Why the Network Matters for Virtualization and Cloud Computing

January 5, 2011 - 1 Comment

I’ve written before (here, here, and here) that Cloud Computing is more than some cool software running on a server. Sure, the applications are the sizzle on the steak (+ all the marketing terms – dynamic, elastic, on-demand, etc.), but there’s a little more to it than that. A user needs to access the application, get the information quickly (or sent it information), and feel confident that the information was delivered securely. The application doesn’t always know what type of device will access it (PC, Mac, Browser, Tablet, Smartphone, etc.), so it can’t be 100% sure it’ll deliver the best user-experience.  And users will demands that applications continue to run regardless of the mobile device’s location. All those demands on applications get a lot easier, and in some cases require, an intelligent network providing the infrastructure.

But people often forget those details because they have become so accustomed to a robust network always being there. They might struggle to define the value of that network, just as Kodak did in defining “original technology” in the famous Mad Men episode (Carousel).

Don’t take my word for it, hear what Cisco Cloud CTO Lew Tucker had to say during a recent set of meetings with industry analysts – here, here, here, here and here. So what do you if your applications groups want you to approach the network like this? (see slide 19) If they do, here’s a quick list to remind them how critical the network is in making their Cloud applications more awesome than they already are…

Entry Level – “The Basics”

Needs for Anywhere Connectivity

  • Access to data and application is “just expected”, 24x7x365. Ubiquitous network access – 3G, 4G, WiFi, MiFi, home broadband, corporate network – makes this possible.
  • Lower costs of devices (smartphones, tablets, laptops) drives demand for “my experience” anywhere. “Consumerization” driving new demands on IT.
  • Users want access to their data (or applications) to be separated from physical devices.

Needs for Higher Bandwidth (10Gb, 40Gb, 100Gb)

  • Growth of traffic – video, big data analytics, converged networks (storage traffic + data traffic)
  • Growth of 24×7 traffic – global user access; need for faster backups to SAN via converged network
  • Growth of VM mobility (it’s not unusual to have 1500-3000 vMotions in a month for a medium sized company)
  • Growth of video-enabled applications (collaboration, learning/training, etc.)
  • Growth of “big data” analytics applications that not only shuffle information north and south, but increasingly east and west.
  • Consolidated servers (lots of VMs) require more bandwidth per physical server

Needs for Security

  • The number of devices continues to grow exponentially, with many companies now allowing “Bring your own” device policies. Managing security on all those endpoints is a losing battle for IT. The network provides a broader, more manageable way to address security.
  • With virtualization, applications and virtual machines can now move within and across data centers. Security policies and enforcement need to be able to move with the application to prevent unexpected security holes.
  • As Data Center infrastructures consolidate and host multiple resources (applications, customers, etc.) it is critical to provide appropriate levels of security across all resources.

Needs for Quality of Service (QoS)

  • Consolidated traffic (multiple tiers of business apps) requires differentiated levels of service (SLAs). The network is able to view all traffic and allocated end-to-end resources appropriately.
  • Dynamically moving virtual machines (vMotion, Live Migration) traffic needs dynamic prioritization, but not dedicated bandwidth.
  • Virtual Desktops are a mix of data and real-time traffic (voice, video, collaboration) and a single tunnel might not be the best approach for delivering real-time traffic.
  • Consolidated storage traffic needs guaranteed service-delivery levels (FCoE, iSCSI, NAS)

Now that you have those in your “why the network matters” bucket, let’s move on to some more complex scenarios:

More Advanced – “Working without a Net”

Needs for Multi-Tenancy

Moving to shared infrastructure and shared services models is a two-way street. For IT, it’s about driving efficiency of the assets, consistency of operations and flexibility for new innovative services. But for the users, they may still have concerns about their toys (important data) playing nice in a big sandbox. So it’s important to be able to deliver an infrastructure that can provide end-to-end isolation.

Needs for Mobility

  • As more application get virtualized and the infrastructure (servers, networks, storage) become more consolidated, the ability to load-balance VMs around that infrastructure increases. But limiting that mobility to a single rack can impact next-generation High-Availability (HA), Disaster Avoidance and Disaster Recovery architectures. Cloud networks need to allow workloads to be mobile – across racks or within data centers and between data centers.
  • (See Needs for Security bullet #2) VM mobility is really helpful in the data center especially when it’s linked to automated resources to drive greater efficient use of assets. But efficiency without control can be dangerous. Be sure to give your network and security teams the ability to keep things safe and fixable if problems occur.

Needs for Hybrid Cloud / Mixed Location architectures

For as many definitions of Cloud Computing exist today, there might be nearly as many for “Hybrid Cloud” (some listed here). This isn’t a bad thing, because it means customers will have choices that fit their cost, risk and flexibility needs. It means that your network is going to extend outside of your data center. It may require VPN-like capabilities (often called “Virtual Private Cloud”) between your private resources and public cloud resources.

  • It means that your network is going to extend outside of your data center. It may require VPN-like capabilities (often called “Virtual Private Cloud”) between your private resources and public cloud resources.
  • It means that your network could benefit from Application Acceleration capabilities over those network links, to improve performance and reduce costs. And if those acceleration capabilities can be virtualized into software so they could be put almost anywhere, that’s even better.
  • And if you really want to impress the boss and show them how your consolidated, virtualized, automated cloud computing model can start to reduce the dependencies of data-location to physical-location, show him these examples (here and here).

So fold this up, keep it in your back pocket and bring it out the next time the application teams want to tell you that the network isn’t that important to them. Tell them that you’re going to make them and their applications look like rockstars!!

Please visit Cisco’s Cloud Homepage.

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.


  1. Thanks so much for the good advice.