sommai - Fotolia
AWS Machine Learning can accelerate model development and use, but developers first must carefully prepare data, test it and refine it to maximize the value of machine learning features.
Once developers create and train a machine learning application, Amazon Web Services (AWS) provides batch and real-time predictions using wizards and APIs. The batch API looks for a large number of predictions, such as customers who might buy a certain product. Batch predictions are handled offline and all of the predictions are returned together.
The real-time API works with on-demand or as-needed predictions, such as the integrity of a transaction, email or a component's performance check. Results are processed and returned immediately. With AWS Machine Learning, real-time models can request up to 200 predictions per second, though actual throughput depends on data size, model complexity and computing demands from other simultaneous tasks. There are no real limits to the total number of batch predictions AWS Machine Learning can produce.
AWS Machine Learning costs $0.42 per hour for compute power (data analysis and model building), as well as $0.10 per 1,000 batch predictions or $0.0001 per real-time prediction. There is an added charge of $0.01 per hour for each 10 MB of memory allocated to the model and separate charges for any data stored in Amazon Simple Storage Service, Relational Database Service or Redshift.
AWS Machine Learning availability and security
AWS claims high-availability with redundant capabilities for model training, evaluation and batch prediction. However, this doesn't address availability for real-time prediction processing. As with all public cloud services, unexpected provider issues, disruptions to connectivity and Internet access can all affect availability.
AWS uses encryption for models and data both in transit and at rest. All requests to the console and APIs are made using SSL connections. Additional AWS Machine Learning features such as identity and access management tools can help to authenticate users and limit access to machine learning resources. Businesses that require additional security may want to take extra steps with data leakage protection, access and control logs and more.
What skills do developers need for machine learning?
What are some real-world uses for machine learning?
Dig Deeper on AWS big data and data analytics
Related Q&A from Stephen J. Bigelow
Access management is critical to securing the cloud. Understand the differences between AWS IAM roles and users to properly restrict access to AWS ... Continue Reading
Containers have rapidly come into focus as a popular option for deploying applications, but they have limitations and are fundamentally different ... Continue Reading
ALM and SDLC both cover much of the same ground, such as development, testing and deployment. Where these lifecycle concepts differ is the scope of ... Continue Reading