| |
|
   |
|
        |
|
 |
|
CLUSTERS
High Availability (HA) - HA typically provides a fail-safe environment through redundancies in hardware, software and middleware. High Performance Computing (HPC) - HPC typically embraces large scale parallel applications to aggregate computation processing power, memory, or I/O subsystem. Logical Compute Farms (LCF) - LCFs typically consist of many identical compute nodes whose numbers can vary with demand over time and jobs allocate through load balancing. High Availability All high availability solutions rely on some amount of built in redundancy. At the simplest level this redundancy might involve replication of business critical data. At the other extreme high availability involves complete duplication of the solution stack to a physically separate location. Many other high availability solutions lie between these two extremities and involve redundancies in hardware, software, storage and network components. High Availability clusters usually involve two or more systems connected together via a common interconnect (or heartbeat), share a common storage subsystem and have equal access to available resources. In an event of component failure, high availability clusters necessitate failover of the defective component or a subset of the solution stack onto one or more alternative systems. High Performance Computing HPC clusters are typically made up of a large number of compute nodes connected through high cluster interconnect. Clusters numbering into several hundreds compute nodes are not uncommon. There are two dominant architectures in parallel computing: shared-memory systems and distributed-memory systems:
HPC clusters are typically distributed memory systems that may use SMP systems as a building block. Very large application data may be required to be distributed across the HPC cluster and every compute node must interact with the others to move data between processors. Large blocks of contiguous memory requires a high speed interconnect whereas small message packets require low latency interconnect to accelerate parallel program execution. Logical Compute Farm |
|
RELATED LINK HP
High Availability Solutions
|
|
The
concept of computer clusters has been established in many guises
over many years. The degree of cluster sophistication
varies greatly from established proprietary OpenVMS clusters, noted
in part for its maturity and high degree of fault tolerance, to “do-it-yourself” high
availability clusters based on commodity hardware components. The
concept of clustering, in its simplest form, joins two or more disparate
systems in such a fashion to pool available system resources (CPU,
memory, I/O etc), provide a scalable architecture for evolving capacity
management, or to build in redundancy for achieving a higher level
of availability to reduce unplanned downtime. Clusters are commonly
categorised into the three cluster types. However, a clear cluster
distinction can be somewhat indistinct:
Great competitive demands are being placed on
corporate IT resources as most research, product development and
mission critical business
applications rely heavily on the availability of computational
resource, project data and business databases. Failure of IT
systems can quickly
cascade into an operational failure across an entire business.
Moreover, server and applications are expected to be available
24 hours a day,
seven days a week with no room for downtime.
For the last three decades, supercomputer design has focused on expensive
specially designed vector computing and massively parallel symmetric
multiprocessor computing platforms. Recently there has been a shift
toward parallel cluster computing that uses commodity “off-the-shelf” components
connected together by a high speed internal network. HPC clusters
are typically deployed for parallel computing to aggregate more
processing power or effective memory for a solution of problems
within the scientific or research and development arenas. The trend
to deploy HPC clusters is clearly demonstrated in the number of
clusters listed in the “TOP500 Supercomputer Sites” list,
which lists the 500 most powerful supercomputers in the world,
and makes clustered systems the most common high performance computer
architecture.
Logical Compute Farms (LCF) provide a single interface to a loosely
coupled set of commodity compute nodes that can dynamically increase
or decrease in response to application demand. Common examples
of Logical Compute Farms include Web server farms and Internet
messaging
services. Jobs are typically allocated onto the LCFs through
batch queues or server load balancing network switches. This
kind of
cluster also provides significant and transparent redundancy
through the
horizontal scalability of compute nodes and has many attributes
of high availability clusters. In other ways, LCFs aggregated
computational power has many attributes of high performance compute
cluster.
