Difference between revisions of "Front end"
(A start...) |
(Copyedit, add more) |
||
Line 1: | Line 1: | ||
− | The term '''front end''' usually refers to a small computer attached to a large one, which is used to offload [[input/output]] work from the larger machine. (Hence the alternative terms '''I/O processor''' and '''peripheral processor''') | + | The term '''front end''' usually refers to a small computer attached to a large one (usually a [[mainframe]]), which is used to offload [[input/output]] work from the larger machine. (Hence the alternative terms '''I/O processor''' and '''peripheral processor'''.) |
− | + | There are several advantages to offloading I/O activity to a separate computer. First, the obvious one is the ability to bring additional processing power to bear. Second, large mainframes are typically optimized for doing large-scale computations, not doing I/O. Also, the overhead involved in processing an [[interrupt]] from a device (e.g. saving [[register]] contents) can be significant, so avoiding bothering the main CPU with this (often larger in a mainframe than in a smaller machine) can produce significant efficiencies. | |
− | The front end machine can communicate with the main CPU in a number of ways. It may have direct access to the [[main memory]] used by the main CPU, a technique used in the PPUs of the [[CDC 6600]], the [[DL10]] of the [[PDP-10]] family, etc. | + | The front end machine can communicate with the main CPU in a number of ways. It may have direct access to the [[main memory]] used by the main CPU, a technique used in the PPUs of the [[CDC 6600]], the [[DL10]]-[[PDP-11]] combination of the [[PDP-10]] family, etc. |
{{stub}} | {{stub}} |
Revision as of 19:55, 20 April 2018
The term front end usually refers to a small computer attached to a large one (usually a mainframe), which is used to offload input/output work from the larger machine. (Hence the alternative terms I/O processor and peripheral processor.)
There are several advantages to offloading I/O activity to a separate computer. First, the obvious one is the ability to bring additional processing power to bear. Second, large mainframes are typically optimized for doing large-scale computations, not doing I/O. Also, the overhead involved in processing an interrupt from a device (e.g. saving register contents) can be significant, so avoiding bothering the main CPU with this (often larger in a mainframe than in a smaller machine) can produce significant efficiencies.
The front end machine can communicate with the main CPU in a number of ways. It may have direct access to the main memory used by the main CPU, a technique used in the PPUs of the CDC 6600, the DL10-PDP-11 combination of the PDP-10 family, etc.