To use an analogy from the automotive sector: Road cruisers, which consume a lot of fuel due to their dimensions and weight, are often the first choice for a comfortable journey. But what if you are in the city centre with narrow streets?
Then a small car with an economical engine will do. The same rule applies to the allocation of storage. Near-line SAS drives that offer a thousand times faster speed than Fast Flash are not the first choice for every application due to their high price. Sometimes it must be decided on whether the service is really necessary for the respective case.**
Currently, data is distributed to the respective data carrier systems according to already defined policies. This process is already established at the beginning and will be adapted if necessary. What this approach cannot offer is a timely, up-to-date considerations and the resulting conclusions at any time.
Conventional methods reach their limits
Given the complexity of today’s database architectures, it is no longer possible to assign which data is stored on which medium manually. This is where artificial intelligence (AI) comes into its own. It collects the number of active users at a given time, how many resources they are using, and takes this information to forecast the performance and storage required.
Since most solutions are scalable, resource overheads should be avoided, i.e. the use of unneeded power or storage space. The user should only pay for the resources he actually uses. Artificial intelligence also saves the budget by relieving database administrators of routine tasks and allowing them to devote themselves to other, strategically more important and lucrative tasks.
Artificial intelligence decides
For example, these systems use the neural cache. This machine learning algorithm finds hidden connections in data access patterns and allocates resources as needed. Here, data can be stored with frequent access to faster media, which can be swapped out to cheaper - but also slower - storage media. It is differentiated whether data is stored on near-line SAS drives, fast flash or conventional disk drives.
In general, the more data or "experience" it has, the higher the performance of artificial intelligence. In this way, data that is made available to users automatically and anonymously helps to identify trends and obtain information on how, for example, storage space can be improved. Valuable information about user behavior or similar configurations can also be collected using artificial intelligence.
The findings not only help to use resources more efficiently. They also improve customer support. The more information available to support, the better and more targeted are the steps to be taken in a support case. Under this aspect, too, there is no way around artificial intelligence in the storage area.
Written by Eran Brown, CTO, INFINIDAT.