Knowing which server to buy can help save your organizations money, space, and several headaches over the course of the machine’s useful life. In the sections below, we talk about the different types of servers available, and how to decide server hardware configuration based on your unique needs.
Types of Server Configuration
Servers can be found in three main configurations:
- Rack Servers — Rack servers are configured into standardized racks. These racks can be up to 10 feet tall, and can be efficiently packed into relatively small data centers (with appropriate cooling). Rack servers can be easily expanded based on your capacity needs, and can act as stand-alone systems with their own memory, CPU, and power source.
- Blade Servers — Blade configurations are circuit boards that perform as individual (or clustered) servers within their blade enclosures. Again, these are great for packing a lot of processing equipment in a small amount of space, and can host applications, hypervisors, and web services.
- Tower Servers — Towers live in a chassis, and are great for small businesses due to the minimal components within their configuration. They look like desktop computer towers, and can be customized to the specific purpose for which you need them.
10-Step Guide to Choosing a Server
When I get calls about buying servers, I start off by gaining an understanding of the customer’s infrastructure and what OEM’s they standardize on. We want to get as much information as we can to make sure we can provide the customer with the most cost-effective solution.
- Has your company standardized on OEM?
- What type of applications will you be running and how many?
- What type of processors do you need, and how many?
- Will this server be used for virtualization?
- What type of switch ports do you need to support the server, and how many?
- Is CPU speed or core count more important? Or both?
- About manageability—do you need remote access?
- What kind of connectivity will you need: 1GbE, 10GbE, 10Gb SFP+ or FC?
- How much memory do you need?
- How much power do you really need?
The server you select depends entirely on the applications you will be running, and these applications tend to be very robust. If you don’t have the correct hardware to support the application, you can experience latency, errors, and even worse: downtime.
Configuration Factors
Always refer to each application’s OEM for the hardware requirements that you may need. For companies with many applications (i.e., file, email, CRM and database), you will need a stronger tower, rack, or blade server.
Processor Factors
When it comes to processors, consider the application you are using. In most cases, the application will require a particular type of processor, core, and speed. The last thing you want is to buy a processor that is too powerful—you don’t want to spend money on something you don’t need. Worse yet, you don’t want to get a processor that doesn’t provide a balanced server because it will give you a bottleneck instead.
Yes, the latest x86 processors up performance levels substantially, but do you need all that horsepower to fuel core business applications? Most enterprises still sport relatively low overall server and CPU utilization across their infrastructures. One study of large enterprises (with more than 100 physical servers) showed an average x86 server CPU utilization of between 8% and 20%.1
Depending on the task, you may be paying for performance you don’t really need. So, carefully assess what type of processor, core, and speed are needed to keep applications running in optimal order. Processors can represent 30% or more of the total cost of server hardware.2
Avoiding the highest-performing processor model in favor of a lower-premium choice can deliver hardware savings of about one-third while only impacting peak performance by 10%. This is the kind of informed tradeoff that most end-users can accept.
The bottom line: High-end processors are pricey, and there’s no valid reason to pay for capacity you won’t use or really need.
Switch and Port Availability
Additionally, customers will sometimes order too many servers only to find that they don’t have enough ports on their switch to plug it into. Remember: do an empty port count in your existing network to see if you need to buy another switch to support your new deployment of servers.
Memory Factors
As far as memory goes, more is better. It will cost initially, but that’s better than taking down your network to install and max out the memory on your server when you could have gotten it done on the initial deployment! It’s best to max out your server with memory to avoid the hassle of taking down your network.
A different approach is recommended when evaluating memory capacity in servers. As a rule, when it comes to memory, more is better. The initial cost may be higher, but it avoids having to take down the network later to max out memory.
That said, prices of dual in-line memory modules (DIMMs) can go topsy-turvy, and purchasing can be further exasperated by a seemingly endless menu of models. Be aware that each DIMM may be priced a little differently, so do your homework.
Price per GB among models may even vary wildly for the same memory capacity. Time spent doing all the calculations upfront could prove eye-opening and put dollars back in your budget. For example, last year, analysts reported finding 30% savings when cost-comparison shopping for 8x 16GB DIMMS vs. 16x 8GB DIMMs – with no ill impact on net memory capacity or performance.3
Such volatility means the best deals can change quickly and do so month by month. Traditionally, server memory prices drop over time, but sudden and unexpected shortages can cause rapid price increases. Consider a best practices approach that includes calculating DIMMs’ current lowest cost option for each purchase. Never blindly place an order based on previous procurement.
Power Factors
Finally, calculate how much power you will need for your server so you can get the adequate power supply. Power consumption—or conservation rather—hot topic in any environment. Don’t ignore the fact that your server may draw a lot of power depending on the components you install. Review the power draw per component, and pick based on the needs of your environment.
5 Questions to Ask Yourself When Selecting Your Next Server
Often times we don’t efficiently support the servers that we are running. The processor is not robust enough, we don’t install enough memory, or sometimes we don’t consider port count. In order to ensure the best support for the servers in your network, check out these 5 questions everyone should ask themselves when choosing a server.
1. What type of applications will you be running, and how many?
The server you select depends entirely on the applications you will be running. Applications tend to be very robust. If you don’t have the correct hardware to support the application, you can experience latency, errors, and even worse: downtime. Always refer to each application’s OEM for the hardware requirements that you may need. For companies with many applications (i.e., file, email, CRM and database), you will need a stronger tower, rack, or blade server.
2. What type of processors do you need, and how many?
When reviewing the type of processors for the server that you want to use, consider the application you are using. In most cases, the application will require a particular type of processor, core, and speed. The last thing you want is to buy a processor that is too powerful—you don’t want to spend money on something you don’t need. Worse yet, you don’t want to get a processor that doesn’t provide a balanced server because it will give you a bottleneck instead.
3. What type of switch ports do you need to support the server, and how many?
Customers will sometimes order too many servers only to find that they don’t have enough ports on their switch to plug it into. Remember to do an empty port count in your existing network to see if you need to buy another switch to support your new deployment of servers. Verify that you have enough ports on your switch to meet and support the amount of servers you plan on deploying.
4. How much memory do you need?
Let’s face it… when it comes to memory, more is better. It will cost initially, but who wants to take down their network to install and max out the memory on their server when you could have gotten it done on the initial deployment? It’s best to max out your server with memory to avoid the hassle of taking down your network. Nobody wants to de-rack and re-rack servers on their weekend. NOBODY.
5. How much power do you really need?
Finally, calculate how much power you will need for your server so you can get the adequate power supply. Power consumption—or conservation rather—hot topic in any environment. Don’t ignore the fact that your server may draw a lot of power depending on the components you install. Review the power draw per component, and pick based on the needs of your environment.
Learn How to Choose the Right Server with Professional Guidance
There’s a handful of items that configure a machine. We build everything configured to order. We have CPU, Memory, RAM, hard drive space, and connectivity.
What additional components are there? Well, we build everything specific to the customer’s need. So that’s why we really try to dig in. Sometimes that applications are requiring a large memory count, or sometimes their whole need for a machine is to actually have disc drive space. So we really drill down as best we can.
Contact us for a quote on your business server today!
1 According to Gartner, “However, even with the widespread adoption of virtualization over the last five years, most enterprises still report relatively low overall server and CPU utilization across their infrastructure. One Gartner study of large enterprises showed an average x86 server CPU utilization between 8% and 20%.” Gartner, 10 Ways to Reduce the Cost of Acquiring Servers, Daniel Bowers, 27 October 2017.
2 Gartner says, “On an x86 rack server with two processors, the processors can represent 30% or more of the total cost of the server hardware. In the example shown in Figure 2, avoiding the highest-performing Intel Xeon Platinum 8180 processor in favor of the lower-premium 8168 processor yields a server with a 35% reduction in hardware list price, but only a 10% reduction in peak performance.” Ibid.
3 Gartner reported, “Lower-cost ways to achieve a desired memory amount can be discovered by analyzing the multiple combinations of DIMM counts and sizes available. In one example, on an x86 server model in mid-2017, choosing eight 16GB DIMMs instead of 16 8GB DIMMs resulted in a 30% lower cost for memory, with no difference in net memory capacity or memory performance.” Ibid.
4 According to Gartner: “By revisiting the choice of SSD types, including considering new SSDs built from cheaper or more reliable flash technologies introduced since the initial configuration selection, buyers can save 40% or more on SSDs.” Ibid.
GARTNER is a registered trademark and service mark of Gartner, Inc. and/or its affiliates in the U.S. and internationally, and is used herein with permission. All rights reserved. Gartner does not endorse any vendor, product or service depicted in its research publications, and does not advise technology users to select only those vendors with the highest ratings or other designation. Gartner research publications consist of the opinions of Gartner’s research organization and should not be construed as statements of fact. Gartner disclaims all warranties, expressed or implied, with respect to this research, including any warranties of merchantability or fitness or a particular purpose.
