Home / Products / ThinkRMS
Unified Monitoring & Alerting Platform

ThinkRMSUnified monitoring for domestic GPUs

See compute, storage, network and accelerator resources in one platform. Unified collection, visual dashboards, multi-dimensional alerting and AI optimization — natively supporting NVIDIA and multiple domestic GPUs, delivered offline for government and enterprise intranets.

rms.thinkalike.com.cn / overview
1,286Online nodes
38Active alerts
99.98%Availability
CPU/MemoryStorageNetwork/SNMPGPU

Four Resource Domains, One Platform

Panoramic compute-resource monitoring and fault prediction for AI data centers, government, enterprise and research institutes

Compute Monitoring

Per-core CPU, memory, load and process-level metrics with built-in host dashboards and alert templates — out of the box.

Storage & Disk Arrays

Dynamic monitoring of partition capacity, IO and usage trends, with unified management of hardware RAID arrays and disk SMART health.

Network & Switches

NIC traffic, error packets and concurrent-connection tracking; SNMP collection of multi-vendor switch CPU and port metrics out of the box.

GPU Accelerator Monitoring

Utilization, memory, temperature, power and ECC/XID deep metrics; a unified view for NVIDIA plus domestic cards such as Ascend, Hygon and Cambricon.

Fault Prediction & Self-Healing

Multi-dimensional tiered alerting, suppression/silencing and subscription rules across 20+ notification channels, with configurable auto-run scripts for self-healing.

AI Optimization Advice

LLM-driven root-cause analysis plus rule assistance; connects to domestic LLMs on-premises so data never leaves your domain while producing optimization advice.

Domestic GPUs monitored out of the box — one view of all compute

Accelerator cards from different vendors use different metric names. ThinkRMS decouples native collection from normalization in two layers, converging NVIDIA, Ascend, Hygon, Cambricon, Moore Threads, Iluvatar and MetaX into unified metrics for cross-vendor comparison — no per-card dashboards needed.

  • 7 major vendors in a unified metric view; when real cards arrive, just change the scrape target.
  • Dashboards & alerts unchanged: swapping card types only tunes normalization rules, zero business impact.
  • Xinchuang needs met: satisfies the dual demand of NVIDIA stock plus domestic incremental for government and enterprise.
Unified accelerator view · accelerator_*
NVIDIA82%
Ascend NPU67%
Hygon DCU54%
Cambricon41%

Command-center dashboard + editable boards

A full set of dashboards and low-code visual big screens with native drag-and-drop design, server-side persistence across multiple screens, and presets wired to real data. Paired with an AI copilot and inspection reports, it turns massive metrics into actionable optimization advice.

  • Low-code big screen: drag-and-drop building, business-group linkage, full-screen command mode.
  • Inspection reports: utilization scoring, capacity forecasting and report generation.
  • Multi-tenancy: business-group permission isolation for tiered operations.
Visual big screen

Three tiers, on-premises by demand

From single-node lightweight to multi-datacenter group scale, one unified offline installer, one-click intranet deployment

Lite

≤ 50 nodes

Single-node all-in-one, out of the box

  • Host / network / storage base monitoring
  • Built-in dashboards & alert templates
  • Offline installation package
Standard · Recommended

50 – 500 nodes

Domestic GPUs + inspection reports

  • External time-series store, high-performance storage
  • Normalized unified view for domestic GPUs
  • AI optimization + inspection reports
Group

Multi-datacenter

Edge engines + custom big screen

  • Center + edge-down multi-instance clusters
  • Traffic analysis & command-center big screen
  • License authorization & multi-tenancy
4 domains
Compute/Storage/Network/GPU
7 vendors
GPU normalization
20+
Alert notification channels
1 click
Offline intranet deployment

Make every unit of compute visible

Experience ThinkRMS's full-stack resource monitoring, unified domestic-GPU view and AI optimization advice.

Start Free Trial

Frequently Asked Questions about ThinkRMS

What is ThinkRMS?

ThinkRMS is a unified resource monitoring and alerting platform by ThinkAlike, covering compute, storage, network and GPU accelerator cards with unified collection, visual dashboards, multi-dimensional alerting and AI optimization.

Can ThinkRMS monitor domestic GPUs?

Yes. It natively supports NVIDIA plus Ascend, Hygon, Cambricon, Moore Threads, Iluvatar and MetaX, converging them into a unified metric view through normalization rules.

Do I have to rebuild dashboards when switching GPU vendors?

No. Dashboards and alerts stay the same — you only adjust normalization rules, with zero impact on the business.

Does ThinkRMS support offline intranet deployment?

Yes. A unified offline installation package enables one-click intranet deployment; data never leaves your domain, and it can connect to domestic LLMs for optimization advice.