Wondering what it takes to architect a modern data center? Whether you’re just curious or already in the active planning phases of building out a data center, you may already understand that choosing the right data center infrastructure can be a complex process.
You’ll need a wide range of equipment, from storage options to network and security hardware and cooling solutions. How do you know, without months or years of extensive trial and error, which options to choose and how to get it all set up correctly?
At ServerLIFT, we work with data centers across the globe on a daily basis. We respond to their needs, hear their complaints, celebrate with our clients as they come up with innovative solutions, and help them out when it’s time to move or install new hardware. We’ve put together this guide as a starting point for planning the architecture needed for a modern data center, so you can get an idea of what the most successful data centers do when it’s time to expand or open up a new location.
The following categories cover, in broad strokes, the equipment you will need to get up and running. Yes, it’s the basics, but that’s always a good place to start.
Building and Physical Security
As you choose your building space, factor in the weather patterns and possible security issues your data center may face in that location. Setting up a data center in a flood plain, for example, should be avoided because, at best, your service would be disrupted whenever there was a heavy storm, and, at worst, you could lose most or all of your expensive, hard-to-replace equipment if the building isn’t flood-proof. Of course, this can’t always be avoided since foul weather or weather disaster zones need essential IT services too.
Earthquakes, tornadoes, fires, and hurricanes are just some of the other considerations one might face. In all these cases, planning the building and infrastructure in ways that can resist or withstand the local, specific weather conditions is a critical starting point. Consulting with local building experts is highly recommended. When possible, find or modify a structure, as best as you can, to withstand these elements without power disruptions, interruptions in service, or equipment damage.
You’ll also need to consider security as you choose a location or building. Is it in a safe part of town? If you’re choosing a building, determine if it has existing security measures, such as a camera network, biometrics, and lockable, secure interior rooms. If not, you will need to add these features yourself.
A lack of weatherproofing and security in an existing building doesn’t necessarily have to be a dealbreaker, as long as you’re prepared to bear the effort and expense of having them installed.
One of the most important considerations is the actual space that your building will provide for a variety of functions.
- How big is the area that will house your servers?
- Will you have room to expand in the future?
- Do you have a place to store highly sensitive systems that require more restrictive security than equipment in the rest of the building?
- How will different personnel access security-appropriate areas? Facilities maintenance people and other vendors may not need or shouldn’t have access to the rooms that actually house the racks and server equipment, but they need to regularly be in other areas of the building. Does the building allow for easy segregation of space based on security access levels?
- Is there a locked area at each entrance where visitors must stop and provide credentials in order to be cleared to enter the data center?
- Does the building have adequate storage areas so you can keep redundant systems on hand for use at a moment’s notice?
- What is the overall footprint of the property? Even if the building space is just what you need for today, having a lot of land on which to expand is a big bonus. Many modern data centers increase capacity with outdoor containerized modules. These can be laid out and even stacked for a dense, affordable way to expand without pouring foundations and building new walls.
Though it may get frustrating shopping around for the perfect real estate to house your data center (or modifying your existing building to suit your changing needs), this effort will be worth it in the long run.
Operations and Management
Modern data center infrastructure consists of so many complex, interrelated parts and systems, that it is imperative to monitor each system closely so malfunctions in one don’t affect the others.
Data center operators need constant, real-time updates on the health and security of their data centers. This means setting up a management system that relays information such as the condition of HVAC, voltage, power levels, backup power systems, security threats, RFID hardware tracking, and hardware performance back to an operations center that is manned 24/7 by attentive employees.
The servers you choose will depend on several factors, including your own individual needs. Shop extensively before you buy, considering how much space you will have in your data center, the supplemental equipment each server will need to be compatible with, and how you will install and uninstall those servers when the time comes.
Once you know what servers will best suit your needs, you can shop for racks or cabinets to house them. Price, space, configurability, and compatibility with planned server equipment and mounting systems will be your most important considerations here. Though it may sound tedious, it’s helpful to arrange your equipment in the space where you want to house it and create a diagram of where everything will need to fit to leave room for air flow and cables, before you install the equipment in the racks.
As you know, data center network architecture consists of switches, routers, network cables, and any physical network security devices, such as data diodes, that you plan to install. Your network will, of course, need additional security in the form of software once you boot it up, but before you can do any of that, you must set up and properly configure the physical network hardware that will be taking up space in your data center.
Think about placing critical networking components in the most restricted, limited-access parts of your building to reduce the risk of anyone being able to deliberately tamper with your configuration or accidentally mess something up.
Connectivity and bandwidth coming into the building must be sufficient for current operations and for long term growth and expansion. Even if the availability today has more than what you’ll need, check with the local service provider to find out what they’re planning in the medium to long term. We have a colocation client who located a very small portion of their space in a large metropolitan area in one diminutive downtown building. Why? Because that building sits directly on top of that city’s termination point and junction for the Trans-Pacific Express telecommunications cable. The building literally has the fastest bandwidth for hundreds of miles—a major draw for certain customers.
The name of the game here is uptime. In pursuing the unobtainable, theoretical goal of 100% uptime, power redundancy is key. In fact, in a highly functioning data center, there is almost no such thing as too much redundancy.
Decide on your power systems, including PDUs, and then have a backup for those power systems. Then, find other diverse power solutions that you could get up and running quickly if an outage, surge, or another issue takes down your main power supplies.
Speaking of surges and outages, you will also want at least one UPS solution to prevent equipment damage and data loss when power drops or spikes unexpectedly. Consider redundancy here, too; a backup UPS system is never a bad thing. And this cannot be overstated: ensure that those backup systems are regularly maintained and tested.
Your server and rack setup, your desired rack power density, your budget, and the space you have to work with in your building will all influence your choice of cooling systems.
If you are able to implement it, direct to chip liquid cooling is your best cooling option. The most effective direct to chip (also known as direct-on-chip) liquid cooling is a waterless, two-phase method that moves heat off chips and away from servers.
Heat damages equipment. The better your cooling system, the longer your servers will last. Since power and computing demands increase every year, consider advanced cooling at least one of the data center infrastructure solutions you can’t afford to skimp on.
Part of achieving data center modernization is developing an efficient, organized, and damage-free way to keep cables under control. But cable management is often a complex operation in itself. A truly efficient cable system is more than just keeping your cables from getting tangled. It should organize your cables so you know which goes where, and provide an easy system for upgrading cables and equipment or moving them around as needed.
Take a look at large, vertical cable managers. These often provide the organization you need while leaving plenty of room for airflow, allowing you to rearrange as needed, and supporting both copper and fiber cables without damage.
As you make your list of equipment, think about how you’ll install and rearrange everything when the time comes.
One crucial part of data center infrastructure design is planning how you will install, uninstall, rearrange, and move all of the equipment you’ve just picked out.
Server racks and cabinets, as well as the servers and other hardware that goes in them, are not exactly light and portable. Ideally, you’ll pick the perfect spot, set up your equipment, and rarely if ever have to move it around again.
In practice, however, plans change. You’ll eventually need to upgrade your hardware, improve cooling, implement a better cable management system, or even migrate your entire setup to a new location.
To accomplish this safely and efficiently, with a minimum of stress and risk, you will need safe and efficient tools capable of lifting and transporting heavy, unwieldy equipment. Depending on your unique needs, rent or buy a server-handling device before you attempt any major moves or installations. When you rent or buy from ServerLIFT, you can pick from front-loaders, side-loaders, devices that reach up to the tallest server rack heights and all the way down to the floor, and even specialty attachments that allow you to move extra wide loads with all of the necessary support.
If you need help or aren’t sure which data center lift is right for you, feel free to contact ServerLIFT for support.