by Angela Guess
A recent article takes a look at some of the strengths and weakness of NoSQL in the context of three enterprise applications: portal, reporting and analytics. Starting with portal, the article states, “Anything but the simplest static HTML page has some sort of data access module at the back-end that formulates a query to fetch relevant records or composite views from one or more data stores. Generally, the overall fetching operations on many NoSQL technologies could be several orders faster than RDBMS as one can fetch the data from the shards in a scale-out parallel manner.”
Moving on to reporting, the article continues, “Reporting applications inherently employ more complex queries than portal. A typical read query may have multi-table joins, nested joins, filters etc. Any write or update operation may cascade to multiple tables and will have to maintain referential integrity constraints. In general, NoSQL is a poor choice to support these applications (but products like HBase or GFS supports time interval-based and condition-based queries). Further, if you need to prepare your data for reporting by running elaborate batch applications to transform data with many-to-many mappings, there is no easy alternative to trusted SQL-based loading and transformation engines.”
On the topic of analytics, the article states, “The sweet spot for many NoSQL data stores is on our third set of applications: those that heavily use analytics, especially when it comes to processing to large amount of unstructured data (aka big data). Used in conjunction with MapReduce, NoSQL can parallelise and quickly perform complex operations such as statistical aggregates, filtering, grouping and sorting. The scale-out capabilities of NoSQL databases – e.g., Cloudera and Aster Data Systems — allow handling of large volumes of unstructured data suitable for analytics. In terms of price/performance, NoSQL quickly gains ground over traditional data warehouses and BI appliances.”
