Page 252 -

Appendix A
Green IT Checklist and Recommendations 217

3. Exploit virtualization to reduce the amount of storage net-
working equipment. This includes SANs and using virtual I/O
connections within servers such as IBM’s HiperSockets™, Virtual
Ethernet for Systems p and i, and OSA integrated layer 2 and 3
switching.
4. Use integrated approach to server consolidation to optimize
savings. More than a single methodology needs to be applied to get
the fewest number of servers. IBM’s Enterprise Computing Model
for reducing thousands to about 30 large centralized servers used
the following approach to consolidation:
♦ Migrate servers delivering largest savings first (for example,
stranded infrastructure). This primes the pump and generates
enthusiasm and savings for other green projects.
♦ Eliminate assets with lowest utilization first. These assets are not
pulling their weight when measured by watts/logical image or
other common metrics to compare servers.
♦ Identify assets with an upcoming compelling event to mitigate
expense (upgrade, move, asset refresh). It is always easier to have a ptg
positive ROI and be green within the normal refresh of assets.
♦ Aggregate by customer work portfolio to leverage strong customer
buy-in. Ease of migration assists speed and successful workload
migrations.
♦ Start with oldest technology first because it uses the most power
and provides the least performance.
♦ Focus on freeing up contiguous raised floor space. This enables
growth and the addition of energy-efficient new IT and facilities
equipment.
♦ Provision new applications to the mainframe or another large cen-
tralized server.
5. Drive to high-utilization rates. Virtualization and management of
workloads are key. The operating system must manage W/L to busi-
ness priorities and dispatch in an automated manner. The average
Wintel server is used only 5 percent to 15 percent of the time. No
manager would allow his people to work 5 percent to 15 percent of
the time. With new technologies and automation, utilization rates
can go beyond 50 percent and at the same time improve flexibility
and responsiveness as more resources can be tapped for peaks.