.com
+86 137 0968 6019

Verification code

Other Applications