玉面飞龙的BLOG

InfiniBand architecture has the following communication characteristics:

1. User-level access to message passing
2. Remote Direct Memory Access (RDMA) in read/write mode
3. Up to a maximum of 2-Gb message in a single transfe

zt from ” Oracle 10g RAC Grid,Services & Clustering ”

The demands of the Internet and distributed computing are challenging the
scalability, reliability, availability, and performance of servers. InfiniBand
architecture represents a new approach to I/O technology and is based on
the collective research, knowledge, and experience of the industry’s leaders
and computer vendors.
InfiniBand architecture specifies channels that are created by attaching
host channel adapters (HCAs) within a server chassis to host channel adapters
in other server chassis. This is done for high-performance IPC and to
target channel adapters connecting Infiniband-enabled servers to remote
storage and communication networks through InfiniBand switches. Infini-
Band links transfer data at 2.5 Gbps, utilizing both copper wire and fiber
optics for transmission. It can carry any combination of I/O, network, and
IPC messages.
InfiniBand architecture has the following communication characteristics:

1. User-level access to message passing
2. Remote Direct Memory Access (RDMA) in read/write mode
3. Up to a maximum of 2-Gb message in a single transfer

The memory protection mechanism defined by the InfiniBand architecture
allows an InfiniBand HCA to transfer data directly into or out of an
application buffer. To protect these buffers from unauthorized access, a process
called memory registration is employed. Memory registration allows
data transfers to be initiated directly from user mode, eliminating costly
context switches to the kernel. Another benefit of allowing the InfiniBand
HCA to transfer data directly into or out of application buffers is that it can
remove the need for system buffering. This eliminates the context switches
to the kernel and the need to copy data to or from system buffers on a send
or receive operation, respectively.
InfiniBand architecture also has another unique feature called a memory
window. The memory window provides a way for the application to grant
28 2.1 RAC components
remote read and/or write access to a specified buffer at byte-level granularity
to another application. Memory windows are used in conjunction with
RDMA read or RDMA write to control remote access to the application
buffers. Data could be transferred by either the push or pull method (i.e.,
either the sending node would send [push] the data over to the requester, or
the requester could get to the holder and get [pull] the data).
While InfiniBand is a fairly new technology, it promises tremendous
potential and benefits in a clustered configuration where high-speed data
transfer is required via the cluster interconnect. While RAC has been
tested to work under InfiniBand, as the popularity of this technology
evolves, more and more implementations using this technology will
become commonplace.