Today everything is bottleneck by data's that stored digitally (e.g., machine learning, artificial intelligence, genome analysis, graph analytics, databases, video analytics) . Every application especially large applications are overwhelmed by data . The key features such as , performance , efficiency and scalability are upon the amount and quality of the data movement . Nowadays one of the most hot topics in technology is ML and AI which highly upon stored datas . Day after day we advance in the software to welcome the advancing of ML and AI , But unfortunately we are not advancing rapidly in the case of hardware . So in here we are talking about how can we simply make our architectures to be better for storing data .
How Should our data be ?
Our datas should store , access and process well . But How ? How can an architecture store , access and process data well . We recommend to make our architectures
- data-centric
- data driven
- data-aware
Data Centric
We mean making the system ensuring that data does not overwhelm its components. For achieving this we should produce new and highly intelligent algorithms to design not just a system but the whole architecture to ensure that data and process is the center of the system not a component means the data is the center and we highly focus and spend energy on it .
Data Driven
This means creating and intelligent architecture and algorithms to take advantages of datas and metadata to continuously improve it decisions to make the data centric . So the architectures is driven by data not component .
Data Aware
This means creating highly intelligent algorithms for architecture to handle datas differently , meaning categorizing datas to be handled differently .
How is the architectures of today ?
Today the center of architectures are component which is processor . So every data should go through processor to process and then to output or storing . So we have no process in any of other components and data is not center of the system , Everything should go through processor to process . So during this process we loose a lot of energy and efficiency . Components such as RAM and ROM is just to store and move data , so to a simple process we should move the data to processor and process there and the moving to desired place , in this process we loose a lot of energy and timing . The purposed three suggest to process simple instructions to newly designed DDR5 RAMs that can handle this instructions which means decreasing energy and efficient and timing .