Cisco UCS M5
Today’s data center is made up of hypervisors that control and provision compute resources and software that defines and allocates storage capacity. The software defined data center still needs an engine – and that engine is the server. For the large data center and enterprise selecting the right server platform is more critical than ever.
The Server Requirements of the Software Defined Data Center
The software defined data center brings flexibility to the enterprise but it needs a strong foundation from which to start, the server. The compute power is needed if the organization is looking to converge compute and storage or if it has decided to have dedicated tiers for each resource but still have those tiers be software defined.
When selecting the server hardware for the software defined data center, there are some key capabilities that IT planners should consider. First, the hardware by itself is only part of the story. How well it works with the software defined options available on the market, as well as the various monitoring and management tools available, is critical. The organization wants these two sides of the coin, server hardware and software, to eventually become one, a system that solves problems.
That system needs to be versatile so it can drive the various workloads the organization has. The server vendor needs to offer a variety of physical solutions as well as make sure those solutions each address as wide a range of workload types as possible. Those various server options, though, need to be unified in the way they are managed. IT can’t afford to manage a hundred or a thousand different compute points.
The server’s biggest challenge may be the rapidly evolving data center. The organization can’t afford to change out physical hardware as quickly as it is adopting new software or upgrading it. There needs to be some modularity to platform so it can grow to meet new challenges without necessarily being replaced.
CISCO UCS M5 Servers
To meet the requirements of the software defined data center Cisco recently announced its 5th Generation of UCS Servers. There are five servers in the portfolio, two blade servers and three rack servers. There are two key themes coming out of this announcement, more storage and more graphics processing units (GPU) for virtual desktop infrastructures (VDI).
IT needs more storage capacity and they want that capacity closer to the CPU. This demand is being driven largely by convergence and hyperconvergence, as well as distributed application. The demand for more storage is particularly difficult to deliver in a blade server configuration because of the tight space customers don’t want to have to use a portion of the storage capacity for the booting of the system and the storage area for applications.
Cisco is easing the blade storage problem by enabling customers to either use a M.2 flash device or SD Card as the boot and application code storage area. By leveraging these small form factor flash technologies, Cisco is dedicating the rest of the storage to data or the organization can put no additional storage locally and access all data from a shared storage device.
The dedicated boot area also creates space for the implementation of an additional GPU for B200 M5 and two for the B480 M5, it is actually two additional GPUs which addresses the second request.
The rack servers are also aggressively addressing the demand for more storage bays, more NVMe slots and more GPU slots. There are 32 (C480 M5) vs. 2 (C460 M4) drive bays capable of NVMe support. From a capacity perspective, the C480 M5 supports 246.4TB of NVMe storage vs. 7.6TB in the C460 M4.
In addition, the entire portfolio leverages the latest Intel CPUs to increase IO capabilities and memory bandwidth.
The Software Makes a Server a System
Great hardware is only part of the solution. Server vendors need to provide or partner with software solutions that can dynamically manage and monitor the servers, the workloads running on the servers and the environment that surrounds the servers, like networking and storage.
Cisco is partnering with Turbonomic to provide workload optimization. Workload Optimization Manager is an intent-based decision engine that can dynamically monitor and manage utilization. It can scale workloads up or down based on utilization requirements. It can also reclaim and reassign resources in real-time.
Trubonomic’s works by abstracting data center resources into a supply chain of buyers and sellers. The “sellers” are the various resources the data center has like compute. The “buyers” are the various applications, workloads and users within the organization as a whole. Service levels are established to achieve a desired state for the data center.
What makes the Workload Optimization Manager unique is it is proactive instead of reactive. Most solutions on the market wait for an alert to occur. Workload Optimization Manager instead continuously adjusts based on current conditions. The goal is to eliminate, or at least reduce, the over-provisioning of data centers.
Cisco and Turbonomics have fine-tuned the solution for UCS so the servers are providing a very deep level of information which enables Workload Optimization Manager to be even more accurate in its management of the environment. Cisco claims that with Workload Optimization Manager, workload density (VMs per server) can increase by as much as 30%.
Cisco also updated its infrastructure automation solution UCS Director to version 6.5. It essentially enables the data center to provide infrastructure as a server. It provides automated workflow schedule management. The new release includes enhancements to the workflow designer and an improved HTML5 interface with intuitive resource utilization heat mapping.
The new release also integrates with UCS Workload Optimization Manager so Optimization Manager can take over the execution of defined workflows. It also includes infrastructure deployment for FlexPlod, Enhanced automation for Cisco HyperFlex and, of course, added support for UCS M5 and S-Series servers.
Servers are still the engine that drives the data center, software defined or not. But today’s servers need to meet a different standard than in years past. They need greater flexibility, increased storage density, increased storage performance and long term future proofing.
Cisco’s UCS line of servers clearly has the momentum in the data center. It seems every storage vendor we talk to is partnering, in a big way, with Cisco on its USC servers. In addition, according to IDC, Cisco ranks first in integrated infrastructure and first in the U.S. in blade servers. It has also risen to third in global revenue for x86 servers.
Cisco’s server progress is impressive, especially considering that USC first came to market in 2009, over a decade after most of its competitors. But it is also a testament to the fact that data center modernization requires a new way of thinking, and servers need to become systems to participate in that modernization effort.