Virtualization on standardized hardware is a key IT trend that began in the late 1990’s with the consolidation and virtualization of storage using SAN and NAS technology. Costs plummeted, customers had greater control and deploying/re-provisioning became a more seamless and agile process. Then in the early 2000’s, compute became virtualized on x86-based platforms, bringing those same benefits to the processing front. Today, the final step of virtualization – network virtualization – is in vogue but as this technology comes into prime time, there may be differences in how it is deployed and how quickly customers move to it. The intersection of network virtualization with changing server form factors, most notably in the largest cloud customers, may present some interesting challenges. Does the answer to network complexity live in network virtualization or rethinking the server chassis? I address that in the following, but if you are looking for a deeper dive, you can find a detailed paper here.
As the market looks to SDN as a solution to the networks’ inability to keep pace with today’s business needs, there are other forces at play within the server space that will have an impact on the potential viability for SDN. We are recommending the following:
- Customers with standard data centers (ie rackmount servers from Dell,Hewlett-Packard, International Business Machines Corp.) are the best equipped to take on SDN with the least amount of management complexity (beyond the fact that you are now managing two separate levels of networking, the physical and virtual.)
- Customers looking to deploy blades (ie Cisco Systems, Inc., Dell, HP, IBM), hyperscale servers (ie Advanced Micro Devices SeaMicro, Calxeda, Dell DCS, HP Moonshot, IBM NeXtScale) or some other aggregated/disaggregated form factors (ie Intel Rack Scale Architecture) should consider the management complexity before tackling SDN. The greater the percentage of these products in the overall data center mix, the more customers need to take them into consideration when making SDN decisions.
- But for those looking to fundamentally re-invent their IT infrastructure with alternative strategies like hyperscale compute platforms, investigating the OpenFlow support of these platforms will be critical before making a move to SDN.
We cannot stress enough that management is a key component of the entire solution and something that should not be quickly dismissed. With each new layer that is added to the data center, the ability to manage across layers becomes more critical.
Again, as we have said before, SDN is a promising technology, but there are still many pieces that are in flux at this moment. Until more of the market is sorted out, investigation of SDN should continue, but deployment may not be the best strategy at this moment unless all of the pieces that are required for your data center are available.
In the mean time we recommend that those investigating SDN factor in any chassis changes and dense form factors to ensure that they have the proper financial metrics and the proper systems management visibility during the project scoping. While this recommendation is an overview, you can find a detailed analysis here.
John Fruehe contributed to this article.