Compromise is tough. I'd like a new car that gets over 50 miles per gallon, with 500 horsepower, comfortably seats 8, fits in a small parking spot, looks great, and costs less than $25,000. Not happening anytime soon.
Big data architects want it all. Efficient, fast, scalable, convenient, innovative, and inexpensive. They may have better options with Hadoop than I do with car buying. While HDFS and Hbase have proven themselves as sound choices for file system and NoSQL database approaches, Kudu is stepping up as a relational database that fits squarely between them.
Kudu is a simplified approach to offering updateable analytics databases (alongside Impala) or real-time applications (paired with Spark). Introduced publicly more than a year ago, but quietly developed for much longer, Kudu is great for common use cases like time-series, machine data, or an operational data warehouse for reporting.
Of course, most open source offerings these days have sponsors that help them along, and Cloudera's investment in Kudu looks to be shaping up to benefit both the company's depth of portfolio and their customers, too. I look forward to seeing what advancements the next year will bring....