The preface
We can focus on my public number, the number public in the layout better, read more comfortable.
Body part
Here it clear that a lot of mess website or when people reprint articles not indicate the source, I do not notice. What's wrong deserve.
We all know the high cost of SAP HANA, the reason is the memory databases, in-memory computing, largely thanks to hardware support
The SAP product there is also a great deal of data, that is, BW, now also launched a BW4
But in this age of advanced Internet, the data want to quickly increase, and therefore the question arises
SAP provides the option (dynamic tiering) heat in the data stored in the disk-based stores managed directly by the HANA
This is much cheaper than HANA in memory, thereby increasing the price of the solution to the proportion of memory
With the growth of data storage and processing needs of customers, SAP must provide a method of utilizing Big Data / Hadoop as cold storage
A common strategy is to use SDA to access Hadoop through Hive or Spark
Most people do not realize is that this is not a good way to use Hadoop (here should be proud face, O (∩_∩) O ~ ha ha)
By simply sending a query to Hadoop, the data processing will be returned to the HANA
HANA is sending data, the SDA is not the right way to big data
HANA need to be able to be injected into the query Hadoop node and use the processing power of Hadoop
For example, to very large data sets in, you can not load it into memory HANA
Therefore, an important guest appearance, is the main coffee played, our SAP Vora came
SAP Vora is located in the top layer in Hadoop Spark.
The role of SAP Vora just let HANA use Hadoop to handle intensive work.
Of course, there are other effects, such as support in line with HANA hierarchy and currency.
Vora shortcomings in its size requirements for Hadoop cluster is three times, and increased SAP licensing costs.
This may temporarily harm SAP's big data.
However, Vora still new, over time, we will see more improvements, perhaps more viable SAP licensing policy.