 |
In a recent post I discussed the topmost mobile layer of an analytic stack and how it can relate (securely) to the bottom-most layer of an IT infrastructure. The value proposition of Pivotal is its ability to reduce the software development time of big data analytic applications. In order to do this the developer must be abstracted away from the underlying IT infrastructure. In this post I'd like to spend a bit more time diving into the interplay between the Cloud M&O layer and the PaaS layer, especially as it relates to the deployment of Big Data apps. In particular, there are two items worth mentioning:
The diagram below highlights how the second layer (in this case VMware) accomplishes item #1: VMware's Big Data Extensions are a great way for a cloud orchestration platform to surface data analytic entry points such as HDFS. In fact, EMC's OIL team has leveraged the Big Data Extensions as part of their Hadoop Starter Kit (which leverages Isilon). This approach by VMware presents HDFS to an application developer. But how can an application developer leverage this access point without tying themselves to the cloud that is producing the entry point? The graphic below highlights the Cloud Foundry approach, which facilitates the export of HDFS to the Data Fabric in a cloud neutral manner: With a few simple commands an application can be deployed in a cloud-neutral way to consume an exported service (e.g. HDFS) produced by a cloud orchestration layer such as VMware. I was at EMC's 2-day leadership summit this week and Cloud Foundry got a lot of air time. In a future post I'll compare what's happening with Cloud Foundry in the industry in the context of historical application and storage interaction. Steve EMC Fellow |
