Why Exalytics? Why not Exalytics? - Part 1

Why Exalytics? Why not Exalytics? - Part 1

Guest post by

Karan Chadha, Senior Associate Consultant, Infosys

Performance has been a major pain point for numerous BI deployments. Poor report performance, apart from being a reason for lower adoption rate of BI, has at times caused entire BI projects to fail.

BI developers have adopted multiple strategies like caching, aggregation, materialized views, database tuning etc. for performance enhancement. However, despite these strategies being fairly effective, consistent sub-second response times and speed-of-thought analytics has always been a distant dream.

Exalytics, the engineered BI machine by Oracle, attempts to put an end to these perennial performance bottlenecks. Leveraging Times Ten's in-memory database technology, Exalytics is aimed at providing a potent solution for conquering performance demons. Speed-of-thought analytics is ultimately a reality in the world of Oracle BI!

If we analyze a query life cycle, most of the time goes into fetching the result set from the database. With in-memory database technology, the data is stored in the RAM of the machine hosting the BI application. Consequently, with data fetch becoming extremely fast and networking bottlenecks out of picture, query performance shoots upwards multiple times.

Exalytics comes with 1TB RAM, out of which only 0.4 TB can be used for data storage. Times Ten has data compression capability of up to 5X implying that a maximum of 2 TB data can be stored in RAM. But what if my data is more than 2 TB? How do I choose the most optimum dataset to reside in RAM? How do I push data from my data sources into Exalytics' RAM? Read on to check how Exalytics is put to use under different scenarios.

1. OBIEE is reporting on top of a large size OLTP databaseAn OLTP database will typically contain large amount of transactional data at lowest level of granularity. However, BI users would usually analyze data at an aggregate level. So, the most optimum 'hot' data to be pushed into RAM would be aggregate tables. Exalytics provides a Summary Advisor utility that recommends a set of aggregates to be pushed into RAM based on usage tracking statistics. Summary Advisor analyses months and years of usage tracking data and suggests aggregates that are likely to bring maximum performance improvement. The utility generates a script to create aggregate tables in the RAM, populates data and do all repository level changes. This script is executed using NQCMD utility. For any data refresh in the source tables to be reflected, the script is required to be re-run.
2. OBIEE is reporting on top of a large size Data Warehouse (DW)       Two situations are possible here. One, the DW has limited or no aggregates, and second, the DW already has all the possible aggregate tables built-in. In the first scenario, the situation is no different from the one discussed above. Summary Advisor can be used to build aggregates in RAM. However, in the second case, with aggregates already in-place, they can simply be pushed into RAM using a data replication tool like Oracle Golden Gate. Repository level changes will have to be manually made to make the BI server aware of the aggregates.
3. OBIEE is reporting on top of a small size DW or OLTP databaseThis situation can bring maximum performance improvement since the entire data can be replicated into the RAM. This would mean that the BI server is able to service every request from the data available in RAM. Another use case under this scenario could be that a particular data mart which is facing performance issues is replicated into the RAM in case DW size is larger than what could be accommodated.

It's a no-brainer to guess that with data residing inside RAM, reports would take a split second to show up. Tests on customer data have established that performance improvement of up to 15 times is common and in some cases it has even been 50 times! So should I go ahead and recommend Exalytics to my client?

In Part 2 of this blog, I will cover the new visualizations that Exalytics offers and some of the situations where Exalytics may not be a good fit.

Source: http://www.infosysblogs.com/oracle/2012/06/why_exalytics_why_not_exalytic.html