What scalability problems have you encountered using a NoSQL data store? [closed]

Answer 1

I have no first-hand experiences., but I found this blog entry quite interesting.

Answer 2

I find the effort to map software domain objects (e.g. aSalesOrder, aCustomer...) to two-dimensional relational database (rows and columns) takes a lot of code to save/update and then again to instantiate a domain object instance from multiple tables. Not to mention the performance hit of having all those joins, all those disk reads... just to view/manipulate a domain object such as a sales order or customer record.

We have switched to Object Database Management Systems (ODBMS). They are beyond the capabilities of the noSQL systems listed. The GemStone/S (for Smalltalk) is such an example. There are other ODBMS solutions that have drivers for many languages. A key developer benefit, your class hierarchy is automatically your database schema, subclasses and all. Just use your object oriented language to make objects persistent to the database. ODBMS systems provide an ACID level transaction integrity, so it would also work in financial systems.

Answer 3

I have used Vertica in the past.It relies on columnar compression & expedites disk reads and lowers storage needs to make the most of your hardware. Faster data loads and higher concurrency lets you serve analytics data to more users with minimum latency.

Earlier, we were querying Oracle database having billions of records & the performance was very sub-optimal. The queries took 8 to 12s to run, even after optimizing with SSD. Hence, we felt the need to use a faster read optimized, analytics oriented database. With Vertica Clusters behind the lean service layer, we could run APIs with sub-second performance.

Vertica stores data in projections in a format that optimizes query execution. Similar to materialized views, projections store result sets on disk OR SSD rather than compute them each time they are used in a query.Projections provide the following benefits:

1. Compress and encode data to reduce storage space.
2. Simplify distribution across the database cluster.
3. Provide high availability and recovery.

Vertica optimizes the database by distributing data across cluster using Segmentation.

1. Segmentation places a portion of data on a node.
2. It evenly distributes data on all nodes. Thus, each node performs a piece of the querying process.
3. The query runs on the cluster and every node receives the query plan.
4. The results of the queries are aggregated and used to create the output.

For more, please refer to Vertica documentation @ https://www.vertica.com/knowledgebase/