Databases Reference
In-Depth Information
the node is completely configured, it is added to the data set. The feature is handled in different
techniques across the appliances available in the market today.
These architectural improvements coupled with new data and workload management techniques
focused on data warehouse-driven workloads and have made the DWA a popular option.
Key best practices for deploying a data warehouse appliance
1.
Clearly documented requirements:
●
Data availability requirements
●
Data scalability requirements
●
Query and analytics requirements
●
Performance requirements
●
Security requirements
●
Compression requirements
●
Data growth requirements
●
Mixed-workload requirements
2.
Implementation architecture. DWAs can be architected and deployed as a fully functional
replacement to the data warehouse or an augmentation to the current infrastructure. Plan this
deployment architecture, as it can have a different impact on the outcome based on how it is created.
3.
Data architecture. Always ensure that data is available at lower grains for the DWA. The
architecture will provide for managing complex formulas and aggregate calculations on-the-fly.
This approach will be very powerful when you look at implementing data virtualization as an
additional integration layer.
4.
Workload architecture. Workload optimization is a delicate art that needs to be handled with finesse
with the type of workload, its frequency of occurrences, and the overall impact on the SLAs. Plan
the workload architecture by documenting the individual workloads, and its infrastructure, data
consumption, baseline performance, and mixed workload performance. Once you add more data,
users, and queries, you can revisit and calibrate the architecture to sustain the overall baseline or
reset the new numbers and amend the new numbers to business users as the new SLA.
5.
Security architecture. Plan the data and user security architecture for the appliance. This is useful
when migrating from other platforms into an appliance platform. It additionally provides a clear
analysis of gaps and workarounds if they are needed.
6.
Compression. Plan for data management by compressing data on a usage basis. If data older than
120 days is not needed at detailed levels and only summary data is accessed, then we can compress
the data older than 120 days and reclaim space. In certain situations, we can alternatively use the
appliance as a backup system and archive data and compress it upon acquisition.
7.
Proof of concept. Execute a holistic proof of concept and perform it through reference checks
from users before you buy a DWA.
8.
BI tools integration. The DWA of today can support a wide variety of BI tools. Plan for utilizing
the native functions in the appliance to perform complex functions. This will truly distribute the
workload into the data layer and enable compute to happen where the data is located rather than
move the data to layers where compute needs to be performed and thereby lose the advantage of
a better architecture. While this means that some rewrites will happen in the current applications,
the benefits outweigh the implementation costs in these situations.
