Database Reference
In-Depth Information
A single rack Netezza system is shown in Figure 4-1. As you can see, it has two
high-end rack-mounted servers called
hosts
, which function both as the external
interface to the Netezza appliance and the controller for the massively parallel
infrastructure. The hosts are connected through an internal network fabric to a
set of
snippet blades
(we'll often refer to them simply as
S-blades
), where the bulk
of the data processing is performed. The S-blades are connected through a high-
speed interconnect to a set of disk enclosures where the data is stored.
If any component fails, be it the host, disk, or S-blade, recovery is automatic.
The range of automatic failure detection and recovery options demonstrates
the advantages of an appliance approach. Netezza has built-in component
and operational redundancy, and the system responds automatically and
seamlessly to the failure of critical components.
A Look Inside the Netezza Appliance
Every component of Netezza's AMPP architecture, including the processor,
FPGA, memory, and network is carefully selected and integrated to yield a
balanced overall system. In addition to raw performance, this balanced
architecture can deliver linear scalability to more than a thousand processing
streams executing in parallel, while offering a very economical TCO. These
parallel streams work together to “divide and conquer” the workload. Let's
examine the key components of a Netezza system, as shown in Figure 4-2.
Netezza Hosts
Hosts are the primary interfaces to the system. The hosts are high-perfor-
mance Linux servers that are set up in an active-passive configuration for
high availability (note that Figure 4-2 has two of them stacked on top of each
other). The host compiles SQL queries into executable code segments called
snippets
, creates optimized query plans, and distributes the snippets to the
MPP nodes for execution. The active host presents a standardized interface to
external tools and applications, such as business intelligence reporting tools,
data integration and ETL tools, advanced analytics packages, backup and
recovery tools, and so on. These external applications use ubiquitous proto-
cols, such as JDBC, ODBC, and OLE DB, to interface with the Netezza host.
Hosts run both high-availability (HA) clustering software and mirroring
technology. All write operations to the active host are mirrored to the standby
host, preparing it to take over system processing at any moment if a failure
should occur. The standby host monitors the active host through Ethernet
links and assumes the primary role should the active host fail.
