Virtualizing Hadoop in Large-Scale Infrastructures

After eight weeks of fine-tuning the virtual HDaaS infrastructure, Adobe succeeded in running a 65-terabyte Hadoop workload—significantly larger than the largest known virtual Hadoop workloads. In addition, this was the largest workload ever tested by EMC in a virtual Hadoop environment on Isilon.


Fundamentally, these results proved that Isilon as the HDFS layer worked. In fact, the POC refutes claims by some in the industry that suggest shared storage will cause problems with Hadoop. To the contrary, Isilon had no adverse effects and even contributed superior results in a virtualized HDaaS environment compared to traditional Hadoop clusters. These advantages apply to many aspects of Hadoop, including performance, storage efficiency, data protection, and flexibility.