Lesson 3: Not all components are made equal

When we first speak to a new customer, often we will be given a list of server components that they have decided they want or have with another infrastructure provider, which we will be asked to match. 

The problem with that is that not all components are created equally. 

I understand why many companies want to take that buying approach; to receive like-for-like comparisons on which to make a balanced decision. Unfortunately, it doesn’t work that way and companies that go down that route will end up making a choice based on an inaccurate view of what’s available to them.

Here’s why. 

Quality and performance vary greatly

Let’s start with the basics. The components that introductory level buyers tend to look at are CPU, RAM, storage and network. When it comes to CPU, if you are comparing the same CPU from Intel or AMD to the same CPU elsewhere there should be very little margin of difference when it comes to performance. The main difference between CPUs comes when infrastructure suppliers offer desktop/consumer grade CPUs in a data center (DC) environment. These chips are not designed for the same use cases, and I would urge anyone considering using a consumer grade CPU in a DC environment to tread cautiously.

As for RAM, the DC specifications will be tied to the choice of CPU. The biggest difference being error correcting (ECC) DIMMs (dual in-line memory module) designed for enterprise/DC use and non-ECC DIMMs used typically in consumer grade hardware.

The quality of storage and networking components amongst providers varies hugely.

For example, there are cheap SSDs and expensive SSDs. As expected, the performance of cheap SSDs is significantly poorer than the more costly option. But how often do you think infrastructure hosting providers give detailed specification information about their SSDs to customers? And how often do you think companies ask for them when buying? 

Almost never. So, when a company selects a server package from the website of a budget infrastructure provider, which includes say 2 x 960Gb SSDs and asks for the same spec from a premium provider, they won’t be able to match the price point.

Likewise, from a networking perspective, budget suppliers will often make cuts in areas that customers typically don’t look at or ask about. You will often see server offers with port speeds of 10Gbit/s or more, which in theory should allow each server in that rack to use almost 100% of that available port speed at the same time.

What is often overlooked is the connection capacity between the top of rack switch and the distribution switch, and in most instances, this is the limiting factor when it comes to throughput. I’ve had this happen to servers that I purchased from a supplier, where they had 30 servers in one rack all ‘capable’ of 10G, but the top of rack and distribution switch were connected by only 20G worth of capacity. It caused packet loss and disruption to our services until the supplier came clean and upgraded the links.

Now, I’m not saying that there isn’t a place in the industry for budget-style hosting. What I would urge you to consider is - what are your needs? Your application may not need high performance storage and you might be handling reliability and redundancy in another way, in which case budget hosting may be a good option for you. But if you do need these things from your supplier, looking under the hood a little more will likely save you from disappointment down the line.

Going beyond the basic components 

As you get more advanced as a buyer, you will also start to expand your specification beyond these components to include the chassis, considering the server model and who manufactures it, as well as which motherboard is used. 

This level of detail is important as it can give you an idea of the downtime that you are likely to experience.

The same rule applies to the data centers in which those chassis are housed. While you could argue it’s not a component, when buying it’s important to look at the quality of the data center that the infrastructure provider uses. There is a big difference between a Tier 1 data center - a rack in a basement with a fan and a padlock on the door - and a Tier 4 data center which is secure, stable and redundant enough to host highly sensitive governmental and military data. 

So, I urge any company or buyer, next time you approach infrastructure providers to get comparable quotes, try to avoid giving them a list of the components you want. Instead, talk to them about what you want to achieve as a business and how you need your infrastructure to support you in achieving those goals. Infrastructure providers can then put together a package of components that meet your specific needs. And expect that those packages may differ between providers at different price points in the market. The one to choose is the one that has listened carefully to your goals and answered them in a way that you feel most confident and secure in.

This is my experience. Do you agree with my perspective? How have you approached specifying components when buying? I would love to hear your feedback in the comments or message me directly.