The term “data fabric” is used in the technology industry, but its definition and application may change. I saw this among the sellers: in the autumn last year, British Telecom (CT) talked about data fabrics at an analyst event; Meanwhile, in the warehouse, Netapp directed its brands to smart infrastructure, but previously used the term. The application platform seller Appian has a data fabric product and the database provider Mongodb talks about data fabrics and similar ideas.
In essence, the data fabric is a combined architecture that abstracts and integrates different data resources to create an uninterrupted data layer. The principle is to create a combined, synchronized layer between your applications, workloads, workloads, and increasingly AI algorithms or learning engines that need to be accessed with different data sources.
There are many reasons to ask for such a boarding. The data fabric functions as a generalized layer of integration, attached to different data sources or adds advanced features such as providing access to these resources to facilitate access to applications, workloads and models.
So far, so good. However, the difficulty is that we have a gap between a data fabric principle and its real application. People use the term to represent different things. To return to our four examples:
- CT defines data fabric as a net -level coating designed to optimize data transmission over long distances.
- Netapp’s interpretation (even with the term intelligent data infrastructure) emphasizes storage efficiency and central management.
- Appian positions the data fabric product as a tool to combine data in the application layer, which allows the user to develop and customize more quickly.
- Consider Data Fabric Principles in the context of data management infrastructure.
How do we cut all this? One answer is to admit that we can approach it in more than one angle. You can talk about data fabric conceptually – determining the need to combine data resources – but without excessive access. You certainly don’t need a universal “Uber-Fabric çocuklar that covers everything. Instead, focus on certain data you need to manage.
If we embrace for several decades, we can see similarities with service -oriented architectural principles that try to distinguish service provision from database systems. At that time, we discussed the difference between services, processes and data. The same applies to now: you may request a service or request data by focusing what is necessary for your workload. Continue to be the simplest of data services, read, update and delete!
I also remind you of the origins of network acceleration that will use cache to accelerate data transfers by keeping data transfers locally instead of accessing the source over and over again. Akamai has established his job on how to transfer unconfigured content such as music and films at efficient and long distances.
This does not argue that data fabrics re -invented the wheel. We are in a different (cloud -based) world technologically; In addition, at least they bring new directions around Meta Data Management, Underpoiring, Compatibility and Safety features. These are particularly critical for AI workloads where data management, quality and provence directly affect model performance and reliability.
If you intend to distribute a data fabric, the best starting point is to think about what data you want. This will not only help you direct you to what kind of data fabric can be the most appropriate, but also this approach also helps to avoid the trap of trying to manage all the data in the world. Instead, you can prioritize the most valuable data subset and think about which data fabric works best for your needs:
- Network level: Integrate data between multiple clouds, in -house and edge environments.
- Infrastructure level: If your data is centralized with a storage seller, focus on the storage layer to provide consistent data pools.
- Application level: To combine different data clusters for certain applications or platforms.
For example, in the case of CT, they found internal value in use to obtain data fabrics from more than one source. This reduces replication and helps to facilitate operations and makes data management more efficient. It is a clearly useful tool to reinforce silos and improve application rationalization.
In the end, the data fabric is not a monolithic, one -body solution. A strategic conceptual layer supported by products and features is that you can apply it in places where it is most logical to add flexibility and improve data presentation. The distribution fabric is not a “set and forget” exercise: not only the software itself, but also the configuration and integration of data resources.
Although a data texture is conceptually present in more than one place, it is important not to unnecessarily reproduce the delivery efforts. Therefore, be withdrawing data along the network, in the infrastructure or at the application level, the principles remain the same: use it in the most suitable places for your needs and ensure that it develops with the data it offers.
