As companies become more reliant on IT, costly network crashes become an everyday threat. However, firms like COL are here to help BUSINESS CONTINUITY and disaster recovery planning have come to the fore in recent years, as the ever-growing reliance of enterprises on information technology has highlighted the large financial losses that can occur when systems fail or are disrupted. Underlying its importance, regulators in some countries have enshrined business continuity into compliance laws, making it mandatory for companies to maintain a level of preparedness against unforeseen disruptions. For companies operating online financial trading services, a network crash could spell closure, unless they have a contract with a facilities vendor that has networked workstations available for emergency use around-the-clock. Probably Hong Kong's biggest such service is COL, an IT service subsidiary to Wharf T&T, which is a fixed-line telecoms operator that can link data centres for disaster recovery. COL has two facilities in Hong Kong providing 650 workstation seats, soon to be increased to 950, co-located with their data centres. Users have three options. The first is to buy dedicated seats, available for disasters, but which can also be used any time for training or demonstrations. A second option is to buy guaranteed emergency access to enough seats throughout Hong Kong, and the third option is guaranteed seats in a particular geographical zone of Hong Kong. Deciding the best option is best done as part of a disaster recovery plan. Norris Hickerson, COL's vice-president of data centre and engineering services, said: 'For example, a company might need some dedicated seats, with its applications mirrored real time to the COL data centre, and use this for training. In addition, it might need some guaranteed seats for general disaster recovery and some shared seats in one location in case a local network goes down.' One solution for Hong Kong enterprises with an in-house data centre is to outsource data replication - providing a perfect mirrored image of their stored data - to a specialised company such as COL. For mirroring data, COL and its sister company, Wharf T&T, will offer its data centre and necessary telecoms link at a bundled monthly price. 'Telecommunications charges are still significant, but they are coming down dramatically,' Mr Hickerson said. Disaster recovery is a growing sector, but most organisations only commit to it when pressured by the need to comply with regulations, auditors, or customers that insist on protection of key data. According to COL's market analysis, the average small to medium-size (SME) firms in Hong Kong has only three PCs. 'Out of 300,000 SMEs in Hong Kong, I would be surprised if 1 per cent has disaster recovery,' he said. Disaster recovery plans need testing to be effective. Even in the United States, where the number of businesses with contingency plans has grown to 68 per cent, the benefits are doubtful, because only about 43 per cent actually test their plans. 'The lesson is clear. If you don't test your plan regularly, you don't have a plan,' Mr Hickerson said. What matters most in disaster recovery is how fast you can do it. IBM classifies its storage into seven tiers, ranging from tape storage up to site mirroring with automated recovery. In practice, three types of solution are on offer: Continuous Availability; Rapid Data Recovery; and Backup/Restore. Whatever recovery performance companies specify for each application, it is probably that as the data gets older it will be accessed less and therefore can be moved to a slower tier. IBM's Tivoli Storage Manager automatically backs up and archives data and enables easy switching of data between tiers. This year, IBM announced the Tivoli Storage Manager Express, which is a less costly application suitable for SMEs and branch offices. Most of the time, business continuity is much more important than disaster recovery. IBM research showed that enterprises typically had six times more planned downtime than unplanned downtime, and the cost to the business of planned downtime was almost as high as unplanned. An IBM spokesman said: 'More companies could benefit by designing their systems for high availability, using fault-tolerant, failure-resistant infrastructures and clustering software.' Alex Tse, Hitachi Data System's director of Business Continuity Practice, Global Solution Services, Asia-Pacific, said before venturing into business continuity solutions, it was important to develop a compelling case for the investment. 'Companies should conduct business impact analysis and risk analysis, to justify the investment in disaster recovery preparation,' Mr Tse said. Computer disasters were more likely to be caused by power outages than fire, typhoons or terrorist attacks. 'But equipment failure is less likely than human procedural failure, which may be due to lack of training.' Other human interventions include virus attacks and website sabotage. The risks can be identified, their business impact quantified in dollars and they can be prioritised when planning defensive measures. 'We advise clients to put controls in place, including failsafe systems, staff training and other preventative measures,' Mr Tse said. Disaster recovery usually involves the duplication of stored data from the primary storage system on the main premises of a company to a secondary storage system at a safe remote site. The storage systems must share a telecoms link that can handle the peak data flow to storage. For applications such as online trading, the two storage systems will be synchronised so that data and transactions are recorded in real time at both sites. For less critical applications, the data from the primary storage system will be backed up with a small time delay. This saves money because the memory cache that holds data before transmission and the bandwidth of the link between the data centres can be smaller. Usually data replication is done by a 'push' approach, which means that data stored on the primary system is copied into a small memory cache and then pushed to a secondary system, but this adds an extra burden to the primary system and requires substantial bandwidth on the connecting lines to accommodate the peak data flow for replication/modification. Mr Tse said HDS took a 'pull' approach for data replication. 'Our Hitachi Universal Replicator (HUR) makes use of the secondary system to pull data from the primary system. HUR leverages disk-based storage on the primary system, which can store terabytes of data, to act as a large buffer, enabling the data to be pulled on to the secondary storage system more evenly, reducing the peak telecoms traffic and cutting the cost. This change allows the primary system to use more of its capacity to run business applications and significantly increases the replication system's performance and operational resilience,' Mr Tse said.