Amazon Timestream pricing

Amazon Timestream is a fast, scalable database service that helps you store and analyze time-series data for applications like Internet of Things (IoT) and real-time monitoring. With Timestream, you can set up, operate, and scale time-series databases in the cloud in only a few steps. Timestream offers a choice of fully managed, time-series database engines for workloads from low-latency queries to large-scale data ingestion. 

Timestream provides both serverless and instance-based configurations, supporting a variety of operational needs. Its pay-for-use and on-demand pricing model provides the maximum flexibility and cost efficiency, allowing you to choose the database engine and configuration that best suits your specific requirements. Whether scaling up for high performance or optimizing for cost reduction, the customizable approach of Timestream allows you to select the best time-series database engine for your workload. You only pay for the resources you use, making it an ideal solution for dynamic and efficient data management.

Amazon Timestream for InfluxDB is a managed, highly available, open source, time-series database that makes it simple to set up, operate, and scale InfluxDB instances in the cloud. On-demand database (DB) instances allow you to pay for compute capacity per hour that your database instance runs with no long-term commitments. This frees you from the costs and complexities of planning, purchasing, and maintaining hardware. Your database instance and storage charges are based on the configurations you choose.

Timestream for InfluxDB supports both Single-AZ and Multi-AZ deployments, optimized instance classes, and storage types that you can choose to match your workload requirements. 

• For both Single-AZ and Multi-AZ deployments, pricing is per DB instance-hour consumed from the time a database instance is launched until it is stopped or deleted.
• For Multi-AZ deployments, instance charges apply to both Timestream for InfluxDB primary instances and secondary replicas.
• For partial DB instance-hours, they are billed in 1-second increments with a 10-minute minimum charge after a billable status change, such as creating, starting, or modifying the DB instance class.

Database storage is billed per gigabyte per month. The amount is determined by the selected storage type and allocated storage volume for the database. Timestream for InfluxDB databases support the following storage types:

Influx I/O-included storage: This storage option is designed to meet the needs of your most I/O-intensive InfluxDB time-series workloads. There are three options to choose from: Influx I/O-included storage with 3K IOPS included (minimum size 20 GiB), Influx I/O-included storage 12K IOPS (minimum size 400 GiB), and Influx I/O-included storage 16K IOPS (minimum size 400 GiB).

• Data transferred between Timestream for InfluxDB and Amazon EC2 instances in the same Availability Zone:

 Free. 

• Data transferred between Availability Zones for replication of Multi-AZ deployments:

Free.

• Timestream for InfluxDB instances outside VPC:

Data transferred between an EC2 instance and Timestream for InfluxDB instance in different Availability Zones of the same AWS Region have no data transfer charge for traffic in or out of the Timestream for InfluxDB instance. You are only charged for the data transfer in or out of the EC2 instance and standard EC2 Regional Data Transfer charges apply ($.01 per GiB in/out).

• Timestream for InfluxDB instances inside VPC:

For data transferred between an EC2 instance and a Timestream for InfluxDB instance in different Availability Zones of the same Region, EC2 Regional Data Transfer charges apply on both sides of transfer.

Amazon Timestream for LiveAnalytics is a serverless time-series database that automatically scales to ingest gigabytes of time-series per minute. With Timestream for LiveAnalytics, you only pay for the resources consumed by the workload. You are charged separately for writes, memory store, magnetic store, and queries.

For writes, memory store, and magnetic store, you are charged based on the amount of data written and stored. For queries, you are charged for the compute capacity used by your applications. The compute capacity is measured in Timestream Compute Units (TCUs). One TCU comprises of 4vCPU and 16 GB of RAM. TCUs are billed on an hourly basis with per-second granularity (with a 30-second minimum). Data stored in magnetic store is billed by gigabyte-months and for a minimum of 100 GB.

Timestream for LiveAnalytics charges based on the following operations: 

• Writes: The amount of data written from your applications (rounded to the nearest KiB) into a table
• Queries: The duration of Timestream compute units (TCUs) used by your application in TCU-hours
• Memory store: The amount of data stored in the memory store of each table billed in GB-hour
• Magnetic store: The amount of data stored in the magnetic store of each table billed in GB-month

Queries are charged based on the duration of Timestream Compute Units (TCUs) used by your application in TCU-hours.

The memory store charges are calculated based on the amount of time-series data stored in the memory store of each Amazon Timestream table.

The magnetic store charges are calculated based on the amount of time-series data stored in the magnetic store of each Amazon Timestream table.

As part of the AWS Free Tier, you can get started with Timestream for LiveAnalytics for free. Upon sign-up, new customers receive 50 GB for data ingestion, 100 GB of Magnetic tier storage, 750 GB-hour of Memory tier storage, and 24 TCU-hours for query usage during a 1-month free trial of Timestream for LiveAnalytics.

Envision a real-time infrastructure monitoring system designed for managing over 5,000 uninterruptible power supply (UPS) devices. This system is geared towards analyzing vital metrics in real time to ensure the reliability and efficiency of these devices.

Each UPS device transmits data every 10 seconds. This data encompasses 5 key fields: input/output voltages, load capacity, temperature, and battery charge level. Alongside these fields, each data packet includes four essential tags—deviceId, deviceModel, location, and manufacturer—adding layers of contextual information to the metrics.

The monitoring process is executed through a sophisticated dashboard system. These dashboards are programmed to refresh every 60 seconds and run a total of 15 queries in each cycle. This setup provides a near real-time view of the status and performance of all connected UPS devices, allowing for timely decisions and actions.

Another important feature of this system is its automated anomaly detection capability. Tasks are set to run every 10 seconds, meticulously scanning the incoming data for any deviations from predefined thresholds. Upon detecting any anomalies, the system promptly triggers alerts, ensuring immediate attention to potential issues.

To support the demands of this extensive data processing and real-time monitoring system, a robust database infrastructure is essential. A db.influx.2xlarge instance equipped with 400 GiB of Influx I/O-included storage and capable of handling 3,000 IOPS would be ideal for this workload.

Single-AZ deployment cost (monthly)

• Compute cost: $697.88
  • Based on a db.influx.2xlarge instance
  • Operating 24 hours for every day in a month
  • Rate: $0.956 per hour
  • Total hours: 730 (1 month)
• Storage cost: $40.00
  • 400 GiB Influx I/O-included storage (3K IOPS)
  • Rate: $0.10 per GB-month
  • Single-AZ configuration
• Total monthly cost: $737.88
   

Multi-AZ deployment cost (monthly)

• Compute cost: $1,396.49
  • Based on Multi-AZ db.influx.2xlarge instance
  • Operating 24 hours for every day in a month
  • Rate: $1.913 per hour
  • Total hours: 730 (1 month)
• Storage cost: $80
  • 400 GiB Influx I/O-included storage (3K IOPS)
  • Rate: $0.20 per GB-month
  • Multi-AZ setup
• Total monthly cost: $1,476.49

Consider a real-time security monitoring service designed for a network of over 80,000 IoT home automation and security customers. Every 60 seconds, the home control central unit sends a comprehensive set of data, comprising 6 tags and 15 fields, to an InfluxDB-based database system. 

The data received from these central units is aggregated and analyzed at a larger scale to generate insightful trend reports and recommendations. These reports and recommendations are then provided to end users, aiding them in optimizing their home security and automation setups.

Besides the routine data, the home devices also generate a distinct category of real-time events, particularly focused on security aspects. These security events are monitored in real time by a specialized control center. The control center is equipped with advanced dashboards that run approximately 50 queries, refreshing every 30 seconds. This setup offers a near-instantaneous overview of the security status across all connected homes.

Additionally, the system is designed to automatically scrutinize certain types of events at 5-second intervals. This feature allows for the generation of immediate automated notifications to external systems when specific conditions are met, significantly enhancing the system's responsiveness in security matters.

To accommodate these continuous data processing requirements, along with real-time security event monitoring, a robust database infrastructure is necessary. A db.influx.8xlarge instance, outfitted with 600 GiB of Influx I/O-included storage configured with 12,000 IOPS, would be well-suited to handle this workload.

Single-AZ deployment cost (monthly)

• Compute cost: $2,792.25
  • Based on db.influx.8xlarge instance
  • Operating 24 hours every day for a month
  • Rate: $3.825 per hour
• Total hours: 730 (1 month)
  
• Storage cost: $210
  • 600 GiB of Influx I/O-included storage (12K IOPS)
  • Rate: $0.35 GB-month
  • Single-AZ configuration
• Total monthly cost: $3,002.25

Multi-AZ deployment cost (monthly)

• Compute cost: $5,584.50
  • Based on Multi-AZ db.influx.8xlarge instance
  • Operating 24 hours every day for a month
  • Rate: $7.65 per hour
  • Total hours: 730 (1 month)
• Storage cost: $420
  • 600 GiB Influx I/O-included storage (12K IOPS)
  • Rate: $0.70 per GB-month Multi-AZ setup
• Total monthly cost: $6,004.50

Consider a moderately sized ecommerce platform with a focus on optimizing conversion rates through user behavior analysis. The website boasts a user base of 4 million registered individuals with approximately 200,000 active daily users. This activity translates into approximately 10 million events each day.

The website's telemetry system is designed to capture real-time events, each generating records that consist of 8 tags and 3 fields. User identification is achieved through a combination of user IDs (which can be either known or anonymous), device type, and UTM tags. Meanwhile, each telemetry event is characterized by three main attributes: the type of event, the time it takes to load, and the source of the event (such as ads, reviews, or searches).

This rich time-series dataset is then aggregated and analyzed to enhance various aspects of the user experience including optimizing product recommendations, rotating ads, and strategically placing banners. The system performs over 100 queries every 5 minutes, leading to diverse types of data aggregations. These aggregations are essential for the continuous updating and optimization of machine learning models.

A db.influx.16xlarge instance configured with 800 GiB of Influx I/O-included storage capable of 16,000 IOPS would be suitable to handle the workload.

Single-AZ deployment cost (monthly)

• Compute cost: $5,584.50
  • Based on db.influx.16xlarge instance
  • Operating 24 hours for every day in a month
  • Rate: $7.65 per hour
  • Total hours: 730 (1 month)
• Storage cost: $400
  • 800 GiB of Influx I/O-included storage (16K IOPS)
  • Rate: $0.50 per GB-month
  • Single-AZ configuration
• Total monthly cost: $5,984.50

Multi-AZ deployment cost (monthly)

• Compute cost: $11,169
  • Based on Multi-AZ db.influx.16xlarge instance
  • Operating 24 hours for every day in a month
  • Rate: $15.30 per hour
  • Total hours: 730 (1 month)
• Storage cost: $800
  • 800 GiB Influx I/O-included storage (16K IOPS)
  • Rate: $1 per GB-month
  • Multi-AZ setup
• Total monthly cost: $11,969

Imagine a video streaming service operating on 2,000 EC2 instances. This service is focused on analyzing streaming metrics in real time to assess application performance, identify anomalies, resolve issues, and uncover optimization opportunities. Every EC2 instance continuously emits four critical metrics: video startup time, rebuffering ratio, video playback failures, and average frame rate. These metrics are captured every second, providing a granular view of the streaming quality.

To detect anomalies, the system is configured with 7 queries that run every second, scrutinizing the previous 10 minutes of data for irregularities. Additionally, a dashboard with eight widgets has been set up to display the last 6 hours of data, offering a comprehensive view for effective monitoring. This dashboard is designed for concurrent access by five users and is refreshed every hour.

The Timestream for LiveAnalytics database plays a central role in this setup. It stores 6 hours of data in the memory store and 6 months' worth in the magnetic store, ensuring both immediate and long-term data availability. Moreover, the choice of multimeasure records for data storage optimizes costs without compromising on efficiency.

Monthly cost estimate

Data ingestion

• Records ingested per second: 2,000
• Write requests rate: 20 per second (2,000 records/100 batch size)

Costs breakdown

• Writes cost: $183.96
  • Using multimeasure records, the size of a typical time series record is 70 bytes (each record will contain 3 dimensions of type varchar (10 bytes) and 4 measures of type double (8 bytes), 8 bytes for timestamp).
  • Number of records ingested per second: 2000 (each EC2 machine emits a records). We recommend to ingest these records in batches of 100 in a write request.
  • Calculation: 20 requests/sec * 7 KiB/request * 3,600 sec/hour * 730 hours/month * $0.50/million writes of 1 KiB
• Memory store cost: $74.47
  • Calculation: 70 bytes/record * 2,000 records/sec * 3,600 sec/hour * 730 hours/month * 6 hours * $0.036/GB-hour
• Magnetic store cost: $66.22
  • Calculation: 70 bytes/record * 2,000 records/sec * 3,600 sec/hour * 730 hours/month * 6 months * $0.03/GB-month
•  Query Cost: $1,588.19
  • Alerting Queries: $1,512.56
    • 7 QPS (~18.4 million queries/month) require 4 TCU.
    • Calculation: 4 TCU * $0.518/ TCU-hour *730 hours/month
  • Dashboard Queries: $75.63
    • 40 concurrent dashboard queries require ~24 TCU.
    •  Calculation: 24 TCU * (30 seconds/3600) hour * $0.518/TCU-hour * 730 hours/month

Total Monthly Cost

• Total: $1,912.84/month

Consider a dynamic payment processing system linked to numerous point-of-sale terminals across major US cities. This system is designed to capture and analyze transaction data in real time, aiming to detect fraudulent activities and provide answers to critical business queries, such as identifying the busiest and least active point-of-sale terminals, determining peak transaction hours in each city, and recognizing the city with the highest transaction rate per hour.

The system handles a substantial volume of data, processing around 5,000 transactions per second. Each transaction, stored in Amazon Timestream for LiveAnalytics, comprises 100 bytes. The analysis is powered by queries that run every second for fraud detection, queries that aggregate, slice and dice the data to draw business insights.

The application runs 14 queries per second to derive the insights that are visualized in a dashboard by multiple users. The analysis is done for the 8 business hours the stores are open. The data retention strategy includes a two-hour period in the memory store and a six-month period in the magnetic store.

This system also utilizes multi-measure records for efficient data storage, and batches the data in groups of 100, with about 35% data repetition in each batch. To enhance query efficiency, only about 2% of the data accumulated over the past hour is scanned.

Monthly cost estimate (with single-measure records) 

Data ingestion

• Records ingested per second: 5,000
• Write requests rate: 50 per second (5,000 records/100 batch size)

Costs Breakdown

• Writes Cost: $216.00
  • Number of records ingested per second: 5,000. We recommend to ingested these records in batches of 100 in a write request.
  • Calculation: 50 requests/sec * 10KiB/request * 3600 sec/hour * 8 hours/day * 30 days/month *$0.50/million writes of 1KiB.
• Memory Store Cost: $34.99
  • Calculation: 100 bytes/record * 5,000 records/sec * 3600 sec/hour * 2 hours * ~9 hours/day (charged only per usage) *30 days/month * $0.036/GB-hour.
• Magnetic Store Cost: $77.76
  • Calculation: 100 bytes/record * 5,000 records/sec * 3600 sec/hour * 8 hours/day * 30 days/month * 6 months * $0.03/GB-month.
• Query Cost: $994.56/month
  • 14 QPS requires ~8 TCU
  • Calculation: 8 TCU * 8 hours per day * $0.518/TCU-hour * 30 days/month
    

Total Monthly Cost

• Total: $1,323.31/month