Silicon Graphics ICE X, User Manual

Page: 64 / 142

44 007-5806-004
3: System Overview
configurations. Users login or connect to submit jobs to the compute nodes. The node also acts as
a gateway from InfiniBand to various types of storage, such as direct-attach, Fibre Channel, or
NFS.
Optional Lustre Nodes Overview
The nodes in the following subsections are used when the SGI ICE X system is set up as a Lustre
file system configuration. In SGI ICE X installations the MDS and OSS functions are generally
on separate nodes within the ICE X system and communicating over a network.
Lustre clients access and use the data stored in the OSS node’s object storage targets (OSTs).
Clients may be compute nodes within the SGI ICE X system or Login, Batch or other service
nodes. Lustre presents all clients with a unified namespace for all of the files and data in the
filesystem, using standard portable operating system interface (POSIX) semantics. This allows
concurrent and coherent read and write access to the files in the OST filesystems. The Lustre MDS
server (see “MDS Node”) and OSS server (see “OSS Node”), will read, write and modify data in
the format imposed by these file systems. When a client accesses a file, it completes a filename
lookup on the MDS node. As a result, a file is created on behalf of the client or the layout of an
existing file is returned to the client. For read or write operations, the client then interprets the
layout in the logical object volume (LOV) layer, which maps the offset and size to one or more
objects, each residing on a separate OST within the OSS node.
MDS Node
The metadata server (MDS node) uses a single metadata target (MDT) per Lustre filesystem. Two
MDS nodes can be configured as an active-passive failover pair to provide redundancy. The
metadata target stores namespace metadata, such as filenames, directories, access permissions and
file layout. The MDT data is usually stored in a single localized disk filesystem. The storage used
for the MDT (a function of the MDS node) and OST (located on the OSS node) backing
filesystems is partitioned and optionally organized with logical volume management (LVM)
and/or RAID. It is normally formatted as a fourth extended filesystem, (a journaling file system
for Linux). When a client opens a file, the file-open operation transfers a set of object pointers and
their layout from the MDS node to the client. This enables the client to directly interact with the
OSS node where the object is stored. The client can then perform I/O on the file without further
communication with the MDS node.