Data Centre Redundancy
Your data center is the backbone of your organization. Without it, everything grinds to a halt. Given how critical it is, it makes sense that many organizations work with Alpha Systems to support critical servers, minimize risk, and maximize uptime.
Data center redundancy is the name of the game, but given that there are multiple redundancy models and that they are not all the same, it’s challenging to make an informed decision regarding your needs. Let's discuss data center redundancy, redundancy design, and more.
What Is Data Center Redundancy?
Data center redundancy is the practice of leveraging duplicated hardware – power supplies, servers, cooling systems, etc. – necessary to keep things up and running in the event of an outage, hardware failure, or disaster. In most cases, a redundant data center architecture only duplicates the most critical components. And it’s important to remember that a fully redundant design that duplicates all aspects and ensures a seamless transition in the case of an adverse event can be unreasonably expensive.
Thankfully, you don’t have to duplicate every component in your data center to guard against downtime related to outages. However, you do need to understand the various types of data center redundancy levels and design options.
5 Different Types of Data Center Redundancy
Data centers have several types of redundancy, each designed to provide different levels of protection against disruptions. The most common types of redundancy are:
Power Redundancy:
This ensures that multiple power sources are available to the data center. In a power outage, backup power sources, such as generators and batteries, will take over to ensure an uninterrupted power supply.
Cooling Redundancy:
This is often overlooked but just as important because technology needs to operate at certain temperatures. So in case of a cooling system failure, backup cooling systems will take over to maintain the data center’s optimal temperature.
Network Redundancy:
This ensures multiple network paths are available for data transmission. In case of a network failure, traffic is rerouted to alternate paths to prevent data loss or disruptions.
Storage Redundancy:
Multiple copies of data are stored across different storage devices. In case of a storage device failure, data can be recovered from other storage devices to prevent data loss.
Server Redundancy:
This redundancy ensures multiple servers are available to run applications and services. In case of a server failure, another server provides uninterrupted service.
Importance of Data Center Redundancy
When it comes to data center redundancy, it’s important to ensure that specific components are protected first. Power supplies are among the most critical components to duplicate. And, with the increasing frequency of natural disasters, data center power redundancy should be one of your prime concerns. Without uninterruptible power supplies (UPS) it’s impossible to operate your data center, regardless of what other systems you duplicate. In addition to UPS systems, you should also duplicate:
-
Backup generators (in case of power outages)
-
Cooling systems
-
Individual servers (and their associated data)
Keep in mind, you can always use third party hardware maintenance on your primary AND redundant systems, but we understand that OEMs go to great lengths to instill Fear, Uncertainty, and Doubt (FUD) about using a third-party maintainer. However, using third-party support on your redundant data center systems is a no-brainer, as any perceived risks should already be mitigated by your system redundancy.
Another critical factor of data center redundancy design that can protect against regional calamities is geo-redundancy. Designing your data center components and data architecture so that a climate event or natural disaster in one country or continent won’t hault your operations can be essential to the success of your business. Geo-redundancy is a common practice among cloud service providers today, and is generally good practice for international and global companies that maintain their own infrastructure.
Data Center Redundancy Levels
When it comes to data center redundancy, there is no one-size-fits-all solution. To account for that, different data center redundancy levels have been developed. These include:
-
N
-
N+1
-
N+2
-
2N
-
2N+1
What Is N in Data Center Redundancy Design?
Before discussing the actual data center redundancy design levels, let’s define what “N” stands for in these designations. What is N? It’s the minimum capacity needed to power or cool your data center under a full IT load. So, if your data center needs five UPS units for full operation, then N would be 5 in your case. Each data center will have an N rating that correlates to its specific power and performance needs.
It’s also important to realize that N does not account for redundancy. This is your base level of operation. If your data center’s N was five and you lost a UPS unit, you would not be able to operate at full capacity. So, N is the bare minimum needed for operation. Obviously, that’s not a tenable situation, so businesses plan beyond N when it comes to redundancy.
What Is N+1 Redundancy?
Now that you know what N stands for, it’s probably easier to understand what N+1 redundancy might refer to. It’s just increasing the number of components by one. So, if you need five UPS units to operate, an N+1 plan would give you six. That way, if one UPS unit fails, you have an extra that can kick in to pick up the slack. This applies to all types of components, though, including servers, storage hardware, HVAC systems, and generators to name only a few.
But does N+1 really work? It depends. Current design standards argue that for every four components, you should have one redundant component that can be used in the case of failure.
Of course, that still leaves some risk to your organization, which is where the next redundancy level comes in. N+2 gives you your full aggregate of components, plus two spares. However, that may still not be enough to achieve IT resiliency.
What Is 2N Redundancy?
2N redundancy is not just the next step up from N+1 – it’s an exponential increase. It mirrors your original setup, so if you had five UPS units, 2N redundancy would give you five more for a total of 10. In addition to the UPS units, you would also have two separate distribution systems so that if something happened, you could switch over without a problem.
Data Center Redundancy: N+1 vs. 2N
When it comes to redundancy, N+1 offers some coverage. However, because it only provides a single spare, there is still considerable risk that a cascading fault could mean that you cannot operate at full capacity. For a growing number of organizations, 2N is the better choice for avoiding significant data center challenges.
What Is 2N+1 Redundancy?
2N+1 delivers the fully fault-tolerant 2N architecture plus an extra component for an added layer of protection. Not only can this architecture withstand multiple component failures, even in a worst-case scenario when the entire primary system goes down, it can sustain N+1 redundancy.
Data center tiers
The N system is a helpful way to measure redundancy, but in practice, achieving maximum uptime is about more than simply adding components. Uptime Institute created a tier system to "explain the infrastructure required for data center operations."
There are four major tiers.
Tier I data centers
Tier I data centers are the most basic. These facilities have enough redundant infrastructure to run efficiently but need more redundancies. They can withstand disruption from human error but not an unexpected outage and must shut down for maintenance.
Tier II data centers
A Tier II data center includes some cooling and power system redundancy, providing more uptime. Employees can remove components without shutting the data center down, but large failures still take the facility offline.
Tier III data centers
Tier III data centers ensure redundancy for every component in the facility. A failure at any one point does not affect data center operations. Shutdowns are not needed to replace or maintain equipment.
Tier IV data centers
Tier IV data centers represent maximum uptime. These facilities have several independent and isolated backup systems requiring 2N or 2N+ redundancy levels. Downtime is unlikely in these data centers, though maintaining them is costly.
Is there a simpler way to look at tiers?
Yes. Although data center tiers officially reflect these diverse elements, most clients care most about the guaranteed uptime. Uptime is usually the most critical factor in any service level agreement. The other tier criteria indicate how and why the facility will be able to deliver on its uptime promise, with greater redundancy, fault tolerant, compartmentalized, and otherwise high-quality infrastructure.
It is generally accepted that the tiers represent the following uptime guarantees:
-
Tier 1: 99.67%
-
Tier 2: 99.74%
-
Tier 3: 99.98%
-
Tier 4: 99.995%
Although set in terms of uptime, it can be helpful to consider what each tier means in terms of unavailability. For example, a Tier 1 facility is available for all but 28 hours per year. Such facilities or portions thereof may go offline for short periods for scheduled, routine maintenance. Depending on the notice requirements and the business needs, this may have limited impact.
At Tier 3, the client expects only 95 minutes of downtime per year, and Tier 4 sees less than 30 minutes. These are truly high-availability options.
Which tier is right for me?
The tier decision usually comes down to cost-benefit analysis. As the facility’s tier increases, the annual contract costs rise as well, often exponentially.
As a rule of thumb, Tier 1 and 2 data facilities are meant for smaller businesses for which cost is as important a factor as reliability. SMBs are often willing to accept a few hours of downtime here and there, and even the possibility of an unexpected outage, to make ends meet.
Tiers 3 and 4 data centers are the “go to” for larger enterprises and various institution, such as banks, that require continuous availability. The consequences of unexpected downtime for a hospital, to take one example, could be life or death. Large internet retailers lose significant revenues whenever a customer cannot complete a transaction, so they often prioritize Tier 4 availability as well.
Choose your own adventure
There are also opportunities for blended solutions. In deciding where to house IT hardware and associated functions, it is essential to evaluate business needs and the true cost of failure.
Selecting Tier 4 for everything because “it’s the best” could quickly outstrip reasonable profit margins. Even where Tier 3 and 4 facilities are required for certain IT needs, non-mission critical workloads can often be housed in Tier 1 and 2 centers to lower the associated overhead costs.
Different choices will suit different IT organizations. The tier structure, however, helps facilitate universal “apples to apples” comparisons of cost and options across providers, so each IT department can make the best decision for their particular situation.