The Business Case for Storage Networks [Electronic resources] نسخه متنی

Capacity Planning and Utilization

During the boom times, it is safe to say that little was done in the way of capacity planning. With triple-digit growth rates in 1999 and 2000, the sole objective was to provide basic storage support for environments before the applications (and the business units depending on them) came to a halt.

In more recent history, the process of yearly and quarterly capacity planning for storage purchases took many hours of research and measurement for each environment. Purchases were closely examined and justified as part of either growth or a specific project or initiative. Historical trends were analyzed in an attempt to project growth rates for each quarter of the fiscal year, and although it was a labor-intensive process, it generally yielded fairly accurate data for planning purposes.

Today, storage capacity planning is typically handled as part of the ongoing consolidation strategy. All purchases are closely scrutinized; however, buffers included in the annual consolidation purchase typically carry most of the business verticals for several months after the initial consolidation. One exception to this rule is the Oracle 11i development environment, which tends to grow from 30 percent to 100 percent annually. Capacity planning for this environment is handled separately and purchases are made on a case-by-case basis apart from the quarterly planning cycles.

In 2001, allocation efficiency was estimated to be between 20 percent and 30 percent with no estimates for utilization efficiencies. Growth was manually tracked for most environments, but the inability to share storage between groups created a stove-piped environment, thereby hampering any concerted attempts to increase utilization rates.

In 2002, storage allocation efficiency reached the 40 percent range with the creation of larger SAN islands and the decommissioning of more DAS environments. Allocation efficiency since the MDS deployment is now in the 50-60 percent range and another project in the coming year will migrate the remaining DAS environments to SANs.

Storage Recovery

In late 2003 and early 2004, estimates for overall allocation efficiency approached 60 percent. Remaining DAS devices are increasingly consolidated into collapsed-core SAN fabrics built on MDS 9509s, boosting allocation efficiency rates. Data collected by the Storage Recovery Programa cross-functional program designed to address the demand side of the storage equation by reducing the number of purchases required by application groupsindicates that allocation efficiency might also be as high as 70 percent.

Note

Note that in some companies, storage recovery efforts are referred to as storage mining.

Raising the utilization efficiency rate is one of the main objectives of the Storage Recovery Program at Cisco. Piyush Bhargava, manager of the Enterprise Data Integrity team, has initiated a novel approach to the problem of storage reclamation by creating the Bytes Back Club, a rewards-based system whereby system administrators, database administrators, and project managers are eligible to receive financial rewards for releasing storage to be reused by other groups. In its first year, the SRP team, in conjunction with the Bytes Back Club, reclaimed an estimated $2.9 million dollars of unused storage.

As of February, 2004, Cisco had migrated off of Brocade and McDATA Fibre Channel switches and on to MDS-based SANs. Additional long-term goals include the migration of SANs in each datacenter into a consolidated SAN infrastructure (with separate virtual SANs [VSANs] for logical management of disks assigned to different business verticals), and the migration of all remaining DAS to either Fibre Channel or iSCSI SANs.

Cost and management savings, as well as performance increases, will be sought as part of a rollout of Linux blade farms for high availability and low-cost compute services. The concept of virtualization applies here insofar as increasing the utilization of the server resource, the CPU, lowers the TCO for the server, just as increasing disk utilization lowers the TCO for the disk.

In addition to multiple storage consolidation efforts, the storage team intends to migrate away from modem-based, dial-up access for remote monitoring of storage frames to an IP-based model that utilizes VPN access methods. The Enterprise Storage Services team will drive other projects, which include deploying iSCSI and virtualization products, outlining a template for end-to-end architecture, instituting a service-level management framework for storage services, and selecting a solution for long-distance replication.

In addition, the Enterprise Storage Services team is in the process of deploying an SATA solution to lower the TCO of one the largest single environments at Cisco, a multi-terabyte repository for code binaries.