Database Reference
In-Depth Information
enclosures are connected to the S-blades through high-speed interconnects
that enable all of the disks in Netezza to simultaneously stream data to the
S-blades at the maximum rate possible.
There are two independent data paths from each S-blade in the chassis to
each disk. Each drive is mirrored in a RAID 1 configuration. In the event of a
disk failure, processing is completed on the mirror without any interruption
of service. I/O is simply redirected to the mirror drive by the storage subsys-
tem. Spare drives are included in each appliance so that it can transparently
self-heal by selecting a spare to take the place of the failed drive and regener-
ating its contents to restore redundancy.
Network Fabric
Internode communication across Netezza's MPP grid occurs on a network
fabric running a customized IP-based protocol that fully utilizes the total
cross-sectional bandwidth of the fabric and eliminates congestion even under
sustained, bursty network traffic. The network is optimized to scale to more
than a thousand nodes, while allowing each node to initiate a large data
transfer to every other node simultaneously.
Netezza's custom network protocol is designed specifically for the data
volumes and traffic patterns that are associated with high-volume data ware-
housing. It ensures maximum utilization of the network bandwidth without
overloading it, thereby allowing predictable performance close to the data
transmission speed of the network.
Traffic in a Netezza system flows smoothly in three distinct areas:
•
From the host to the snippet processors (1 to 1000+) in broadcast mode
•
From snippet processors to the host (1000+ to 1), with aggregation in
the S-blades and at the system rack level
•
Between snippet processors (1000+ to 1000+), with data flowing freely
on a massive scale for intermediate processing
The Netezza network fabric infrastructure has the same level of redun-
dancy as the other major components. In fact, each Netezza appliance has
two completely independent networks, and because each of the redundant
hosts is connected to one of these redundant networks, the loss of any piece
of the network fabric can be overcome by switching to the standby host. The
data network is not only redundant, but also physically distinct from the
separate management network, enabling the system to assess the health of its
components even if the data network experiences problems.
