<img alt="" src="https://secure.bomb5mild.com/193737.png" style="display:none;">

Turbonomic Blog

Finding The Desired State: Cisco UCS Fabric Interconnect

Posted by Matt Vetter on May 23, 2016 7:30:41 AM

Earlier this week, I was on a demonstration discussing the concept of a desired state with a particularly interesting customer. Ordinarily, most customers and prospects I talk with consider the concept of a desired state to be a pretty simple concept. Due to the tools we have in the marketplace today, most people consider the desired state somewhere between no active alerts and no phone calls from end users. But the concept is much more complicated than that. This prospect in particular was challenging me on the concept that not only is there no single desired state, but that there was also a desired state for every level of virtualization, and the only true way to assure performance was to find a desired state for both at the same time in real time. We had only been talking at that point on how to assure service performance while maximizing the efficiency of the hypervisor layer, when the prospect brought up Cisco UCS. He mentioned that, occasionally, he finds that performance degradation occurs on his fabric layer, but does not manifest itself as an alert in his vSphere environment. He then challenged me and asked, can VMTurbo not only control a hypervisor in a desired state, but also control in a desired state for Cisco UCS?

How It Works

The answer is yes. By tying the Cisco UCS blade/chassis system into VMTurbo, the brain behind the control platform can now drive decisions using the same common algorithm as it uses for the hypervisor level. VMTurbo can not only see and understand new metrics from the I/O Module, Fabric Interconnect, Domain and Chassis system, but much more importantly can drive new decisions within these layers to prevent performance degradation. To use an example, VMTurbo can now make intelligent decisions on how to prevent network congestion on the southbound server ports by understanding the relationship between the UCS Fabric Interconnect and the IO Module, and making appropriate placement decisions accordingly, especially in a complex multi-chassis datacenter. In another situation, VMTurbo can prevent issues around northbound ports between the Ethernet uplinks in the UCS Fabric Interconnect that connect to LAN switches, assuring that, regardless of network traffic, VMTurbo can maximize the performance of the end-to-end network in the virtual datacenter.

What About Licensing Costs?

In addition to the performance enhancements that VMTurbo provides in regards to the blade/chassis system of UCS, VMTurbo can also help reduce port licensing costs and defer future capital expenditure. As the market naturally determines positioning for VMs across the UCS Fabric Interconnects in real time, it will also be able to determine unnecessary or underutilized ports that can be decommissioned. This capability allows the user to determine exactly how much capacity is necessary in real time, preventing overspending on UCS licensing. As the environment continues to grow and scale out, VMTurbo can also help determine exactly when to add more UCS Fabric Interconnects and where, in addition to the traditional blade provisioning it will also provide. By helping the administrator to understand true demand for resource utilization, VMTurbo’s UCS control module can help reduce and delay future licensing costs as a natural part of the decision engine.

Unifying the Datacenter

In addition to network traffic and licensing costs, one of the major benefits of the UCS control module is the ability to unify the multiple layers of virtualization into a single common data model. By understanding a Virtual Machine’s path to physical, network, and virtual resources beyond the traditional ESX layer, VMTurbo can assign a price to each of these values, giving a VM many more resources to purchase for its basket of goods to assure performance. If any of these new metrics become constrained, VMTurbo can understand exactly where and when to move a VM to get better access to this newly constrained commodity, while also not sacrificing the performance of the VM across any other metric. For example, if a VM on a blade finds that southbound traffic between the UCS Fabric Interconnect and IO Module is becoming limited, and therefore expensive, VMTurbo can place the VM on a blade with less constrained traffic, while also preventing any sort of rippling effect across the rest of the compute, storage or VM layers.

In conclusion, the benefits of the Cisco UCS Fabric Control Module are numerous. Not just for the insight, but, much more importantly, for the control that it can provide for the datacenter, ensuring that VMTurbo can extend its reach to be a true demand management platform for the virtual environment.

Topics: Features

Subscribe Here!

Recent Posts

Posts by Tag

See all