Grid for Apps is our event-driven framework for data processing, which enables customers to run applications directly on the object store. This is a radical approach compared to other compute models available on the market, but it makes a lot of sense for those tasks and automations that involve data manipulation, metadata enrichment, and other jobs that can be easily managed when triggered by an event.
A different compute model
This is not a fancy way to run your SAP environment or your Oracle DB, but it can be considered complementary to them. At the same time, it can easily improve data management and become a powerful framework to build next-generation applications.
Use cases include:
- Metadata enrichment (create additional information or modify existing metadata during ingestion),
- Data transformation (data set filtering and normalization, real-time analysis, transcoding),
- Monitoring (alarming, real-time monitoring, data quality control),
- Compliance and security (data scrambling, pattern recognition, real-time data masking)
And this is only the beginning
We are working on some amazing use cases, such as real-time video transcoding, which helps to save huge amounts of storage capacity. Or image/data recognition through machine learning techniques, to get descriptions of images and add them directly to the metadata, so it is easier to search and analyze content. The same technique can be used also to change content, such as to blur faces or license plates to enhance privacy.
Just data and applications
This is just data and applications, next to each other with a serverless approach, simplifying the entire infrastructure and offloading several tasks to the storage system while increasing its overall efficiency. Look at this diagram, and tell me you don't like it. This is the dream of all developers: removing most of the layers that usually require a lot of management.
Grid for Apps is possible because of the lightweight design of SDS core (which can run with a minimal configuration of 512MB RAM and 1 ARM core). Few resources are used for storage functions, and the rest is available for other tasks, making the system particularly efficient.