I have recently started research work in Big Data analytics. I was wondering what kind of combination of hardware and software acceleration techniques are there to speedup the Big Data computation for analysing them?
FPGA based implementation is one solution; you could analyze hot-spots of algorithms and implement them in FPGA hardware while leaving the other parts in software.
There is a large range of solutions in Big Data analytics field, and the choice of approach largely depends on problem on hand and available resources (hardware and software). MapReduce and its implementation Hadoop have been greatly popular and there is number of solutions building on top of it such as Pig, Hive, and Spark. Database vendors also integrate Hadoop approaches with their database offerings. There are also NoSQL and NewSQL data store that emerged for storage of Big Data.
GPU processing is also getting more popular (http://www.nvidia.ca/object/cuda_home_new.html).
Combining distributed processing over number of nodes with GPU processing looks promising.
Katarina
Hybrid architectures combining processing and IO acceleration seem to be the right solution. Look at heterogeneous systems such as IBM Power8 mixing CPU and GPUs and even FPGAs in the future. On the IO side there is HP's IO Accelerator for low latency big data analytics.
as Katarina Grolinger say's MapReduce and its implementation Hadoop have been greatly popular and i think it can be a good solution.
- Badges
- Science topic
- Similar topics
- Mathematics
- Statistics
- Statistical Analysis
- Data Analysis
- Dataset
- Big Data