Big Data: So What! That’s Why You Virtualize
How data virtualization enables Big Data volume, variety, velocity and value
Mar. 11, 2013 10:00 AM
Big Data! Yes it's BIG!
The volume is BIG! The variety is BIG! The velocity is BIG!
And hopefully the business value is BIG!
New Opportunities Bring New Ways to Leverage Proven Technology
There is no shortage of media articles, analyst reports, tradeshows, blogs and other source of Big Data technology insight and advice.
But it strikes me that in our search to be on the leading edge, we may be overlooking some great existing technology.
In fact, some technology, for example data virtualization, is even more useful in a Big Data world.
What Is Data Virtualization?
Data virtualization is an agile data integration approach organizations use to gain more insight from their data. This includes traditional sources such as transaction systems, data warehouses and more as well as new sources such the cloud and Big Data.
Unlike data consolidation or data replication, data virtualization integrates these diverse data types without costly extra copies and additional data management complexity. Seriously, if the data is already big, why make it even bigger by copying and storing it again and again?
With data virtualization, you respond faster to ever changing analytics and BI needs, fast-track your data management evolution and save 50-75% over data replication and consolidation. In other words, you deliver value, the most important V but often not listed with the 3 Vs of Big Data (Volume, Velocity & Variety).
Variety Is Big Data Integration Challenge #1
Often, the biggest Big Data integration challenge is variety, not volume. Consider all the different Big Data types that may require integration:
- Massively Parallel Processing based Appliances - Examples include EMC Greenplum, HP Vertica, IBM Netezza, SAP Hana, and more
- Columnar/tabular NoSQL Data Stores - Examples include Hadoop, Hypertable, and more
- XML Document Data Stores - Examples include CouchDB, MarkLogic, and MongoDB, and more
- Key/value Data Stores - Examples include Cassandra, Memcached, Voldemort, and more
Fortunately integrating heterogeneous data sources is the original raison d'etre of data virtualization. Why do you think many still call it data federation?
Volume Is Big Data Integration Challenge #2
As listed above there are many ways to store and manage big data. Similarly, a plethora of analysis tools exist such as MapR, Karmasphere, Alpine Data Labs and more.
The biggest volume challenge is how to query large data sets from these high-volume sources at speed in order to feed these analytics?
The answer is data virtualization.
Data virtualization platforms use sophisticated rule- and cost-based query-optimization strategies that automatically create a query plan that optimizes processing and performance, with minimum overhead.
Advanced Query Optimization Is the Key to Data Virtualization
Here are but a few of the query optimization strategies and techniques data virtualization provides:
- Pushdown - Data virtualization offloads as much query processing as possible by pushing down select query operations such as string searches, comparisons, local joins, sorting, aggregating, grouping into the underlying data sources. Thus you can take advantage of native capabilities.
- Parallel Processing - Data virtualization optimizes query execution by employing parallel and asynchronous request processing. After building an optimized query plan, the data virtualization server executes data service calls asynchronously on separate threads, reducing idle time and data source response latency.
- Distributed Joins - Data virtualization detects when a query being executed involves data consumed from different data sources and tries to employ distributed query optimization techniques to improve overall performance and minimize the amount of data moved over the network. A variety of sort-merge, semi, hash and nested-loop joins are leveraged depending on the nature of the query and data sources.
- Caching - Data virtualization can be configured to cache results for query, procedure and web service calls. When enabled, the caching engine stores the cached result sets and queries them as appropriate.
- Advanced Query Optimization - Data virtualization provides a number of additional techniques and algorithms include data source grouping, join algorithm selection, join ordering, union-join inversion, predicate pooling and propagation, and projection pruning.
- Integrated Network and Database Optimization - Even in a Big Data world; network bandwidth is generally the scarcest resource in the query processing pipeline. So reducing the amount of data that needs to be transferred has a significant impact on the latency and overall performance. Data virtualization optimizes the network and the query processing capabilities of underlying big data sources intelligently, in combination.
Value and Velocity are Big Data Integration Challenges #3 and #4
Big Data itself only has value when the data is analyzed. This analysis provides value by uncovering drivers for growth, finding better ways to attract and retain customers, and identifying opportunities for innovation and costs reduction.
As such the fastest path to Big Data analysis is also the fastest path to business value.
But everyone knows that providing analytics with the data required has always been difficult, with data integration long considered the biggest bottleneck in any analytics project.
The Data Warehousing Institute confirms this lack of agility. Their recent study stated the average time needed to add a new data source to an existing BI application was 8.4 weeks in 2009, 7.4 weeks in 2010, and 7.8 weeks in 2011. And 33% of the organizations needed more than 3 months to add a new data source.
Data Virtualization Provides Velocity along with Analytic Value
According to Data Virtualization: Going Beyond Traditional Data Integration to Achieve Business Agility, data virtualization significantly accelerates data integration agility. Key to this success is data virtualization's
- Streamlined data integration approach
- Iterative development process
- Adaptable change management process
Using data virtualization as a complement to existing data integration approaches, the ten organizations profiled in the book cut analytics project times in half or more.
This agility allowed the same teams to double their number of analytics projects, significantly accelerating the business value delivered. In other words, value with velocity!
Variety, Volume, Velocity and Value
Big Data is all the rage. And at first glance, the Big Data variety, volume, velocity and value challenges may seem extraordinarily difficult.
Proven technologies, such as data virtualization, provide proven approaches to addressing these "big" challenges.
So if Big Data is on your agenda, don't forget to make a big commitment to data virtualization. You'll be glad you did.