Zero memory buffer copying in a reliable distributed computing system
Reducing buffer space that is reserved when a VM attempts to write to disk. A shared memory maintains stored information available to all VM's, allowing translation between accessible to RSD's, independent of how that VM's storage is divided into VSD's. When a guest OS writes to a VSD, an “extent”, designating a location in the shared memory and amount of data to be transferred, which can be passed directly from the guest OS, through each element involved in translating the operation into one or more operations on RSD's, keeping copying to a minimum. No substantial changes need be made to the guest OS.
1. A method,
in a reliable distributed computing system, said system including one or more virtual machines, each said virtual machine having access to a plurality of virtual storage devices emulated in a memory commonly available to all said virtual machines,
of minimizing an amount of buffer space in the commonly available memory reserved for virtual storage device operations, said method including steps of
coupling the commonly available memory to said virtual machines, said commonly available memory maintaining information accessible to said virtual machines, wherein said information is allocated among said virtual storage devices;
receiving data with respect to an operation directed to one or more of said virtual storage devices;
allocating an extent within said shared memory, said extent including one or more regions of said shared memory and metadata defining said extent including designating locations and sizes of said one or more regions; and
sending said metadata defining the extent without data in said extent along a call chain disposed to perform said operation, said steps of sending along said call chain including steps of moving said metadata defining the extent from at least one process or thread disposed at a guest operating system to at least one process or thread disposed at a host operating system;
wherein said data with respect to said operation is not itself sent along said call chain.
2. A method as in claim 1 , including steps of
by a first said guest operating system, providing a first pointer to a first block of data in said shared memory;
by a second said guest operating system, providing a second pointer to a second block of data in said shared memory; and
by a host operating system, defining a first extent in response to said first pointer and defining a second extent in response to said second pointer, notwithstanding that said first pointer and said second pointer have distinct formats.
3. A method as in claim 1 , wherein
said extent is defined by said metadata, said metadata including
a starting location in said shared memory and a length of said data for which said operation is to be performed.
4. A method as in claim 1 , wherein
said guest operating system provides said metadata defining said extent in said shared memory,
said metadata being responsive to one or more virtual storage devices accessible to said guest operating system, and
said guest operating system sends said metadata defining said extent and a description of said operation to be performed to a process or thread in said host operating system; and
said host operating system receives said metadata defining said extent and said description of said operation to be performed,
said host operating system provides further information responsive to one or more real storage devices accessible to said host operating system,
in response to said metadata defining said extent and said description of said operation to be performed, and
said host operating system performs said operation to be performed in response to said further information to said one or more real storage devices.
5. A method as in claim 4 , wherein
said metadata defining said extent in said shared memory is responsive to a pointer to a data structure,
said pointer being formatted according to said guest operating system, and
said metadata being transformed from said pointer formatted according to said guest operating system into a data structure formatted as defined by said host operating system.
6. A method as in claim 4 , wherein
when said operation includes a read operation,
said host operating system provides data to be read in response to a result of said operation in said one or more regions of said shared memory defined by said extent, and
said host operating system returns a response to a result of said operation with respect to said metadata defining said extent to said guest operating system.
7. A method as in claim 4 , wherein
when said operation includes a write operation,
said guest operating system provides data to be written in one or more regions of said shared memory, and
said guest operating system provides said metadata defining said extent in response to said one or more regions.
8. A method as in claim 1 , wherein
when said operation includes a read operation,
data to be read need not be copied from a first portion of said shared memory to a second portion of said shared memory, and
data to be read need not be copied from said host operating system to said guest operating system.
9. A method as in claim 1 , wherein
when said operation includes a read operation,
said process or thread disposed at said host operating system fills said extent with data after said operation is performed.
10. A method as in claim 1 , wherein
when said operation includes a write operation,
data to be written need not be copied from a first portion of said shared memory to a second portion of said shared memory, and
data to be written need not be copied from said guest operating system to said host operating system.
11. A method as in claim 1 , wherein
when said operation includes a write operation,
said process or thread disposed at said guest operating system fills said extent with data before said operation is performed.