The SAP NetWeaver BI Accelerator offers two main benefits to the companies that deploy it alongside an SAP NetWeaver BI system:
- Enhanced productivity: The increased performance and other features of the accelerator enhance the productivity of BI system users.
- Reduced TCO: The total cost of ownership (TCO) of the BI system is reduced by introducing the accelerator compared with other strategies for boosting performance.
It’s worth examining these benefits in detail.
The productivity of SAP NetWeaver BI system users is enhanced with the accelerator for several reasons:
- Performance: Response times delivered by the BI system for analytic queries are often much faster. Even for huge InfoCubes with billions of rows, results can be returned in seconds.
- Stable response times: The response times for repeated executions of a query or for execution of similar queries remain stable even under high system load. The response times also vary predictably under changes in data design and query design.
- Scalability: The BIA system can easily be extended incrementally with additional hardware.
- Robustness: The BIA blade hardware is reliable and requires much less routine administration than legacy architectures.
- Usability: The performance and stability features encourage users to explore their data. With the accelerator, they can drill down to the facts and explore their data in an ad hoc manner, without fear of timeouts or overconsumption of system resources.
Let’s look at these five reasons one by one.
Compared with typical response times for standard company SAP NetWeaver BI systems in which InfoCubes have been supplemented with manually defined aggregates, analytic query response times with the accelerator are often tens or even hundreds of times better.
Scenarios with challenging response time requirements are numerous. For example, a call center for handling customer calls needs to be able to extract relevant information from customer records within a few seconds. In such scenarios, the IT department faces demanding service level agreements, and high performance is not a luxury.
Stable Response Times
The response time improvement is consistent. If the system is not under heavy load, response times for a given query are constant whenever it is executed. Also, similar queries are answered in similar times. By contrast, response times with aggregates often vary unpredictably, depending on whether an existing aggregate exactly matches a query or whether a database optimizer finds a perfect index for the query. The problem with relying on a database optimizer is that it is a “black box” product containing proprietary algorithms that impact performance in ways a user cannot anticipate. With the accelerator, users learn to expect fast responses to new queries. This encourages ad hoc reporting and creative analysis of the data. This in turn extracts the added value hidden in the huge amounts of data generated by a modern business.
The average response times of the accelerator for increasing amounts of data increase in proportion, with near-linear scaling. And as the number of simultaneous users increases, average response times increase gradually, again with near-linear scaling. For example, if the single-user response time for a class of queries is less than 1 second, the response time for the queries under high system load may be less than 4 seconds (Project Jupiter, Section 6.1). In any case, good performance under increasing load is achieved without hitting a wall.
Nowadays, many companies have accumulated data repositories of several terabytes and expect to accumulate data at an increasing rate in future. Companies in the retail sector with point-of-sale (POS) data, utilities with detailed records for millions of customers, and telephone companies with billions of call records are some obvious examples. In other application areas, new technologies such as radio frequency tagging (RFID) will generate an order of magnitude more data in the near future. For such scenarios especially, good scalability is a must.
This near-linear scalability is achieved by modularization. Each blade server is a complete hardware server module in a standard and compact format. Companies deploy as many blade servers as they need to handle their user load and can later add blades one by one to handle increasing load. Because the blades are produced and sold in large numbers, they can offer massive processing power and memory capacity at commodity prices that break through previous price-performance barriers.
The new hardware paradigm leveraged by the accelerator is a great improvement on earlier hardware. Modern blade servers require very little routine administration, are highly reliable, and offer automated failover. Thus a few administrators can be responsible for a large landscape. Further, instead of rebuilding their entire landscape to introduce new hardware, companies that deploy the accelerator begin with an encapsulated scenario where the risks are contained and the benefits are instantly visible.
On the software side, BIA indexes typically require far less maintenance effort than the aggregates they replace. Change runs and roll-ups are required only for one object per InfoCube and generally run much faster for BIA indexes than for database aggregates because the work is parallelized over the accelerator blade nodes. This also allows a company to schedule much more frequent data alignments and to improve performance in scenarios where the process of building and updating aggregates would be far too cumbersome.
The opportunities opened up when the accelerator is in daily use are game changers. Knowledge workers asking analytic queries can get their answers more quickly and hence waste less time waiting, but that is only the start. With accelerator performance, complicated navigations over large data structures that were previously impractical become feasible and new patterns in the data become open to discovery. An additional benefit of accelerator deployment is that knowledge workers no longer need special training to follow the constraints of prebuilt BI aggregates.
As an example of a new opportunity, consider a retail company that uses its accelerator to perform assortment analysis at the stock keeping unit (SKU) and store level. If the accelerator is suitably sized, the company can choose to perform the analysis at point-of-sale (POS) till receipt and SKU level, for basket analysis and customer segmentation. This enables the company to pursue a much more refined and responsive assortment strategy.
Moving to the other half of the value proposition, companies can expect to enjoy reduced TCO for their SAP NetWeaver BI systems by deploying the accelerator rather than other productivity boosters for several reasons:
- No aggregate maintenance: Without aggregates to maintain, fewer resources are needed for routine administration.
- System consolidation: Dependencies between the database and reporting are relaxed, so data loads and reporting no longer need to be scheduled serially. Because the accelerator takes over the reporting load, system copies made to handle that load can be consolidated.
- Longer hardware usage: Investment in new database hardware can be delayed or avoided. Growth in reporting demand can be accommodated without adding load to the existing BI landscape.
- Ease of reporting: Long and inflexible approval processes for reporting can be replaced by shorter and more open processes. Changes in reporting requirements can often be accommodated at no additional cost.
- Simplified data modeling: Because manual performance tuning is no longer necessary, companies can afford to develop and release new analytic scenarios more quickly to their lines of business.
Again, let’s look at these reasons in detail.
No Aggregate Maintenance
The accelerator creates only one BIA index for each InfoCube, so companies no longer need to waste their administration resources maintaining multiple aggregates. Because the definition and creation of aggregates is a task requiring a high skill level, and the ongoing task of maintaining them is laborious, the human resource costs for this approach to performance optimization are high. By contrast, both the initial indexing effort for an InfoCube and the subsequent maintenance effort for change runs and roll-ups are an order of magnitude lower. Furthermore, in most cases, the performance gain obtained is greater and more consistent.
Some companies currently copy BI systems to manage high overall system loads. This creates new complications regarding data replication and synchronization and drives up administration costs. The costs can be avoided by deploying the accelerator to handle the reporting load on the systems. The accelerator can raise the effective capacity of a BI system sufficiently to allow multiple systems to be consolidated.
Longer Hardware Usage
Deployment of the accelerator enables companies to delay or avoid investment in more traditional hardware for increasing capacity. The lifetime of an existing database solution can be extended significantly when an accelerator is available to take some of the load. And since BIA indexes replace disk-based aggregates, the database consumes less disk space, which may be significant if database licensing is based on data volumes. Moreover, the accelerator enables companies to profit from the new hardware without premature write-down of their existing hardware investments.
Ease of Reporting
The cost reduction can be dramatic. Consider a typical scenario in a large company where the board consumes regular reports from the controlling department. Without the accelerator, a set of reports needs to be defined and prepared, then supported with tuned aggregated that must be maintained on a regular basis. This ongoing effort is a significant cost driver that severely impairs flexibility. If the board wants a new report, the workflow consequences in controlling can be disruptive. By contrast, with the accelerator, the situation is transformed. The board still gets its reports but now the controlling staff who sweated to deliver them can work much more efficiently. Also, the other company users who analyze the data more intensively can now work more creatively. They can hunt down new opportunities for optimization and raise an alarm faster if necessary. The board, too, can ask for new reports at short notice without generating excessive costs.
Simplified Data Modeling
When accelerator performance makes manual performance tuning less critical, not only can companies develop and release new analytic scenarios more quickly to their lines of business but they can also afford to be less performance-aware during the design of InfoCubes. Although good design is still important for the InfoCubes, it becomes less critical. This means in turn that the granularity of the data in an InfoCube can reach down to line item level, rendering the InfoCube sufficient for all reporting needs and allowing DataStore objects to be optimized for writes, not reads, and simplifying the modeling task in the data warehouse. In this way, accelerator deployment offers an opportunity for a company to reconsider its data warehousing practices more generally. Any companies who are starting from scratch with a new SAP NetWeaver BI project can benefit immediately from the increased flexibility in data design and modeling.
The Benefits in a Nutshell
To summarize, the SAP NetWeaver BI Accelerator value proposition is that deployment brings two main kinds of benefits:
- User productivity is increased due to the improved query performance, more stable response times, high scalability and robustness, and the opportunities created by enhanced usability.
- The total cost of operation for analytic reporting is reduced by savings in aggregate maintenance, system consolidation, longer write-downs for existing hardware, ease of reporting, and simplified data modeling.
SAP provides a tool for calculating the benefits a company can expect by deploying the accelerator. The ROI calculator for the SAP NetWeaver BI Accelerator is an easy-to-use tool to qualify and quantify the value of BI accelerator deployment by measuring the return on investment (ROI) to expect from introducing the accelerator into a landscape. With the right input data, the tool calculates the ROI in precise numerical terms. The calculator is built with BusinessObjects Xcelsius as an intuitive standalone tool embedded in a Microsoft PowerPoint file that can be customized for use in larger presentations. It is available on request from SAP account executives and others.