Using optimized packages

Many of the functionalities in base R have alternative implementations available in contributed packages. Quite often, these packages offer a faster or less memory-intensive substitute for the base R equivalent. For example, in addition to adding a ton of extra functionality, the glmnet package performs regression far faster than glm in my experience.

For faster data import, you might be able to use fread from the data.table package or the read_* family of functions from the readr package. It is not uncommon for data import tasks that used to take several hours to take only a few minutes with these read functions.

For common data manipulation tasks—like merging (joining), conditional selection, sorting, and so on—you will find that the data.table and dplyr packages offer incredible speed improvements. Both of these packages have a ton of useRs that swear by them, and the community support is solid. You'd be well advised to become proficient in one of these packages...