We all are facing economic pressures to do more with less, and sometimes we run into walls coming up with ideas. In this month's earlier post, I raised the question to TechRepublic members about putting switches inside of server racks. With plenty of user comments, it became clear that the answer just depends. However, there was one common theme that there is concern about unused ports in each rack on a pricey switch. One thing I observed at my prior position as a consultant in supply chain software for one client was an incredibly clever yet simple solution to this issue. This client selected to install 48 port managed switches in every other rack -– instead of each rack. This strategy reduced the likelihood of unused ports, reduced the number of switches, yet allowed for some creative redundancy configurations. A server could get two copper interfaces, from each adjacent switch. Figure A below shows how this could be used in practice: Figure A
Click to enlarge.
The particular client was very particular on cable labeling and invested a small amount of in-rack cable management so that each rack-crossing cable followed the same path. You can imagine how this could quickly get out of hand, but good policy and cable management can make it quite intuitive.
One silver lining benefit of the every-other-rack switch is that there are fewer runs back to the main network switch. Of course the practice could be extended to an every third rack or possibly further. The main factor is the port count requirement that is typical per rack. Given that environments vary widely today and that various levels of virtualization may change it greatly, there is no clear guidance. One limitation is that it almost assumes that the only networking media requirement would be satisfied by the single switch in every other rack. This may not be a viable option to environments with fiber channel SANs using LC fiber or other fiber media. Again, we could put the fiber interface panel in every other rack or only in the regions where necessary.
I am personally a fan of in-rack switching in lieu of central patching or home-run cabling. Share your comments below related to keeping costs down while using in-rack switching.
Rick Vanover is a software strategy specialist for Veeam Software, based in Columbus, Ohio. Rick has years of IT experience and focuses on virtualization, Windows-based server administration, and system hardware.