Storage telemetry,understood.
HSP turns performance counters, internal FTDS statistics, logs and alarms from OceanStor Dorado arrays into explained anomalies, evidence-backed diagnoses and ready-to-send reports — on premises, fully offline.
One pane for everything your arrays emit.
Performance dumps, REST inventory, SSH diagnostics, SmartKit archives and live alarms flow into a single time-series and analytics stack — with every heavy task queued, retried and audited.
- SourcesREST API · SSH / SFTP · performance dumps · SmartKit archives · alarms
- CollectorsInventory sync every 5 min · alarm poller every 60 s · perf watcher · log collector
- StorageVictoriaMetrics time series · PostgreSQL job queue & history · object storage for artifacts
- IntelligencePattern detectors · ML root cause · performance envelope · health & fault engines
- OutputsGrafana · web console · Excel / Word reports · AI assistants via MCP
Root cause, with the reasoning shown.
A latency spike is a symptom. HSP looks at the peers around it, tests which signals moved first and how far they left their normal range, and tells you which one most likely caused it — with every number on the table.
Pattern detectors
FC port transmission latency, high utilisation, write flow control and baseline “find the difference” — grouped into episodes and incidents, run automatically after every upload.
Two-phase root cause
Peers are selected automatically, then ranked by Granger causality, permutation importance and deviation from baseline. Results in ~30 s across ~2,500 series.
Guard rails, not guesses
Eleven audited post-model filters, an independent “extreme movers” check and lead / lag in seconds keep the verdict honest.
Performance envelope
Quantile models learn each array’s IOPS, bandwidth and latency ceilings, with drift detection when the workload changes shape.
Detect
Envelope and pattern detectors flag a deviation the moment data lands.
Correlate
Findings collapse into episodes and incidents instead of an alarm storm.
Explain
Ranked causes with confidence, lead time and links to the exact charts.
Report
Engineer-grade evidence and management-ready Excel and Word reports.
Evidence first. Before the ticket.
From scheduled health checks to offline diagnosis of a full log archive, every finding points to the line of evidence behind it — so the conversation with support starts at the answer, not at the question.
SMART health score, 0–100
Per-disk scoring across 160+ drive models, fleet reports and a 16-section Word report generated from a single archive.
Every verdict cites its source
Offline diagnosis of a collected log archive. Findings carry typed evidence down to the file and line, so engineers can verify in seconds.
Three speeds of “is it healthy?”
Read-only checks over the database, SSH and time series — with a 0–100 score and JSON / XML export.
See inside the controller
HSP decodes the array’s internal FTDS statistics — hundreds of millions of points per array, published in minutes — and surfaces the noisiest IO-path locations and hottest CPU functions.
Know what changed, and when
Controllers, disks, ports, LUN mappings, SFPs, BBUs and PSUs synced over REST; hourly snapshots deduplicated by hash; alarms polled every 60 s across up to 32 arrays in parallel.
One click from array to report
Three collection strategies over SSH / SFTP pull exactly the archives support needs, then chain straight into health check and diagnosis.
Ask your fleet. In plain words.
A built-in Model Context Protocol server gives any MCP-capable assistant read-only access to metrics, health data and internal statistics — so an engineer can ask a question instead of building a query.
mcp tools
Runs where the internet doesn’t.
Identity & control
- Local and LDAP / Active Directory sign-in
- 3 system roles, 20 granular permissions
- Per-array scope and full access audit
Hardened by default
- Array credentials encrypted at rest
- Strict security headers, container image scanning
- Offline Ed25519 licensing with anti-tamper
Air-gapped by design
- Single Docker Compose bundle, no cloud calls
- Self-checking install, update and two-level rollback
- Twelve-gate bundle verifier before every release
DME manages. HSP investigates.
Huawei DME is the vendor’s management and automation layer. HSP is built to go deeper on a narrower job: explaining performance and health on OceanStor Dorado. Most teams will want both.
Manage & automate
Huawei’s lifecycle platform for storage: provisioning with approval workflows, multi-vendor and SAN fabric management, alarm handling, AI risk prediction and forecasting — plus the DME IQ cloud service that connects arrays to Huawei support.
Explain & prove
Deep, evidence-backed analysis of what arrays are actually doing — internal statistics, logs and performance — with explainable root cause and offline operation.
| Capability | Huawei DME | HSP |
|---|---|---|
| Official vendor support, automatic service requests | yesvia DME IQ cloud service | noindependent product |
| Provisioning & orchestration | yesservice levels, templates, approval workflows | noread-only by design |
| Multi-vendor storage, SAN switches, hosts, vCenter | yes | nofocused on OceanStor Dorado |
| Capacity & performance forecasting | yesrisk prediction, forward-looking forecasts | partialperformance envelope (labs) |
| Anomaly detection | yesdynamic thresholds; log anomalies in recent releases | yespattern detectors + ML (labs) |
| Findings explained: thresholds, lead time, confidence breakdown | not in public docs | yesevery number shown, linked to charts |
| Diagnosis of offline log archives & SmartKit collections | not in public docsSmartKit itself inspects offline | yesevidence down to file and line |
| Analysis of internal FTDS statistics | not in public docs | yesnoisy locations, hot functions |
| Telemetry stays on site | partialDME on premises; DME IQ uploads to Huawei Cloud | yes |
| Fully air-gapped operation | not in public docsDME IQ needs a route to Huawei Cloud | yesoffline licence, no cloud calls |
| Open APIs | yesREST, SNMP, Ansible, Kubernetes CSI | yesREST API, MCP server for AI assistants |
yes partial not in public docs — DME column based on publicly available Huawei documentation; capabilities vary by version and licence. Recent DME releases add log-anomaly detection and root-cause location. Where both tools fit, HSP complements DME rather than replacing it.