Do you have a question for our experts?
Submit your question directly to our editors at email@example.com
I have heard that top-of-rack switching can be more costly than alternatives; is this true, and why?
Yes, top-of-rack switching costs can be higher and lead to a large oversubscription of ports. With a fixed switch serving one cabinet or two, you may not be able to use all of your ports. The backbone fiber in the number of ports required can be quite expensive over a large data center, but the real cost is in buying ports you don't need and can't use. I say this because your limit to the number of servers you put in a cabinet comes from power and cooling. An average data center gets about 12 to 14 servers in a cabinet. With a 48-port switch, there are a lot of leftover ports, which not only cost day one, but also carry maintenance costs and power costs even in idle mode. The cabling is generally about 10-15% of the cost savings on unused ports.
Read more of Carrie's advice
Will 10GBASE-T switches eliminate copper cabling?
Converged network infrastructure bids cost and communication benefits
Although it is often touted as a cost savings, you usually come out cheaper centralizing your switches in distribution area zones. This lowers the overall power requirements for your switching architecture, which is generally about 5% of your overall data center power bill. I would suggest looking at TIA 942-A (or the ISO 24764 equivalent) to see how the cabling would work. You should also note that when you buy the direct attached cables with the interconnects in the end (also called Direct Attached Copper, or DAC) those cables typically carry a 90-day warranty as opposed to a 20-year warranty that you would get with a structured cabling system. Reducing the number of switches required (which is very easy with 10 Gigabit Ethernet, or 10GBASE-T) also reduces the number of required uplink ports.
This was first published in July 2012