Storage consolidation projects are permanently on the IT project whiteboard. The process typically starts after IT realizes that their data center is overrun by multiple storage systems from multiple vendors. Storage Switzerland finds that most data centers have 5-6 different storage systems often with 3-4 different vendors represented. The goal of the storage consolidation project is to lower storage expenditures, simplify storage operations (reducing soft costs) and improve services. Usually the storage consolidation project fails at meeting any of the three goals.
Interestingly those same data centers also have multiple server vendors but all claim to have consolidated servers. However, they didn’t all go buy a modern day equivalent to the mainframe, instead they consolidated at the software level with VMware, Hyper-V and other hypervisors. Storage consolidation, to be successful, needs to follow a similar track.
Why is Server Consolidation Successful?
Server consolidation is successful because it provides flexibility in hardware selection. Instead of forcing data centers to buy a giant mainframe for all their computing needs, hypervisors enable IT to purchase a single piece of software that allows disparate hardware to act as one.
Flexibility in hardware selection is a critical element to server consolidation’s success. While virtual machines (VM) can run on any host in the cluster, usually there is a specific host that IT intends the VM to run on most of the time. If the workload is compute or IO intensive, IT can make sure that its default host is equipped with extra capabilities in those areas. In the same way, if the workload is not compute or IO intensive IT can make sure its default server is on a less powerful but more cost effective server.
Increasingly, hypervisors are providing flexibility in workload location by enabling the movement of workloads to other data centers and to the cloud. Assuming (a big assumption at this point) the data is pre-positioned in the target location, the workload can transition to the cloud within a few minutes making the cloud viable for peak workloads and disaster recovery.
The other critical element in the success of server virtualization is the software. The hypervisors perform an excellent job of efficiently virtualizing the server’s computing and memory resources. While bare metal systems still outperform CPUs under control of a hypervisor the difference is negligible for most use cases, especially when factoring in the capabilities of sharing server resources. If the hypervisor only delivered 5% of the performance of the CPU, virtualization’s popularity would be nothing like it is today.
What Did Storage Get Wrong?
Most storage consolidation strategies try to consolidate down to a single storage system, creating a storage mainframe instead of a storage hypervisor. The alternative to the storage mainframe approach is software defined storage (SDS) but in most cases it doesn’t efficiently deliver on the storage IO potential of the devices it manages. The lack of efficiency limits SDS’s success. The next two blogs in this series explain in detail the problems with today’s storage consolidation strategies. The fourth blog in this series provides insight into how a hypervisor-like approach to storage is necessary for a successful long lasting storage consolidation strategy.
Storage Switzerland is also providing a live webinar that discusses the challenges with storage consolidation and discusses the various approaches to creating a sustainable strategy. You can sign up for our on demand webinar “Designing a Storage Consolidation Strategy for Today, the Future and the Cloud” and get a copy of our latest eBook, “Consolidating Storage in the Modern Data Center”, available to registrants.