How Cisco IT Reduces the Cost and Complexity of Server Fleet Refresh

With 12,000 Cisco UCS servers installed in 22 company data centers worldwide, Cisco IT had a big challenge tracking when individual UCS units were due for replacement as part of our infrastructure refresh program. Compounding the problem: Many different server models were installed at different times, making it very difficult for us to manage refresh planning, budgeting, and scheduling.

In the past, we typically used the server’s end-of-sale or end-of-support date as the basis for replacement planning. However, it may make sense to replace or upgrade a server sooner, especially when considering the higher performance and lower power consumption delivered by newer models.

To simplify the refresh planning, we now replace Cisco UCS units on a standard schedule of four years after installation, meaning we will refresh 25 percent of our server infrastructure every year.

We already see significant cost savings from our first-year implementation of this strategy, especially because of the higher capacity and performance of the UCS models. For example, we reduced the number of blade servers by as much as 50 percent, while maintaining the same number of virtual machines (VMs). We also reduced the need for bare-metal servers by one-third to one-half. These reductions mean we use less data center space and cooling and saw 1 MW of electricity savings in the first year.

Establishing a standard refresh schedule also brings several other benefits:

  • We are better able to predict budget needs for server replacement based on the consolidation levels we’ve already seen. We are also able to better forecast the associated data center costs.
  • The effort required to manage servers is now 40 percent lower because of fewer UCS servers installed and fewer differences in models and configurations.
  • We now install defined Cisco UCS server and cluster models, chosen because they are adequate for most application needs and avoid the need to install excess capacity.
  • A standard offering for bare-metal servers makes it easier to standardize security measures. This standard also means we can deliver a bare-metal server in a few days, compared to as many as 90 days previously required for the server ordering process.
  • We gained easier tracking of financial depreciation for the installed equipment because it is replaced on a consistent, known schedule.

In our role as Cisco’s Customer Zero, we work with the UCS business unit to help them understand the capabilities needed for our standard server models. We are also involved in product testing, such as to determine the appropriate VM capacity on an individual UCS blade.

Opportunities for Automation

A standard refresh cycle will allow us to automate many migration processes by providing proactive visibility of assets to refresh, as well as better downtime planning, change management, and actual migration. This automation will save time and effort for IT, reduce potential impact on application availability, and allow faster access to improved server performance levels by application teams and users.

Currently, we have automated migration of VMs from older servers to new UCS blades with no disruption to the application. We are now working to automate the processes for servers that host dedicated application workloads to allow for earlier planning and reduce the migration time and effort. For example, early planning will help application and hosting teams ensure that all software and licenses are ready to move to the new servers as soon as they are installed.

Expectations for Future Savings

For the next three years, we will continue to have a complex mix of servers until the new refresh plan is complete. Throughout this process and into the future, we expect to continue seeing cost savings and operational benefits as our server environment becomes simpler and standardized.

What strategies do you use to simplify a server refresh program? Share your tips in the comments.



In an effort to keep conversations fresh, Cisco Blogs closes comments after 60 days. Please visit the Cisco Blogs hub page for the latest content.

1 Comments

  1. Karan,

    Very good article.

    HAM program under BCSP maintains compute asset inventory from ordering to decommission, and provides visibility to assets by its various dimensions (age, EoL/EoX, utilization). HAM has recently automated quarterly reporting of aging blades (36 months), to help Fleet Managers identify and migrate dependent IT services (app/DB) from aging hardware, and refresh assets periodically instead of as fiscal program. HAM can provide analytics to help compute teams maintain inventory at optimum levels, reduce the risk of running IT Services on aging hardware, improves utilization and thus save money for Cisco IT.

    We would like to partner with you to provide software solutions to help your initiatives.