SAP And Oracle: Get Real About In-Memory Analysis
These tech giants are making plenty of promises about their Hana and Exalytics appliances. But do you really need all that power?
It slices, it dices, it will replace transactional databases, it will do away with data warehouses. Indeed, SAP sometimes makes its Hana in-memory technology sound like a digital Ginsu knife.
In the latest example, SAP announced Wednesday that SAP BusinessObjects Business Intelligence (BI) OnDemand is now powered by Hana, a breakthrough that the vendor's press release said will enable users to "quickly and cost-effectively perform detailed real-time analysis of any data."
More Software Insights
- Mobile DevOps: Achieving continuous delivery with multiple front ends and complex backends in Banking, Financial Services, and Insurance
- Maximize the benefits of virtualization for greater ROI
White PapersMore >>
Well, that's partly true, but I wouldn't connect the words "real-time" with "any data." Similarly, Oracle claimed at its recent Oracle Open World conference that the Oracle Exalytics Business Intelligence Machine due out next month will be able to run any Oracle BI or enterprise performance management (EPM) application. That's partly true, too, but do you need a Ferrari to drive to the train station or go to the grocery store? Probably not.
I think a little more reality and little less hype are in order from both companies about these products. So here's a bit of hype bubble bursting for both vendors, along with the advice for companies to think long and hard about their need probe up-to-the-second data or support complex, multidimensional analyses.
[ Want more on SAP's in-memory-powered applications? Read SAP's Hana Gets Apped. ]
Let's start with SAP's BusinessObjects BI OnDemand environment. The new news is that the paid, $25 per user, per month version of the service now runs on Hana in-memory technology (while the free version still runs on a conventional database). So you can expect real-time analysis if the data is already loaded into the BI OnDemand cloud environment running in SAP's data centers. If the data is in another cloud-based application, you could, theoretically, move it from there to the BI OnDemand cloud within the limits of Web services bandwidth constraints. (For more on that possibility, check out the Salesforce.com integration highlighted at biondemand.com/hana.)
But if we're talking about using data behind your firewall in your data center being part of the analysis, you can forget about "real, real time" within BI OnDemand. Whether it's SAP app data, SAP Business Warehouse (BW) data, or information from any other on-premises source, it will take some time--a few minutes at a minimum--to move that into SAP's BI OnDemand cloud.
A few minutes many not sound like much of a penalty, and for many uses that's probably just fine. But it won't cut it in decision-support scenarios where you're deciding on a financial transaction, dealing with volatile prices, or trying to serve up offers to customers while they're still online or on the phone.
Real-time analysis of up-to-the-second data from within transactional apps is a differentiator for Hana (and something Exalytics doesn't promise), but that doesn't apply if you have to move information from on-premise to cloud-based data centers. SAP is evaluating three possible data-integration technologies to move data from on-premises sources into BI OnDemand. SAP BusinessObjects Data Services can be used to continuously push small batches of data into the cloud via FTP. This is the technology SAP also uses to sync SAP App data with Hana whenever Oracle is used as the underlying transactional database. The Sybase Replication Server is faster than Data Services, and it's SAP's preferred option for quickly mirroring app data into Hana. Unfortunately, Oracle has balked at certifying this Replication Server as a means of moving data from Oracle databases into Hana (and I'm cynically guessing that's because Oracle can't match the speed of that combination or doesn't want to help SAP obviate the need for Oracle databases used for data warehouses). There's also SAP SLT (system landscape transformation), a product used to shuttle data between SAP apps.
SAP has yet to work out which tool will work best with the BI OnDemand service. But the bottom line is that you can expect to be able to move "hundreds of megabytes into the cloud within minutes," according to Jason Rose, SAP's vice president of solution marketing. (I couldn't get any details on whether BI OnDemand customers will face additional charges once they start loading lots of data into SAP's cloud, but I suspect there's no such thing as a free lunch.)
So Hana-powered BI OnDemand is not about real-time analysis of "any data," but once the data reaches the cloud, the in-memory technology will let you slice, dice and do what-if querying and scenario analysis very flexibly and with "speed of thought" response times. This is the other side of the in-memory story: being able to crunch lots of data and look at any combination of attributes without rebuilding cubes or using summarized/aggregated data. With Hana, massive amounts of data can be held in memory, so you can explore and correlate any number of dimensions and quickly drill down to granular levels of detail, meaning particular products, territories, customers, or salespeople.