Application Real-Time Monitoring Service:Benefits

Out-of-the-box monitoring

Set up monitoring in minutes instead of days:

• Instant Kubernetes monitoring: Create a Prometheus instance for any Kubernetes cluster or cloud resource directly from the Application Real-Time Monitoring Service (ARMS) console. A container monitoring setup that typically takes 3 days to build from scratch takes 10 minutes with Managed Service for Prometheus.

• Built-in alerting and dashboards: Pre-configured Grafana dashboards and alerting rules cover common application components and cloud services. No manual dashboard creation or Alertmanager installation required.

• Managed operations: Health inspection, automatic agent upgrades, and visualized configuration are built in, reducing the operational burden of self-hosted Prometheus.

Cost-effective

• Free Kubernetes metrics: Monitor Kubernetes components at no cost with a set of included metrics.

• No infrastructure to manage: The service is fully managed. You do not need to purchase or maintain additional servers, and nearly all O&M overhead is eliminated.

• Fast time to value: Integrates directly with Container Service for Kubernetes (ACK). Start monitoring ACK clusters without building a separate monitoring stack.

Compatible with open-source Prometheus

If you already run Prometheus, your existing configuration works here:

• Native collection rules: Supports Prometheus.yaml scrape configs, ServiceMonitor custom resources, and annotation-based auto-discovery. ServiceMonitor is especially useful for monitoring custom Kubernetes clusters.

• Standard query and data model: Supports PromQL queries, custom multi-dimensional data models, and the HTTP API -- the same interfaces you already use with open-source Prometheus.

• Flexible service discovery: Static file configurations and dynamic discovery mechanisms make it straightforward to migrate existing monitoring setups, integrate new data sources, and discover monitoring targets.

Scalable cloud storage

• No storage ceiling: Cloud-based distributed storage scales with your data. There is no upper limit on data volume, and distributed replication safeguards data reliability.

• Cross-cluster queries: Use Global DataSource and Global View to query and visualize metrics across multiple Kubernetes clusters in a single aggregate view.

High performance

Managed Service for Prometheus outperforms self-hosted Prometheus in both collection throughput and query speed:

• 6x single-replica throughput: On 2-core CPU / 4 GB memory hardware, each replica collects 6 million time points per scrape, compared to 1 million for open-source Prometheus. The collection component is optimized to improve single-replica throughput while reducing resource consumption.

• 20x collection efficiency: A lightweight agent architecture requires only a single process per Kubernetes cluster, improving collection performance by 20 times over open-source Prometheus. The Prometheus agent is deployed on the user side, leveraging native collection capabilities to minimize resource usage.

• 18-22x faster queries: Querying 0.6 billion time points takes 8 to 10 seconds, compared to 180 seconds with open-source Prometheus.

• Separation of collection and storage: Decoupling these layers lets each scale independently, removing the single-process bottleneck of open-source Prometheus.

• Auto-scaling agents: The number of agent replicas increases or decreases automatically based on workload, distributing collection tasks across replicas.

High availability

• Dual-replica architecture: Data collection, processing, and storage components each run with multiple replicas to maintain high availability across core data paths.

• Horizontal scaling: Elastic scale-out adjusts capacity based on cluster size. No manual intervention required.

• Automatic data retransmission: A built-in retry mechanism preserves data integrity and accuracy, even during transient failures.

Managed Service for Prometheus vs. open-source Prometheus