Data consolidation is a powerful tool for improving the efficiency of a data center. By using virtualization to put more data on fewer servers, organizations can reduce the cost of hardware, power consumption and space in the data center. But consolidation must be properly planned. Here are some tips for IT managers who are considering data management as part of an overall data consolidation project.
Tip 1: Calculate the costs. Data consolidation is never free. While many data centers can consolidate without adding new servers, that isn't always the case. Some organizations find they need fewer, but more powerful servers. While this represents savings in the long run, there is still the upfront cost of adding new servers.
More common is the need to make infrastructure upgrades, such as changes to networking or storage. For example, consolidation typically increases bandwidth requirements for the parts of the network connected to the newly consolidated servers. Find out about these costs before you commit to data consolidation.
Tip 2: Employ active data management. While it’s possible to simply dump all the organization’s data onto its consolidated servers, it makes more sense to actively manage the data during the process. A good place to start is a data census. Take an inventory of the organization’s existing data and how much of each type of data is stored. The inventory doesn’t need to be overly elaborate or precise, but it should include growth trends in data types so the IT department can effectively plan its consolidated servers for the future.
Tip 3: Target which data to consolidate. The first instinct is to consolidate all data on as few servers as possible, thus maximizing the benefits of consolidation. However, that’s not the best strategy for a number of reasons. Some data shouldn't be consolidated because of the performance hit some applications will take from sharing hardware resources. This includes transactional databases, which are heavy consumers of computing resources.
In other cases, security considerations dictate that some data should reside on separate servers. For example, sensitive data, such as personally identifiable information, is best kept on a separate, more secure server. And some data isn't worth consolidating; for example, outdated information that is cluttering up the storage system, or low priority data that will seldom be accessed. If you don’t want to discard this data, archive it to tape or other media.
Tip 4: Decide where the data will be stored. Determine the physical location of the organization’s consolidated servers. To minimize performance hits, the IT staff will want a topology that limits the number of hops (among subnetworks) between the data source on the consolidated server and the data’s consumers. If the organization uses multiple servers, decide what data is going to go on which server. To minimize the load on server resources — and the possibility of a single point of failure — spread similar kinds of data over several consolidated servers. If possible, don't put data with the same usage pattern or demand for the same sorts of resources on the same server. Divide it among the organization’s consolidated servers to improve overall performance.
Tip 5: Don’t overload the hardware. Finally, while there are very real benefits from consolidation and virtualization, remember that underlying it all, there is real hardware with real limitations on capacity, bandwidth and processing power. This isn't just a matter of spreading the same type of data across several servers. Consider the demands that the data will put on the server, even if it represents different types of data.
Also, pay attention to usage patterns: Some kinds of data will spike usage, putting a load on the server at certain times. A classic example of this is accounting data at the end of the month when the books are closing. If the hardware is overloaded, it will have performance problems no matter how clean the organization’s consolidation plan looks on paper.