Debug live distributed systems with eBPF
We show exactly which kernel function, I/O path and node is the bottleneck, without redeploying, recompiling or instrumenting your application. Then we build eBPF tools so your team can find the next bottleneck.
Talk to an eBPF engineerWhat we offer
Live production debugging
When a system is slow or failing and the cause isn't obvious, we trace block I/O, VFS and kernel function latency on the running system and show where the time goes.
Custom eBPF tools
Tools designed for your system, using bpftrace, kprobes, uprobes and tracepoints, so your team can keep finding new insights.
Fleet-wide observability
Custom tracing across all your nodes and clusters, with dashboards and alerts. It keeps working even when the network drops.
Built for your systems
eBPF tooling built around your fleet
We've already built fleet-scale eBPF tracing ourselves. Here's what we can build for yours.
The question
Why are reads slow on some storage nodes?
What we trace
vfs_read latency as a histogram, across the storage nodes, on the running system.
What you get
The kernel path and the node that are losing the time.
Coordinated across the fleet
Agents on every node, coordinated across clusters and scheduled from one place.
Keeps working through network loss
Each node buffers data locally, so nothing is lost when the network drops.
Measures what you need
Block I/O latency, VFS behaviour and kernel function latency, plus any metrics that matter for your system.
Dashboards and alerting
Scheduled traces, history, per-node and per-cluster views and alerts.
Other solutions
Lustre storage
Design, deploy, automate, tune and benchmark Lustre, then keep it running with ongoing support.
Storage for AI
Keep GPUs busy with KV-cache offload, GPUDirect Storage, and data and checkpoint I/O tuned for your cluster.
Storage slower than your compute?
Send us your topology, workload and a benchmark run. We will get back to you.