Many people complain about high costs but they don’t know where they come from.What if you knew what you don’t know?
Years ago people were already using BI to answer similar questions. Then people used data from just a few sources. It was internal, structured and people waited for hours or days to get the information. Nowadays, data is exploding. People are recognizing this and want to use data from all kinds of sources and in all kinds of formats. Real-time and ad-hoc.
To whom does big data matter? The answer to every industry is: Tomorrow’s data volumes will be even larger than today’s. We have to keep processing time constant while the volumes increase. Traditional solutions do not work. Technologies like relational databases are not flexible because the data has to be structured. Results take time and are obsolete.
So, what will work?
Distributed parallel processing where everything happens in parallel and nothing is shared. By replicating data to several nodes we can simply add more nodes which creates a kind of fault tolerance.
In-memory databases are another option as everything happens in the RAM.
In-memory data grids can manage any sort of data objects and increase performance by reducing I/O.
NoSQL databases, which include a caching function and are designed for big data, are a new category of databases. They only read the relevant data and make the analysis much faster.
Big Data is not just data at rest but also data in motion. This requires a Complex Event Processing (CEP) approach. But, Big Data is not just about infrastructure. It is also about data and processes. Data needs to be really understood. It’s about asking the right questions and using the right tools. Finding the right answers requires a deep understanding of data, its uses and the intended results.
We can warn cars telling them of dangers on the road.
To do this, there is a big stream of incoming events which is sent to the CEP. The relevant data is sent to the in-memory database or an in-memory data grid and a push warning is sent to the cars.
Smart congestion recognition
We know their positions and planned routes and we integrate knowledge of, e.g. sporting events. We can then recommend a different route. Or maybe, a driver is running out of fuel and is directed to the next petrol station for which he has a loyalty card. Again the fast incoming data stream is sent to the CEP engine. The data imported from sources such as forums is sent to the distributed data store by Hadoop. All the information is sent to the in-memory database and a personalized recommendation is sent out.
Safe Driving and Phone Call
The driver’s task plan is taken into account and he taken to a low traffic area, where the system in the car calls the people in the task plan. Here, there is also a big stream of incoming events which is sent to the CEP. The relevant data is sent to the in-memory database or an in-memory data grid and a push message is sent to the cars.
The answer is an individual one. One size does not fit all.
Fujitsu’s target is to mix the technologies to achieve the customer’s goals, e.g. servers, storage or appliance or bundled solutions.
Fujitsu is a one-stop shop which lets you reduce complexity, time and risk. We consult with the customer in order to achieve their business goals.
There are a lot of usage scenarios on show at the Big Data booth. Come to the exhibition area at Fujitsu Forum 2013.