EMC Connectrix EC-1200 Planning Manual page 160

Windows NT Consolidation Design Example
B
Connectrix v6.02 Planning Guide
B-4
Manageability factors include capacity and performance.
Capacity management is primarily concerned with the planning
and executing of new capacity.
Performance management wants to insure that under worst-case
loads (whether it is a specific time of day, as in Exchange at 8
A.M., or during a specific backup window), I/O bottlenecks will
not occur. The dynamic nature of the environment makes it
equally important to consider that bottlenecks will occur and,
when they do, a topology that facilitates diagnosing and
correcting the bottleneck is equally important to the design of the
logical topology.
The availability considerations of the logical topology are illustrated
in Figure B-1 on page B-7. In that example, the Data Center generally
scales its server capacity per application and per department. Each
server runs one application: file, print, SQL, and so on. If the total
workload of multiple departments on a specific application can be
handled by one server, a single server is deployed.
In the example in Figure B-1, three servers fill the print server needs
of six departments. However, the Exchange workload requires the
computing power of a server for each department.
Applying the first logical topology alternative (each Symmetrix port
serves multiple server types) based on this server scaling strategy, the
impact of a failure of a Symmetrix port would be felt across multiple
applications, and most likely multiple departments. The advantage to
this alternative is that not all of any one application or department
will go down. Under the second topology alternative (which
consolidates the Symmetrix port to an application), the impact of a
failure of the Symmetrix port would be the failure of a single
application across all departments.
Given a choice between (a) one department complaining about an
SQL outage, another complaining about print services, and a third
complaining about communication services, or (b) all departments
complaining simultaneously about an SQL server outage, the Data
Center manager in this example chooses the latter, going with the
second alternative. In this case, the second alternative might facilitate
more rapid fault isolation.
The logical topology decision is checked against the capacity and
performance needs of each of the applications.