When displaying geometric primitives I can use vertex arrays way so applications (clients) will continue to send data to opengl kernel (service).
If these data are real-time (or possibly controlled by a short-time timer),Whether the cost of the creating buffer zones in the server and updating operations can match the way of sending vertex array data to a kernel of opengl?

In other word,using buffer object does not bring any benifit on this condition?

Another question,is Buffer memory space located in the graphics board or the ddr3/ddr2 of computer motherboard?