[sheepdog] [PATCH 1/9] Doc. "Sheepdog Advanced" add chapter "more concepts"
sirio81 at gmail.com
Fri Oct 11 17:53:10 CEST 2013
Signed-off-by: Valerio Pachera <sirio81 at gmail.com>
doc/more_concepts.rst | 61 +++++++++++++++++++++++++++++++++++++++++++++++++
1 file changed, 61 insertions(+)
create mode 100644 doc/more_concepts.rst
diff --git a/doc/more_concepts.rst b/doc/more_concepts.rst
new file mode 100644
@@ -0,0 +1,61 @@
+Qemu, as any other process, is going to forward I/O request to the OS
+(reading and writing to disk).
+Sheepdog is a layer between qemu and the OS and it's scope is not to store data
+on the server were qemu is running, but distribute along the nodes composing the
+This layer (sheepdog), is made os several components:
+**QEMU block driver**
+that sends I/O request to sheepdog insted of the OS.
+The requests of "qemu block driver" are handled by the gateway.
+E.g., when qemu makes a read reqeust, not all the necessary objects are stored
+on the node were qemu is running.
+The gateway will take care of retriving objects from the other nodes to
+accomplish the request.
+Viceversa, for a write request, it will calculate were to store the extra copy
+of each single object.
+The gateway doesn't write or read directly form disk. It only manages the
+Tey are forward by the gateway to the Object Manager that will write to local
+Before the implementation of Multy Disk (that allow a single sheep daemon to
+manage more disks), it was necessary to sun sheep N time for N disk in the same
+Each sheep daemon was consider as a node.
+This way, more than 1 copie of an object might been store on the same host.
+To avoid that (to have then 1 copie per object per host), it was necessary to
+associate a zone all sheep daemons running on the same host.
+ sheep -z 1 ...
+.. |Architecture Overview| image:: https://github.com/sheepdog/sheepdog/wiki/architecture.png
\ No newline at end of file
More information about the sheepdog