JMP and Hadoop<meta http-equiv="Content-Type" content="text/html; charset=UTF-8" />I know several have asked about JMP and Hadoop. Here is what Stan and I have come up with so far. Please respond to this thread with any notes from the field or additional comments/questions.
• JMP can access data in Hadoop optimally with the SAS/Access Interface to Hadoop. By using this interface, customers can access and use big data stored in Hadoop through the SAS integration interface in JMP. The usual caveats of data size and performance apply when trying to access these data in JMP.
• We have been marginally successful connecting to Hadoop via. ODBC and by using SAS/Access to Hadoop in JMP internally. The distribution used was Hortonworks with a local test cluster here at SAS. We were able to get as far as connecting to the server and pulling in metadata about the cluster but due to a still-unresolved permissions error we are unable to load data into the cluster or bring data other than the informational tables into JMP (Stan K and Dan V).
• (SAS Company Confidential) Matt Flynn (Aetna) has been working on developing a workflow around JMP and Hadoop. He is using a Hortonworks cluster. He is trying to use SAS/Access to Hadoop (but running into issues around security/permissions). SAS becomes the connection point to Hadoop. JMP -> SAS -> SAS/Access -> Hadoop. SAS interface for random or stratified sampling would be productive. Also considering Netezza as another intermediate step. IT would serve up summarized and cleaned up data to a Netezza appliance and then he would access in JMP via SAS/Access for Netezza.
Re: JMP and HadoopHi Everyone,
Just a quick update. I was successfully able to connect to Hadoop using the out of the box ODBC driver provided by Hortonworks. The process to install and configure was straightforward using the Mac options. The process for Windows is quite similar. I am happy to confirm that the functionality using the sample data provided in the test sandbox with both query builder and the open table platforms worked quite well.
As Dan mentioned previously, our preferred method is using the SAS/Access engine because of the speed and standardization that it provides out of the box.
Please let me know if you have any questions or would like additional information.
Re: JMP and HadoopHi Kemal --
In this case, I'd just recommend using the ODBC driver. No SAS is involved. But the SAS/Access to Hadoop affords greater performance, easier access to sampling tools through the interface and better flexibility in terms of configurability.
You can use the ODBC driver for different distributions to access Hadoop cluster like other databases, but I think our strategy should be to recommend the SAS/Access products for "accessing data in Hadoop optimally".
Re: JMP and HadoopI agree data access/manipulation/viz. would be much easier in JMP . It sounded to me like JMP Hadoop connection needed a SAS bridge, I might have misunderstood. I've heard from customers who wanted to use JMP with Hadoop ( or exploring the opportunity if they could ) but didn't have SAS.
Re: JMP and HadoopHi Kemal --
What would be the use case here? I'd think getting the data straight into JMP via ODBC driver or through SAS/Access to Hadoop would be cleaner?
Seems like most of the data access/manipulation/data viz. examples in this tutorial would be done easier/better straight in JMP. Maybe I am missing something?
Re: JMP and HadoopIt looks like Hadoop and Excel can work together, can this be implemented for JMP; as in JMP -> Excel -> Hadoop?
Here is a link that describes the process:
Hadoop Tutorial: How to Use Excel 2013 to Analyze Hadoop Data
