Data in Memory Important for Deep Data Analytics
It still takes several orders of magnitude longer to read or write data to a magnetic disk when compared to accessing it directly in memory.
In-memory databases are becoming increasingly popular and an ever-more important factor in performance-critical activities such as stream processing and deep data analytics. In this DataBase Journal article, Julian Stuhler takes a look at the world of in-memory databases: a technology thats both reassuringly familiar and intriguingly novel at the same time.
"The basic concept of data in memory is certainly nothing new. Early hard disks were slow by todays standards, but even with the increased rotational speeds and bit density of todays technology it still takes several orders of magnitude longer to read or write data to a magnetic disk when compared to accessing it directly in memory. 'I/O avoidance' is therefore a major preoccupation for anyone concerned about the overall performance of an IT system, and a great way of avoiding I/O is to keep data in memory once it has been read, so that next time its needed it can be accessed immediately."