Time series data is all around us. We use this data every day in our smart homes, in our connected cars and on our mobile devices. And thanks to the steady growth of sensors and connected devices, companies have more time series data at their fingertips than ever before. From stock market analysis and economic forecasting to earthquake predictions and healthcare, the uses of time series data apply to every industry amid the growing need to determine trends over time. If time series data is the future, then the future is now.
Behind The Explosion Of Time Series Data
Two significant factors have contributed to this surge in time series data. The first factor is the rise of connected devices and the Internet of Things. As new devices come online, they generate an ever-increasing amount of time series data. Imagine a common metric from a connected car, like speed in miles per hour. If we change the time interval for collecting this info to every minute or second, the resulting dataset is orders of magnitude larger. Apply this principle across an entire industry of millions of connected devices and sensors, and the exponential growth of this data becomes obvious.
The second factor in the growth of time series data is the way in which companies use data itself. As companies move large volumes of data to the cloud, these systems, processes and containers generate time series data. Once in the cloud, companies use that data horizontally, creating larger workloads and more data. Netflix, for example, started in one geographic area and now spans hundreds of locations across six continents. We see similar trends from other services and industries where persistent data gets reused across expanding networks.
It’s worth pointing out that there are different types of time series data: metrics and events. Metrics are measurements taken at a regular interval; events are measurements collected at irregular intervals. In order to use time series data effectively, you need to be able to handle both types. Collecting only metrics means you could miss out on critical events when they occur, and collecting only events means you could misinterpret anomalies as major events. To avoid these pitfalls, use a solution that can handle both data types.
Who Benefits From Time Series Data
Whether companies design their applications to include time series data from the outset or retrofit time series analysis into established, preexisting applications, the industries and use cases for time series data vary widely:
• Financial services companies can use time series data to monitor for transaction anomalies.
• Industrial businesses can use systems on factory floors to produce a variety of time series data used for real-time alerting, process optimization and forecasting.
• Streaming services can use time series data to identify and prevent issues before they impact end users.
• Telecommunications carriers often rely on time series data for anomaly detection, network telemetry and capacity planning.
As the sources of time series data increase, so do our needs to process, analyze and act on this data. Each of the examples above generates data in different formats. In fact, the same application may pull in data from 10 different sources, each of which is in a different format. That’s also why it’s vital to select a time series solution that can ingest data from a wide variety of sources. The more flexible a solution is when it comes to data ingestion, the more you future-proof your application, saving time and money in the long run.
Download Handouts Click Here
What To Look For In A Time Series Data Platform
Every use case is different, so it’s critical to choose a time series platform that meets the ingest needs of your application. Attributes to look for in a platform include:
1. The scale and velocity of data. This is a key indicator—as time series data grows exponentially, organizations looking to harness it need a platform that can handle it all now and for years to come.
2. The shape of time series data. This can change much faster than other data types, so having a solution that can adapt to adding new data fields without requiring additional development work will save you a ton of time in the long run. In my experience, legacy technologies, such as relational or document databases, are insufficient for managing this type of data.
3. The modernization of legacy devices. The devices producing time-stamped data include consumer devices, such as phones, cars and appliances, as well as industrial IoT devices and processes, like those related to manufacturing or healthcare. But the way companies modernize and upgrade legacy devices also needs to be factored in. Connecting and integrating these devices further adds to the increasing amount of time series data.
