Let's examine five questions storage managers should ask when considering cloud data storage:
Because of the architecture of cloud storage, you can expect high latency on file or directory access requests. So if your next project is a SQL server database or a mail server for a large site, then cloud storage probably isn't the way to go.
But if you're implementing a file server for your remote sales force, then the access time of the cloud is probably consistent with the Wi-Fi or shared Internet connection they would be expecting to use. The cloud also provides access to the data from anywhere that has an Internet connection.
There's a user experience component to this as well: Are users expecting the performance of local storage? It's important to communicate upfront that moving data to the cloud may affect the user experience. If end users have local file servers or even centralized file servers in a data center accessible by a private WAN, accessing data from the cloud will most likely be slower. This can be mitigated by providing multiple Internet connections or a higher level of network service for outbound storage requests.
Most cloud storage providers give a 99.99% service-level agreement (SLA) as their peak. A few, most notably Nirvanix Inc., advertise a 100% SLA. However, when considering these claims, it's important to understand what they are guaranteeing.
Most cloud storage providers reside in co-locations or hosted data centers, and are dependent on yet another vendor's infrastructure. If the hosting provider has connectivity issues, will the potential loss of access to your data be covered under the cloud provider's SLA or will it fall back to the hosting provider?
All cloud storage providers offer encryption for data in flight. But if the data for the cloud is sensitive, then the security of that data must be addressed once it lands on the service provider's infrastructure. Several providers now support encryption of data at rest, but you need to find out if the provider you're evaluating has this capability. Additionally, keys or logins for data encryption may become an issue if there are several pools of data that need to be segregated. Depending on the vendor, user accounts may need to be set up for each individual user in the environment.
Most cloud storage providers will charge customers based on storage access in terms of the type of access request and the space utilized. This becomes important when a large number of search either for file names or text within a file. For example, if files are required to be indexed or scanned for viruses on a regular basis, the cost of cloud storage may increase significantly.
E-discovery data stores, application data caches and temporary file storage are other types of data storage that do not work well with the cloud cost model.
For businesses that require fast access to recent files and slower access to older files, the cloud may still be a viable storage alternative if a caching appliance can be housed in the local network. These machines will store the most recent data sent to the cloud assuming that the data has a higher probability of being accessed in the near future.
This was first published in November 2009