As big data will hit SA, and legislation is pending on how companies must look after personal information, Panduit provides insight on 10 unexpected data centre disasters that directly affect the business and need to be addressed.
The Protection of Personal Information Act has only to be translated into Afrikaans and signed by president Jacob Zuma before coming into effect. It governs how companies deal with personal information, such as names, addresses and identity numbers.
Andrew Oldfield, emerging markets director for Panduit in SA, has provided the following information.
1. Disaster recovery (DR): without continual assessment and regular scenario testing, DR sites often cannot cope with a disaster. This can be due to the rapid changes in today's data centres (DCs) not being reflected in the DR site. A decision often has to be made as to what systems are to be supported and which are non-critical; this needs continual review to ensure it matches the current business need.
2. Backup generators: most facilities managers know about the need to regularly test backup generators, otherwise they will not start when needed the most. Often overlooked are simple things such as whether there is enough fuel for the period they are expected to supply power. How long has the fuel been sitting in the tanks and what condition is it in? What are the plans to bring in more fuel if there is a real disaster?
3. Hunt the server: one of the benefits of virtualisation is that workloads dynamically swap between servers, especially in the case of failure. Without an accurate asset tracking system it is often difficult to find the physical server to swap out.
4. Physical to logical design: when building a DC many companies look closely at the logical design from the network, compute and storage vendors without considering the optimal way of implementing the physical infrastructure, such as cabling, patching, power, racks, etc. Correct design of all elements will allow for the most effective operation and growth of the DC without costly retrofits to live systems.
5. Cooling: many older DCs adopt the "brute force" approach to cooling by just cramming as many chillers into the room as possible to cope with inefficiencies in the design. When building a new DC, it is often possible to raise the temperature a couple of degrees due to the correct design and directed airflow. Many of these principles can still be implemented in existing sites to ensure that, for example, hot exhaust flows are directed away from inlets and that cool air can circulate past active equipment.
6. Cable management: this is important both outside and inside the rack. Cable routing within the DC needs to conform to separation from power cables, meet the requirements of the bend radius specification, and be identifiable and accessible for maintenance/reconfiguration purposes. This is often not the case and can lead to failures that are difficult to locate and costly delays in reconfigured connectivity to meet new needs. Within the rack, lack of cable management increases the risk of incorrect patching, restricting the airflow, causing overheating of active equipment and also blocking access to maintainable parts (fans, PSU, etc) on the equipment.
7. Grounding and bonding: in countries such as SA where lightning strikes are common, it is important that all equipment is grounded to avoid costly damage. Many DCs are built or have expanded without the proper grounding in place. If implemented correctly in the first place, it is easy to expand and add additional racks to the DC that are full protected.
8. Standards: many customers are unaware there are standards for DC design, as well as best practices derived from those standards and practical experience by industry association members. When contracting with a systems integrator or consultant, ask what design standards they will be implementing; key are TIA 942 and associated standards as well as industry bodies Ashrae and UpTime Institute.
9. Security: the DC is the core of an enterprise's business in that it contains much if not all of their business data. Failure to assess, properly implement and regularly review the security policy can leave that data open to exploit. A proper security policy will guard against both malicious and accidental security breaches; it should consider both the technological and human aspects of how to be secure. Once security is breached, the biggest cost is often to the reputation of the enterprise rather than and financial loss and the cost to rectify.
10. Record-keeping: DCs are a dynamic environment, and without proper asset control, as well as infrastructure management, much of the estate will quickly unwittingly become unused and thus become a hidden cost. By implementing proper change control procedures, the chance of incorrect, as well as non-recorded changes, will be lessened, thereby ensuring the routine maintenance and fault-finding is not compromised.
Share