Devs and Data, Part 3: Managing a Large Volume of Data

This article is part of the Key Research Findings from the new DZone Guide to Big Data: Volume, Variety, and Velocity. 

Introduction

Welcome back! In case you missed them, here are some links to Part 1 and Part 2. In today's installment, we check out what our respondents had to say about managing large volumes of data. 

Just as a reminder of our methodologies, for this year’s big data survey, we received 459 responses with a 78% completion rating. Based on this response rate, we have calculated the margin of error for this survey to be 5%.

Data Management

The basis of any data management plan is data storage. According to our respondents, there is a shift going on from cloud-based solutions to on-premise and hybrid solutions. 29% of respondents reported that their data typically resided in the cloud (down 10% from 2018), 31% told us they use a hybrid solution (up 7% over 2018’s report), and 40% use on-premise data storage (another 7% year-over-year increase). In terms of the actual database used to house this data, MySQL proved the most popular in both production (51%) and non-production (61%) environments, though its year-over-year adoption rate stayed rather static. PostgreSQL could be an interesting database to keep an eye on in the coming year, as its adoption rose in both production (42% in 2018 to 47% in 2019) and non-production (40% in 2018 to 48% in 2019) environments.

For filing big datasets, a vast majority of respondents told us they prefer the Hadoop Distributed Files System (HDFS). In fact, 80% of survey-takers reported using HDFS as their big data file system. While this large of a majority among respondents is impressive in its own right, HDFS also saw a 16% increase in adoption over our 2018 Big Data survey. The second most popular response to this question, Parquet, had a 36% adoption rate in our 2019 survey, up from 17% last year. Interestingly, even the least popular of the file systems reported, (O)RC File, saw an 11% year-over-year increase, rising to a 17% adoption rate.

Data Volume and Issues With Big Datasets 

We also asked respondents about the issues they encounter when dealing with such large volumes of data. It turns out that normal files (such as documents, media files, etc.) cause the most headaches, with 49% of respondents selecting this option. Server logs also proved a popular answer, garnering 42% of responses. Data collected from IoT devices, however, saw the largest increase in developer frustrations. In 2018, 20% of respondents reported data from sensors or remote hardware as an issue; this year, 32% of survey-takers reported this type of data as a pain point. Surprisingly, despite user-generated data (i.e. social media, games, blogs, etc.) being one of the largest means of creating and ingesting new data, the difficulty this type of data gives to developers and data scientist seems to be decreasing. In 2018, 33% of respondents said user-generated data was a pain point in their big data operations; in 2019, this fell to 20%.

The types of data that gives developers issue when it comes to large volumes of data also witnessed a good deal of variability over last year. The data type that, according to respondents, causes that most issues —relational data — fell by 8%. Despite this decrease, it still registered 44% of respondents’ votes. Event data also underwent a large swing, only in the opposite direction. In our 2018 survey, 25% of respondents said they had issues with event data; in 2019, this number rose to 36%. This increase in the number of respondents having trouble with event data is intriguing, given that user-generated data was reported as less of an issue than last year, yet much of the event data there is to be collected can be categorized as user-generated.

That's all for our look into data management and data volume. Come back tomorrow for the final part of this four-part series, in which we investigate the last of the Three Vs, variety. 

This article is part of the Key Research Findings from the new DZone Guide to Big Data: Volume, Variety, and Velocity.