The main purpose of a data center is to ensure the highest possible availability of services, preferably 100%. Of course, 100% can be a desideratum, which is why data center providers generally guarantee the operation of services in a percentage close to 100%, for example 99.9%. It is called the Service Level Agreement (SLA) and is a commitment of the service provider to customers regarding the minimum time of service availability, generally calculated monthly.
In order to ensure a high SLA, at the level of a data center, technical and organizational measures are implemented to ensure the redundancy of all elements that contribute to the availability of the services. The elements that ensure the continuous operation of the servers, and which, in case of interruption, will affect the proper functioning of the web services, are: power supply, cooling and internet connectivity.
Of course, we are not talking here about the proper functioning of web services that are available at the operating system level (web server – apache for example, SQL server – MariaDB for example, email server – exim etc), which are software applications that run over a operating system, and which are generally managed by the customer. In this material we will refer strictly to the way in which the data center, in particular CHML data center, implements the redundancy at the level of the physical infrastructure.
Redundancy of power supply is ensured by powering the server room from two different lines (A + B), each line being supplied from a different electrical transformer. We have found in practice that there are many breakdowns of the electricity supplier that affect only one of the two lines. The available power is 200KVA on each of the two lines.
Each of the two power lines is protected by an automated diesel generator. We have a 160KA IVECO generator (line A) and a 60KVA Pramac generator (line B). In case of an electrical outage, the time period of 10 – 12 seconds from the moment of power failure occurs until the generators start is covered by UPSes. We use 8,000VA and 10,000VA Smart-UPS RT 8000/10000 XL APC UPSes. Beyond the high power, a notable advantage of these UPSs is that the batteries are hot swappable, which allows us to replace used batteries without interrupting the power of the servers.
We provide cooling with 10 units (air conditioners) inverter type, Mitsubishi and Fujitsu brands. The cooling capacity of a single unit is between 22,000 and 28,000 BTU, raising the total capacity to over 230,000 BTU. Cooling redundancy is ensured by the large number of units, only half being needed to cool the entire server room. To ensure good operation for a long time, not all units operate simultaneously, usually having 6 units in normal operation, two in standby and two off.
The Internet connection of the data room is a redundant one, both at the level of providers and equipment. We are connected through GTS Telecom and LanSoft Data, with each of these providers having a contracted capacity of 2Gbps, with a total capacity of 4Gbps. We are RIPE members, having our own AS number and 3,840 own IPv4 addresses. These are announced by BGP protocol to both providers.
We use only Cisco network equipment, with dual power supply and two supervisors at the core equipment level. We cannot specify the exact model of the equipment used for security reasons. For security, each client is assigned a dedicated vlan.
Of course, the presentation made here is not extremely detailed, its purpose being to point out how we, the CHML team, have implemented redundancy in the server room. If you want to use our colocation services, we invite you to our headquarters for a detailed presentation and discussions. We are at your disposal at the telephone numbers on our website www.colo.ro.