Chapter 2: Capacity and Availability Management 35
Hardware Upgrades
Exactly when hardware upgrades are required depends on the results of your capacity
planning. If you plan your capacity well, you will be able to predict when hardware upgrades
are required, which is particularly important when there are long lead times on hardware.
Failing to predict when new hardware will be required can lead to severe availability man-
agement problems, for example if a server running Exchange were to fail because you ran
out of disk space.
It can be very beneficial to perform hardware upgrades on a regular basis, standardizing
on new hardware each time. This allows you to keep your hardware consistent across each
Exchange 2000 role and therefore reduce support costs for the new environment. Always
ensure that you test Exchange and Windows 2000 thoroughly on the new hardware to
ensure that there are no unforeseen anomalies.
In many cases hardware upgrades coincide with software changes or with consolidation of
servers. If this is the case, you must ensure that the new hardware is able to cope adequately
with the new environment and with any intermediary changes you need to make to get to
your final environment. As with all areas, ensure that you plan and document thoroughly
any changes that you make to the hardware environment. If you are rolling out hardware
changes across the entire organization over a short period, this can be a very labor
intensive period for your operations department, so ensure that you plan carefully for
the appropriate skills to be assigned.
Having a clustered environment is the best way to perform an upgrade while simulta-
neously minimizing downtime. Under these circumstances, you can perform a rolling
upgrade—manually failing over the system to node A, performing an upgrade on Node B,
failing it back over again to Node B, performing an upgrade on Node A, and then return-
ing the system to normal.
You will also need to ensure that there is provision in your SLAs for hardware upgrades.
Hardware upgrades often involve periods of scheduled downtime, especially if you do not
have clusters everywhere, and you need to allow for this when you define your SLAs. With
proper planning, both scheduled and unscheduled downtime can be kept to a bare mini-
mum.
Summary
This chapter has shown you how to manage capacity and availability in your Exchange
2000 environment. You have seen performance-tuning changes that are easy to identify
and implement to obtain better performance and you have learned how to examine what
areas to consider when implementing hardware changed in your organization.