Databases Reference
In-Depth Information
SAN
(Database
Storage)
Database
Server
Application
Server
End-User
Receives User
Query
Receives User
Query
Receives User
Query
Process query
with optimizer
Generates
execution plan
Sends I/O requests
for data
Waits for response
Scans disk for data
Parses the data
Sends data back to
the database
server of
requestor
Issues a query
Waits for
response
Preprocess
Sends request to
database server
Waits for response
FIGURE 8.1
Classic systems architecture.
picture. This isolated problem addressing technique needs to change for designing and addressing
scalability and performance demands of today's users.
An additional aspect to keep in perspective is the ever-changing nature of data and its associated
structures. The impact of data itself is one of the most ignored areas that affect workloads of the data
warehouse and datamarts. There are two categories of data that adversely affect workloads due to
their volume and special needs for security in certain categories:
●
Persistent data
—data that is needed to be maintained for compliance and regulatory reasons and
cannot be deleted or archived for a period of time; for example, customer data and sensitive data
fall into this category.
●
Volatile data
—data that is of value for a shorter life span, but once the intelligence/analytics are
drawn, the data is no longer useful to maintain; for example, a large portion of transactional data
falls into this category.
Understanding workloads
In a classic systems architecture approach as shown in
Figure 8.1
, the workload management today is
centered on the database. The primary reason for this situation is due to the fact that we can analyze
what happens in the database when a query is submitted and, based on the results, interpret if the
issues are within the database or outside the database. As a result of this approach, today we define
workload from the database perspective alone regardless of what happens in the entire system. The
results that are discussed for workload and throughput revolve around the statistics like CPU usage,
elapsed time of query, the number of SQL executions, and total users on the database. Based on the
