Based on the design of Data Center infrastructure and the way redundancy is applied at every component, Data Centers are categorized in four tiers. How data centers are categorized? To know more about Data Center redundancy you may like to visit "Why Data Centers never go down". So, we have Tier-1, Tier-2, Tier-3 and Tier-4 Data Centers.
 |
First let us understand how data center up time is calculated?
If redundancy in data center is designed in such a way that all services are available for 100% time for the whole year, then the up time is said to be 100%. If in a data center service goes down for 1 hour in a year, then its up time can be calculated as:
Total up time expected in a year = 24 (hrs) X 365 (days) as Data Centers are designed and expected to run through out the year 24 hrs a day.
So total expected up time = 8760 hrs in a year.
In this example if it goes down for 1hr means total up time for this DC = 8760 - 1 = 8759 Hrs.
Hence %up time in a year = (Actual Up time/8760) X 100
= (8759/8760) X 100 = 99.988%
Hence as the table shows tier-1 DC has lowest up time time and as we go up, the up time percent keeps increasing and tier-4 has highest up time and lowest downtime in a year.
| Parameters | Tier-1 | Tier-2 | Tier-3 | Tier-4 |
|---|---|---|---|---|
| Guaranteed Uptime | 99.671% | 99.741% | 99.982% | 99.995% |
| Expected Downtime per year | < 28.8 hrs | < 22 hrs | < 1.6 hrs | < 26.3 mins |
| Component level redundancy | None | Partial power and cooling redundancy | N+1 | 2N or 2N+1 |
| Concurrently maintainable | No | No | Partially | Yes |
| Cost | Lowest | Low | High | Very High |
| Business Type | Small companies and start-ups with basic requirements | Small and Medium Business | Large and expanding businesses | Large enterprises like Google, Facebook etc and government bodies |
This tier level is defined by Uptime Institute which is explained below.
Tier-1 Data Center: A tier-1 DC doesn't have redundancy hence services will get impacted in case of any planned or unplanned disruption. At every level of infrastructure there will be a SPOF (Single Point of Failure) component like one UPS, one DG. Hence for any maintenance work or failure in such components will bring down the data center services.
Tier-2 Data Center: This type of Data Center will have redundant components like if we need (N) one UPS, we will deploy two UPS i.e. N+1. So, in case if one UPS fails or powered off for maintenance work, the other UPS will keep supplying power to computer equipment. But they have single distribution or maintenance of distribution path will bring down the IT infrastructure.
Tier-3 Data Center: Tier-3 DC will have redundant component and will have two distribution path, out of which only one will be active. for e.g. we can have two path for chiller water supply for cooling out of which one will be active and other at standby. In case of any planned maintenance on active path, the standby path can be manually activated and maintenance work can be done. But for any disruption created on active path due to human error will lead to outage for some time till standby path can be activated.
Tier-4 Data Center: Tier-4 DC has redundancy in power, cooling and network paths and all dual paths are always in active-active mode. Even if any accidental disruption happens on one path, will not lead to any service outage as other path is also active and supporting the infrastructure. All components in each path will have N+1 redundancy, e.g. if we need 250KVA supply in one source, we shall use two UPS of 250KVA i.e. need+1. Similar setup is required for second power source. All devices in data center shall have dual power input capability and any device having single power input socket shall be powered through ATS (Automatic Transfer Switch) or STS (Static Transfer Switch) to enable power redundancy. For ATS deployment you may like to read best practice for ATS deployment.
So, these are the basic definitions to understand how data centers are categorized. But when we go for Tier-3 or Tier-4 certification from certifying authorities like TIA-942 certification, whole design is audited at micro level under four categories i.e.
2- Power
3- Cooling &
4- Telecommunication
For example if we have two redundant path of ISP/telecommunication lines coming inside the campus through man holes, they should be 20 meters apart. They should terminate in two different "Meet Me" rooms commonly known as "Mux" or "ISP" rooms. These two meet me rooms shall also be 20 meters apart in building and shall be on two opposite side of the Data Center room. If your data center design adhere to all the criteria but if this criteria is not fulfilled, you cannot get tier-3 certification.
While audits are done at micro level under above four categories, if all four categories fulfill the requirement of Tier-3 or Tier-4 data center, they only you will get Tier-3 or 4 certification. If you fulfill only tier-2 requirements in any one category and in rest you may fulfill tier-4 requirements, still your data center will be certified as tier-2 DC only. That means the final tier certification is based on the weakest component in your data center design.
For more detail you may refer TIA-942 guideline document.
0 Comments