Avatar

Today Cisco is introducing an expanded architectural portfolio and partner ecosystem in support of our successful desktop virtualization solution built on Cisco Unified Computing System (UCS).  Cisco UCS market traction has been phenomenal over the last 3 years. In fact, desktop virtualization has been one of the top workloads deployed on UCS as IT organizations apply the benefits of our stateless, simplified operations model, expansive I/O, and scalable performance to desktop workloads in the data center. Combined with unique product integration and the software eco-system partners such as VMware, Citrix and Microsoft, Cisco has delivered a number of reference designs with our strategic storage partners such as EMC and NetApp. Typically, these architectures were based on designs that easily scale from supporting a few hundred virtual desktops to thousands of desktops.

We have seen an inflection point with the perfect storm of the evolution of storage options, desktop software maturity, and data center architectures. One of the important changes in the storage market is the emergence of flash storage to address performance problems.

Taking advantage of enhanced UCS features and expanding the eco-system of storage partners including Atlantis Computing, Fusion-io, LSI, Nexenta, Nimble Storage and Tegile, Cisco is defining a broader portfolio of data center architectures for delivering desktop virtualization solutions – on-board architecture, simplified architecture and scalable architecture. “Converged” or “Unified” infrastructure stacks such as FlexPod and vBlock have, and will continue to be another successful option for desktop delivery infrastructure.  Let me walk you through each of these architectural approaches.

On-Board Architecture – This is one of the fastest evolving architectures with performance storage moving into the server. In this architecture, the server not only provides the compute, but also the flash storage footprint to service the performance-intensive virtual desktop elements, such as the desktop operating image. This comes about because we recognize that there are two distinct storage modalities, namely storage that serves data demanding high I/O capacity with low latency (performance intensive) and storage that houses less intensive data sets that can be adequately addressed on backend shared storage (for example user persona and data).  This “disaggregation” of monolithic shared storage into distinct and separate data domains, with high-performance data sets being addressed on-server has advantages.

  • Server = Compute + Flash Storage, providing a simplified, integrated solution
  • On-board storage is part of the UCS blade configuration, and therefore benefits from simplified provisioning and management via UCS Manager
  • This approach reduces management complexity, deployed directly on Cisco UCS blade servers
  • It’s ideal for floating or non-persistent virtual desktops
  • Delivers linear virtual desktop storage scalability with consistent, predictable performance
  • Reduces the dependency on  traditional networked storage
  • Improves TCO with reduced costs for storage, power, cooling, rack space, and operations
  • Reduces the marginal cost of virtual desktops

Simplified Architecture – This architecture is often used for virtual desktop “Appliance” or all-in-one POD configurations. The network attached storage is connected directly to the UCS Fabric Interconnect. In this architecture, the UCS Fabric Interconnect behaves as not only the server aggregator, but also as the storage/LAN networking fabric. UCS Manager provides network and server management including storage zoning and networking constructs.

  • “Appliance” or POD model
  • Collapsed network architecture with shared storage connected directly to UCS FI
  • Reduces complexity with single access switching layer
  • Lower initial cost, ideal for environments lacking shared storage and growing into SAN environments
  • Applicable for floating and persistent desktops
  • Scalability limited to single UCS domain
  • Can support FC, FCoE, NFS, and/or iSCSI storage options

Scalable Architecture – This architecture provides a scale-out architecture with the building block that can grow linearly with shared storage. This architecture is consistent for deploying a few hundred desktops to scaling to 1000’s desktops. This solution is based on Nexus access layer for scale-out and aggregate domains and UCS FI provides the demarcation point for the “governance domain” between the server and network environment.

  • One Architecture” that scales from few 100 to 10’s of 1000s of virtual desktop environments
  • Resilient, scalable switched/shared fabric-based approach using shared storage
  • Ideal for customers with existing SAN investment
  • Applicable for all desktop delivery models – floating and persistent models
  • Can support FC/FCoE/NFS and/or iSCSI storage options
  • UCS Manager and UCS Central provides scalable management

Special Webinar: Architectures Explained!

I cover each of these architectures in detail, on a recent BrightTalk webinar I did, “Eliminating the Guesswork With Optimized Architectures for VDI”.  Check it out if you want to get a deeper understanding of the benefits and use cases for each of these architectures being announced.

In conclusion, with this portfolio mindset of eliminating the one-size-fits-all approach and tailoring our solution architectures to a wider spectrum, we believe we’re offering the right architectures for the appropriate use cases and IT environments of all sizes.

You can learn more about this new portfolio of architectures by checking out the assets below.

Please also check out the Cisco webinar “Customer Insights: Desktop Virtualization On Your Terms

Our featured guests include:

  • Jim McHugh, VP UCS Marketing – Host/Moderator
  • Mark Balch, Director UCS Solutions Product Management, Cisco
  • David Johnson, Principal Analyst, Forrester Research
  • Charles Rosse, Baptist Memorial HealthCare
  • Udaya Kiran, WiPro Technologies
  • Robert Dixon, University of Colorado, Boulder