Why Time Series Databases Are Crucial for IoT and Cloud Monitoring

01 Application Scenarios

Time series databases are optimized for time‑ordered data and are used in manufacturing, finance, DevOps, social media, healthcare, smart home, and networking.

Manufacturing: lightweight production‑management cloud platforms collect sensor data to monitor progress and resource utilization in real time.

Finance: trading systems for securities and cryptocurrencies analyze time‑series market data for quantitative trading.

DevOps: infrastructure monitoring gathers metrics from devices and services to assess health continuously.

Social Media: platforms track user interactions and live‑stream quality metrics.

Healthcare: wearables generate health indicators for business‑intelligence analysis.

Smart Home: IoT hubs monitor smart‑device status remotely.

Networking: network monitoring displays latency and bandwidth usage.

02 Requirements of Time‑Series Data

IoT and cloud‑monitoring generate massive data; for example, Huawei Cloud CES monitors over 70 million metrics per region, processing 900 k reports per second, amounting to about 1 PB per year. Traditional relational databases cannot handle such write pressure.

Continuous high‑performance writes: 24/7 ingestion at sub‑second or nanosecond intervals.

High‑performance queries: real‑time analytics such as anomaly detection and predictive maintenance require fast access to large volumes of data.

Low storage cost: data volumes are orders of magnitude larger than typical OLTP workloads.

Support for massive timelines: billions of distinct series must be managed.

Elastic scalability: ability to expand quickly during traffic spikes.

03 Open‑Source Time‑Series Database Capabilities

Popular open‑source solutions include:

OpenTSDB: built on HBase, offers high availability and write performance but has a longer data stack than InfluxDB.

InfluxDB: uses a custom storage engine with inverted indexes for multi‑dimensional queries; its volcano model can limit aggregation performance.

TimescaleDB: an extension of PostgreSQL, supports SQL, joins, and good read/write performance, though it uses fixed schemas and larger storage.

04 GaussDB (For Influx) Architecture

GaussDB (For Influx) addresses high‑performance writes, massive timelines, and compression with a cloud‑native design that separates storage and compute and employs kernel bypass.

Storage‑Compute Separation: leverages distributed storage for reliability and enables stateless compute nodes, reducing scaling time from days to minutes and lowering replication overhead.

Kernel Bypass: custom user‑space storage interface avoids kernel‑user copy overhead, with containerized compute nodes communicating directly over dedicated storage networks.

Performance Optimizations:

LSM‑Tree layout and asynchronous logging improve write throughput by 94%.

Vectorized query engine, ARC block cache, and aggregation result cache boost aggregation queries up to 9×.

Specialized compression algorithm reduces size by 2× versus Gorilla and moves cold data to object storage, cutting storage cost by 60%.

Indexing for massive timelines raises write performance fivefold at tens of millions of series.

GaussDB (For Influx) has been deployed in Huawei Welink and Cloud Eye Service, and future work will focus on efficient analysis of massive data to provide richer insights.