Cisco Blogs

Composable Infrastructure: Cutting Through the Noise

- November 24, 2015 - 0 Comments

Its seems that lately, there is a lot of noise in the industry about composable infrastructure.  Some of others in the industry seem to be busy churning out a bunch of noise around their vision of the future as though just talking about it repeatedly will somehow capture mindshare.   Most of what is out there amounts to a whole lot of talk and hand-waving, but it has been pretty light on real products that customers can buy.  Back in September of 2014 we launched the UCS M-Series and C3000 families of products.  At that time, we not only added the new server products to the UCS portfolio, but also introduced SystemLink Technology.  We set out to lay the foundation for our composable infrastructure vision, a vision that is being noticed by many in the industry.  Check out what Moor Insights & Strategy is advising customers about Composable Infrastructure.  Also, take a few minutes to hear more about the attributes of a composable infrastructure with the video below.




The revolutionary capabilities of SystemLink really lay the groundwork for one of the most important elements required for a truly composable infrastructure.  The Cisco SystemLink ASIC, the first of it’s kind in the x86 Server space, has several key capabilities.

First, it enables server subsystem disaggregation.  This is critical to be able to break down the traditional server, which is traditionally defined primarily by convention and sheet metal, into its basic elemental resources.  By separating the physical components of the server, those resources can then be pooled and programmatically composed into a logical server and then, subsequently decomposed, returning the elements back to the pools allowing for reuse.  This breaking down of the server means that not only can the most efficient and optimal use of resources be made, but also the lifecycle management of those resources is also decoupled.  So, in the case of M-Series, the next CPU generation that would drive a complete replacement of the server with a traditional rack-mounted server would only require the replacement of the CPU and possibly DIMMs to achieve an upgrade.  Subsystems like the local storage, RAID controller, network adapter, power supplies, fans, and cabling are preserved until upgrades of those respective elements would yield benefit to the business.  As you can see illustrated below, this adds up to a significant savings of both capex and opex, while at the same time offering a more agile IT environment.



Another key capability afforded by Cisco SystemLink Technology is the extension of the control plane into the hardware.  After all, simply sawing a woman in half on stage is not a real magic trick unless you put her safely back together at the end.  So too is the case with server disaggregation.  It’s why we chose the term Composable Infrastructure.  The reconstitution of the elements (in a more efficient and flexible form) is where the real benefit comes in.  To accomplish this, without undue complexity or multiple layers of software, you must have a control plane within the server that can arbitrate the connection of the subsystems as prescribed by the policies that are defined by the user / application.

If you would like a more in-depth understanding of how SystemLink works, check out a very informative blog that Steve McQuerry posted back in October of 2014.

Thus, SystemLink’s unique capabilities expose some truly revolutionary capabilities within the underlying infrastructure, but the more evolutionary (read: consistent) part of Cisco’s Composable Infrastructure architecture is the management framework which allows for the policy-based programmatic control of the disaggregated elements.  The UCS Management environment along with a highly extensible set of APIs provides the customer the ability to control this composable infrastructure while at the same time abstracting the added complexity of the larger and more granular pools of resources.




The UCS Management architecture is capable of providing not only the capabilities to orchestrate the Composable Infrastructure, but do so with the common set of interfaces and user experience that customers employing a more traditional model, one that is transitioning between a traditional and a DevOps style, and ultimately the organization that requires ultimate agility.  That means that UCS is uniquely positioned to offer both traditional server elements and composable elements in whatever combination that best meets the needs of the business without siloed tools and processes.  Additionally, consistent UCS Management tools and interfaces means that IT staff need not re-learn new tools to manage an infrastructure even as it evolves from traditional to agile. This allows the IT staff to be treated as the ultimate resource in the infrastructure and be utilized as efficiently as the hardware.

So while the talk about Composable infrastructure continues in the market, at Cisco, we believe that with products like UCS M-Series and the UCS C3260, we are not just defining Composable Infrastructure, but we are walking the walk with shipping products that customers can buy today to begin the journey to the next generation of agility and efficiency.

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.