Slingshot

The Interconnect for Exascale Computing

Contact Us

New Capabilities for New Possibilities

Scientific and technological questions are changing. So how we answer them has to change, too.
For HPC, that change means the ability to run diverse simulation, modeling, AI, and analytics workloads. On one system. At exascale speeds. Simultaneously. Bringing these capabilities to life requires a flexible, heterogenous supercomputing architecture. And a high-speed, purpose-built supercomputing interconnect to support it.

Start with the right partner.

We've been helping visionaries like you solve the most complex computing problems for over four decades.

Not a First, But a Next

If supercomputing is our thing, then so are networks. We started developing scalable, high-performance networks over 25 years ago. Slingshot is our eighth major generation of networking technology.

So yes. We know HPC networks. But it doesn’t mean designing them is just another day at the office for us. Each one represents a hard-won networking milestone.

The genesis of Slingshot dates back to 1992 when we introduced the Cray T3D, our first massively parallel processing system. Our first-ever implementation of adaptive routing followed in the Cray T3E system. Then came high-radix switches, low-diameter networks, and most recently Aries, the first network to implement the Dragonfly topology.


SCALABLE, HIGH-SPEED,
PURPOSE-BUILT INTERCONNECT

New Features Raise the Performance Bar

Adaptive Routing

Each switch can make fast, well-informed decisions about optimal paths for each data packet and adjust automatically.

Congestion Control

The innovative congestion control mechanism requires no tuning and stays stable across a range of HPC workloads

Ethernet Compatibility

Slingshot switches can connect directly to third-party Ethernet-based storage devices and networks


Learn more about what makes Slingshot an ideal network for exascale computing.


Download the White Paper

By using this site, you agree to Cray's terms of service and privacy policy.