While we are quite a far way away from hearing “Houston, tranquility base here… the eagle has landed”, the HP moonshot is definitely pushing us all toward a new class of infrastructure to run more efficient workloads, like Apache Hadoop. Hortonworks applauds the development of flexible Big Data appliances like Moonshot. We are excited about this development as it signals alignment across development, operations and infrastructure within organizations. For quite some time, our team has been accustomed to a natural balance required across these three constituents and now the server the market is joining in on the game.
We agree with our friend, Jeff Kelly at Wikibon in that “Big Data as one example of a workload that requires a lot of low level optimization. One of the main reasons is that Hadoop clusters are scaled over time in response to increased usage, and factors like power efficiency and the physical footprint of servers become major considerations as the environment grows in size.”
Did I just hear Moonshot can enable up to 1800 nodes in a single chassis. Wow! Sounds like physical resource optimization to go along with optimizations provided for compute and storage in Apache Hadoop. To quote one of our Hortonworkers, “this is awesomeness on a stick”. Moonshot seems to be forward looking as well. It will eventually lead to further price/power/utilization optimizations as the price of SSD drops and I/O becomes more widely deployed against flash. The HP Moonshot approach is interesting with sled for servers and sleds for disk enables completely new server and rack configurations to be optimized for Hadoop. We are looking forward to getting our hands on it.
Ultimately, Hadoop workloads are somewhat unique and we are intrigued to say the least at where the future can go with the HP Moonshot approach.