The in-memory batch-processing framework sheds more JVM performance bottlenecks as a major Hadoop vendor eyes Spark as a full-blown replacement for the aging MapReduce Apache Spark, the in-memory data ...