Everything You Need to Know About Data Reduction

Posted by VIOLIN SYSTEMS on Jun 3, 2020 1:05:09 PM

It seems like our entire lives -work and personal- are composed solely of data. There’s just so much of it. With the use of big data analytics on the rise, there’s more data than ever before. In fact, each day, users generate 2.5 quintillion bytes of data. When we talk about big data analytics, this is a massive volume of structured and unstructured data– sometimes petabytes- or exabytes-worth of information to be processed.

All of this data has to go someplace. But too much data can leave systems weighted down, with lots of lag. These lag times are killer for the overall user experience and can lead to frustrating delays unless you have the right systems to handle all of the data coming in.

To manage all of this data, it’s necessary to employ data reduction techniques. What does this mean? What should you know about data reduction, and how it applies to all-flash solutions?

Here’s the low-down on data reduction:

What is Data Reduction?

Before we talk about what data reduction can do and why it’s a crucial part of all-flash solutions, it’s important to know what data reduction means. Data reduction is a process designed to reduce the total capacity needed to store data. With data reduction, you maintain all the information you need, usually with the same level of quality, but you can increase overall efficiency and reduce costs because your data doesn’t demand so much of your system.

You may have heard data referred to in terms of raw capacity and effective capacity. What does that mean? Raw capacity refers to data before it has been reduced. Effective capacity is the term for data after it has been reduced.

How Does Data Reduction Affect Flash Systems?

Because data is accumulating at such a fast rate, flash-based systems have to keep up, which is why data reduction is so necessary. Transaction processing and big data analytics in particular can experience delays or perform poorly if there’s just too much data for the system to handle.

What Does Data Reduction Look Like?

There are different methods of data reduction, all of which function a little differently. Depending on how you plan to use this data, different methods may suit your needs better than others. It’s all about finding what works best for you.

Data Cube Aggregation

Data aggregation is a way to summarize information so that it’s a simpler form. An example of this would be taking data from previous years. Maybe this raw data covers revenue per quarter. If you need the data for annual sales, this data can be aggregated to summarize the total sales for each year to optimize effective capacity. You’re pulling less data, but still receiving all the information you need.

Dimension Reduction

Dimension reduction reduces the data size by eliminating outdated or redundant elements. There are three kinds of dimension reduction techniques: Stepwise forward selection, stepwise backward selection, and a combination of forwarding and backward selection.

  • Stepwise forward selection begins with an empty set of attributes and keeps only the best data.
  • Stepwise backward selection starts with a complete set of attributes and removes the worst remaining data.
  • Combination forward and backward stepwise selection allows users to remove the best and worst attributes and examine all the data that exists in the middle ground.

Data Compression

Data compression reduces the raw capacity size of the files by using encoding mechanisms. With data compression, you can either reduce the size of the data with algorithms that allow you to take effective data and restore the information to its original state (lossless compression), or compress data and lose some of the original data’s state, but keep enough of it intact to be able to retrieve necessary information (lossy compression).

Numerosity Reduction

This data reduction technique replaces data with mathematical models and/or a smaller representation of the full data set. This may mean that only data within certain parameters is stored, or reductions may include clustering, histogram, or sampling of data. Data is still accessible, but because your team will never need all of the data, you keep only what you need.


With data discretization, the continuous nature of data is divided up based on different attributes. All of these constant attributes are replaced with labels, so data is accessed in a concise and digestible fashion.

Concept Hierarchy Reduction

Concept hierarchy reduction takes all of the raw data and replaces low-level concepts with high-level concepts, oftentimes by sorting data into effective, defined categories or “bins.” This might look like taking a low-level piece of data, like someone’s age, and putting them in a larger concept, like a 76-year-old person categorized as a senior citizen.

Why Does Data Reduction Matter?

Data reduction can have all kinds of benefits for your flash system. When you amass large amounts of raw data as a result of big data analytics and value creation activities for employees and customers, you gain all kinds of useful information, and you can even increase user satisfaction. Without data reduction, this can come at a cost. You need maximum efficiency, and data reduction can provide that.

Flash service fees can increase when you take on all kinds of data. It has to exist somewhere. Reducing the total load of all of this data also reduces how much data exists within your flash system.

What do you gain when you implement data reduction techniques?

  • A lower service utilization cost due to improved and reduced data usage
  • Less latency, resulting in increased customer trust in your enterprise system
  • Secure data sharing within your organization
  • Preserved privacy of customer data

Who Benefits from Data Reduction?

If you are looking to increase the efficiency of your flash system or maximize what you can do with your overall workload, data reduction may be for you, especially if you are starting to use big data analytics or considering adding big data to your arsenal.

Another reason to consider data reduction? If you want to reduce the costs of your flash-storage system, data reduction is a way to decrease your total load and create a more lightweight system without losing any of your functionalities.

Access information quickly and enjoy fast turnaround times with big data analytics with a flash system, and keep things moving economically with data reduction. As we say at Violin Systems, flash servers and data reduction work in perfect harmony.

Want to learn more about what data reduction can do for you, or which kind of data reduction is right for your business needs? Let’s talk. Contact us today for more information.

Topics: data reduction