One of the things that companies often face when deploying new technologies are the costs involved. Many times they have to decide which is more important to them; the cost of acquisition or the total cost of ownership. When considering Cloud, Hyperscale and High Performance Computing technologies, the stakes can get considerably higher.
With such a wide range of choices out there on the market, you have to make a series of choices that will impact the way you operate and how much it will cost you over time. The temptation that many companies have is to go with the cheapest solution they can, but in the end, they find out how much that cheap solution costs them much more. Not long ago I met with the CEO of a company that for the longest time standardized on traditional Dell Blades. Always wanting to ensure that he was getting the most for his money he evaluated what he thought would be an equally good solution for his company. The total savings was probably less than five thousand dollars per chassis, but it was enough for him to give it a try. As we sat in his office he was recalling that decision with less than wonderful memories.
“The vendor I had switched to in order to save a little money ended up costing me four days of downtime…before I even got a phone call back from the folks in tech support. Once I did it took another three weeks to have a replacement system shipped to me. I didn’t even unpack it. I sent it back along with the others, and moved back to Dell. Yes, I might pay a little more, but I know when I call for service and support I’ll have my systems fixed in four hours or less. Even if it is a system that is under the next business day support, I know I won’t be down three weeks”.
In this case the total cost of ownership was measured in loss of compute time, and while the cost of acquisition, the loss of compute time ended up costing more that had they stuck with that they had. He also mentioned that deployment of the hardware from this manufacturer had its own challenges and took quite a bit of time, but I’ll avoid mentioning specifics for now as to give it away.
Then we need to look at power consumption and rack footprint. When it a collocation facility, these are the two things you need to watch out for or you’ll run the risk of watching your operating costs go up and up. Using the Dell c6100 or the c6105 as an example, I can put what is the equivalent of four R610’s (taking 4u of rack space) and condense that down into 2u, and then reduce the power consumption by an average of 20% (using similar configurations as calculated on the Dell ESSA tool). Now replace that R610 with a previous generation server and the numbers just go up when you look at something like a PowerEdge 1950 and a c6100 / c6105. What if you don’t need the power of an R610 or a c6100 / c6105? Perhaps a Microserver would fit you better. What is a Microserver? Think about putting eight or twelve servers in a 3u chassis, while using less power. Check out this video from Drew Schulke of Dell Data Center Solutions as he talks about these Microservers.
When you are able to reduce the power you consume for compute power, and you are able to reduce the rack footprint you are now able to take that money and put it elsewhere, like R&D or other areas, your TCO goes way down. Even if it costs a little more to purchase upfront, it often times pays for itself month after month is worth it many times over. When you get a server chassis in a single box ready to rack and plug in without having to put it all together…or better yet, getting the servers racked and cabled in the factory, then shipped ready to plug in to the wall and the network, think of the man-hours you save in deployment time.
It is always good to save money where you can. But in the end you need to always remember that you get what you pay for, and often times, cheap solutions end up costing you more.