• The traditional relational database can be a poor fit in these cloud environments. The promise of the cloud is that you can increase computing resources by allocating additional virtual machines.
• As a result, cloud providers offer simpler, nonrelational databases that can provide elastic on-demand scalability. The major offerings are:
• SimpleDB within Amazon AWS
• BigTable within Google App Engine
• Microsoft’s Azure Table Services and SQL Data Services
• These cloud databases differ in implementation detail, but share common core features:
• A hierarchical structure resembling a B-tree index or hash table. Rapid lookup via a single key value is provided.
• Flexible. Each “row” often can contain different “columns,” and columns may have multiple values or include a more complex embedded structure.
• Automatic geo-redundancy. Elements stored in the database are guaranteed to be replicated across multiple data centers.
• Automatic partitioning across multiple hosts and automatic scale-out as the size of – or demand on – the data store exceeds the capability of a single host.
• However, these databases are missing many of the features we’ve come to expect in a relational solution:
• Joins and complex queries must be implemented in procedural code.
• Transactions are not supported. You can’t create a set of changes that must succeed or fail as a single unit.
• “Eventual” consistency. When you make a change to a data item, it will be visible eventually, but not immediately, in all locations where a copy of that data is maintained.
• Will cloud databases disrupt the RDBMS market? Probably not. But the cloud environments available from Amazon, Google and Microsoft offer significant advantages over traditional hosting or in-house hardware models. Applications that exploit the cloud will find the scalability and zero-maintenance models of these cloud databases appealing. As these cloud applications emerge into the mainstream, the cloud database probably will establish a permanent and successful segment of the database taxonomy.
Posts
0 comments:
Post a Comment