SOVEX
CBDC Data Centers Sovereign AI Tokenization Deep Tech Architecture About Team Request access
AI Data Centers / Networking / Non-blocking topology

Non-blocking topology.

A cluster that looks fast on paper can still crawl if two flows have to fight for the same link. Non-blocking topology is the guarantee that every GPU can reach every other GPU at full rate, no matter what else is on the wire.

Non-blocking means bisection bandwidth scales with the machine

The property that defines a training fabric is how much traffic can cross its narrowest cut at once.

01

Bisection bandwidth

Split the cluster in half and measure the bandwidth across the cut. In a non-blocking design that number scales with node count, so an all-to-all exchange does not saturate a chokepoint as the machine grows.

02

Oversubscription is a policy

Oversubscription is the deliberate ratio of downlink to uplink capacity at a switch tier. Non-blocking sets that ratio at one-to-one on the paths training traffic uses, spending capacity where collectives demand it.

03

Congestion is architectural

Most training congestion is not bad luck, it is a topology that cannot carry the offered load. We treat contention as something the wiring plan eliminates, not something congestion control is left to paper over.

04

Predictable step time

A non-blocking fabric gives every collective the same bandwidth regardless of which nodes participate. That predictability is what lets a long training run hold a stable step time from the first epoch to the last.

The fat-tree is a Clos network that fattens links toward the root

A properly provisioned fat-tree offers full bisection bandwidth between any pair of endpoints.

01

Leaf and spine tiers

GPUs attach to leaf switches, leaves connect to spines, and the aggregate uplink capacity at each tier equals the downlink capacity below it. No tier becomes a funnel that throttles the layer beneath it.

02

Equal-cost multipath

Between any two endpoints there are many equal-cost paths through the spine. Traffic is spread across them so no single core link carries a disproportionate share of a large collective.

03

Full or tapered fabrics

A full fat-tree is non-blocking end to end; a tapered fabric oversubscribes upper tiers where the traffic pattern permits. The taper is chosen against the model's actual communication profile, not applied blindly to save switches.

04

Modular growth

Clos topologies extend by adding pods and spine planes. Capacity grows in defined increments, so a sovereign operator can expand a cluster without re-architecting the fabric that already exists.

Rail-optimized wiring aligns the fabric to how GPUs actually communicate

Modern training traffic is structured, and the topology is built to match that structure exactly.

01

Rails follow GPU ordinals

Each GPU in a server connects to a dedicated rail, and the same-ordinal GPUs across all servers share a rail plane. Collectives that exchange between matching ranks stay on one plane and avoid crossing the whole fabric.

02

Traffic localized per plane

Because ring and tree collectives communicate along ranks, rail alignment keeps most of that traffic on a single rail switch. The core fabric carries only the residual, reducing the load on the most expensive links.

03

Deterministic placement

The scheduler assigns ranks to GPUs so that the job's communication pattern lands on the rails designed to carry it. Placement is a first-class input to performance, not left to chance.

04

Intra-node NVLink tier

Inside a server, GPUs exchange over a dedicated high-bandwidth link mesh before anything touches the rails. The topology is hierarchical by design: keep traffic on the fastest tier that can hold it.

Static balance is not enough; the fabric steers around live hotspots

Even a non-blocking topology can hash two heavy flows onto one link, so routing adapts.

01

Flow-hash collisions

Naive multipath hashing can place two large flows on the same uplink while others sit idle. Non-blocking capacity is wasted unless the fabric detects and corrects these collisions.

02

Adaptive rebalancing

Switches monitor queue depth and reroute flows onto less-loaded equal-cost paths in hardware. The balance is maintained dynamically against the live traffic, not just assumed from the wiring.

03

Topology-aware collectives

The communication library is told the physical layout so it constructs rings and trees that follow real links. Software and fabric agree on the topology instead of the library guessing at it.

04

Deterministic reconvergence

When a link fails, routing recomputes over the surviving fabric and preserves as much bisection bandwidth as the remaining hardware allows. Degradation is proportional, not catastrophic.

Topology is a capital plan the owner controls, not a rented abstraction

For a sovereign buyer the wiring diagram is an owned, auditable asset with a known cost basis.

01

Sized to the mission

The fabric is provisioned for the largest model the owner intends to train in-nation, with defined headroom. Capacity is a deliberate sovereign decision rather than a ceiling imposed by an external provider.

02

Documented as-built

Every cable, port, and rail assignment is captured in an as-built record. The physical fabric is fully documented, which is a precondition for the external audit sovereign infrastructure must withstand.

03

Failure domains by design

Pods and rail planes form explicit fault domains, so a switch or power event affects a bounded slice of the cluster. Resilience is engineered into the layout rather than added afterward.

04

Expansion without lock-in

Because growth follows a standard Clos pattern, the owner can add capacity from multiple hardware sources over time. The topology outlives any single procurement cycle.

Build it sovereign.

Talk to us about non-blocking topology in a sovereign deployment.