Building Software for Tomorrow’s Big Data Challenges

TRANSCRIPT

Distinguished Professor of Computer Science Xian-He Sun: Hermes is a software system for I/O systems, for high-performance I/O systems. 

I/O means input/output. So the I/O systems, you may ask why I/O systems become so important. 

Everybody knows big data. The data becomes bigger, so when data becomes bigger, you cannot just put it into memory. You have to get into the storage, and then you get this input/output. 

So you can see why this I/O problem has become more important now than before—because of big data.

Assistant Research Professor Anthony Kougkas: The main differentiating factor that Hermes addresses is the hierarchical nature of future supercomputers. 

Hardware innovation has given us new, very fast storage devices, like flash-based SSDs and NVMEs; even in our personal laptops we have options to get faster drives.

These new innovations in hardware, they have to be accompanied by innovations in software, and how can we write software that manages this new hardware and simplifies life for the end user is still required.

The main thing that Hermes brings to the table is that it abstracts the complexities of the underlying hardware and makes it easier for the user to leverage this new hardware and get more performance with less coding.