Use Cases
SPARK-XC for Enterprise IT
Enterprise IT
Fits how your team already works
Works alongside the GPU management, scheduling, and facility tools you already run. Nothing to rip out. Each recorded power-cap change becomes a Power Event Record your SIEM can ingest as structured JSON.
Records power-limit changes made by other tools, not just its own. Spark-XC isn't in your workloads' data path. If it goes offline, your GPUs and schedulers carry on, and no Power Event Records are written for that window.
- Works alongside your GPU management, scheduling, and facility tools. Nothing to rip out
- Records power-limit changes made by other tools, not just its own
- Each record is structured JSON, signed with Ed25519 and hash-chained
- Prometheus /metrics endpoint, with a sample Grafana dashboard
Integration Points
Changes by other tools
Recorded too
SIEM ingestion
Structured JSON records
Monitoring
Prometheus /metrics endpoint
Audit
Tamper-evident records you hold
Example scenarios
Example scenarios (illustrative)
Example · IT operations
An OS patch causes a driver regression
An OS security patch is pushed to GPU nodes and the GPU driver stops responding for a while. Spark-XC can't read the GPUs during the outage, so it writes no records for that window and flags the gap.
What the records would show: the last limit recorded and its readback, the flagged gap, then what the GPU reported after the driver came back.
Example · Budget review
A team's power settings drift from the plan
Finance asks whether one team's GPUs ran at the power caps that were agreed. The operator pulls that team's Power Event Records for the quarter.
What the records would show: each power-cap change on those GPUs, when it happened, and the readback.
Example · Security review
Someone changes a power limit with another tool
A power limit is changed directly on a node, outside the usual tools. Records power-limit changes made by other tools, not just its own.
What the records would show: the out-of-band change, when it was seen, and the readback.
Example · Incident response
A power-limit change doesn't take
During a busy period, a GPU doesn't apply a requested power limit. The limit it reports back doesn't match the request.
What the records would show: the requested value, the value the GPU reported, and the mismatch.
Why it matters
The cost of no record
Days
Incident investigation without audit trail
Without a record, teams spend days reconstructing what happened from scattered logs.
1
Record per recorded change
Each recorded power-cap change gets a Power Event Record. If a record can't be written, Spark-XC flags it.
Integration & Deployment
Designed to fit your existing stack
Runs on your hardware
Spark-XC runs on your own GPU nodes. Records are anchored to a local write-once directory by default, and anchors can be sent to off-site storage you control.
Deployment
SIEM & log integration
Each Power Event Record is a structured JSON record. Forward the records to your SIEM, log platform, or archive; the hash chain is preserved in the output.
Evidence chain
Metrics
Per-GPU power draw, temperature, applied power limit, control actions, and safety counters are exposed on a Prometheus /metrics endpoint, with a sample Grafana dashboard.
Prometheus
Changes by other tools
Records power-limit changes made by other tools, not just its own.
Out-of-band
Per-GPU records
Each record names one GPU, so you can look at a single device or a whole fleet.
Per device
Audit hand-off
Share tamper-evident Power Event Records for a time range or a set of devices with internal audit or due diligence. Every record is signed with Ed25519. Auditors only need the public key.
Evidence chain
Integration preview
What a Power Event Record looks like in your SIEM
Each recorded change is a structured JSON Power Event Record. Here's the record structure (sample values):
{
"timestamp": "2026-01-15T14:22:01.003Z",
"device": "example-node-12:gpu:7",
"origin": "other_tool",
"action": "POWER_LIMIT_CHANGED",
"limit_before_w": 320,
"readback_w": 220,
"power_draw_w": 214,
"temperature_c": 71,
"seq": 28491,
"signature": "ed25519:5b0e...a9c4",
"prev_hash": "8a3f...c291",
"entry_hash": "d47b...0f18"
}
Structured JSON
Works with any SIEM that ingests JSON
Enterprise
See the record for one of your power events
We're looking for design partners among enterprise IT teams. Send us one power event or curtailment scenario and we'll show you the record Spark-XC would keep for it.