IBM System z mainframe data centers buy a lot of software, and many data center managers pride themselves or negotiating the best price for each piece of software (or their procurement departments do). Ongoing mainframe software benchmarking based on actual prices paid, however, suggests this is not the case at all. The difference between the average price for a piece of mainframe software and what best-in-class (BiC) shops pay can amount to millions of dollars.
The results of mainframe software benchmarking from the Minneapolis-based ISAM Group may shock you. Since 1996 ISAM has collected over 7 million price data points from over 500 major companies, 13 leading outsourcing firms, six of the top ten largest states, and US and Canadian federal agencies. Among its findings: BiC companies pay less than half what their peers pay for software regardless of data center size. It found, for example, there is an average difference of $4 million on what companies pay for software in each of the Top 10 functional categories.
The top 10 functional categories:
- DB/file mgt.
- Production mgt./system control
- Dev tools/language code generators
- OS application server
- DB/File mgt. utilities
- Dev tools/language extender
- Resource/performance mgt.
- Dev tools/test-debug
- DB access tools.
These ten functional categories account for 57% of the total software spend.
Not surprisingly, IBM’s software market share runs 50%-69% at the average data center, 80% at BiC data centers. Most data centers, Mike Swanson, ISAM head of research, can cut their software costs in half by becoming BiC, which happen to have higher concentrations of spend with IBM and CA. BiC shops spend about half of the average on software.
To become a BiC data center in terms of software cost start by focusing on the top 10 functional areas where the bulk of the software spend lies. The spend variance is greatest, Swanson adds, with the largest categories and with database access tools.
Then look closely at your mix of vendors. ISAM research shows that disparity in product pricing, changes in product mix, and variations in product deployment cause dramatic variations in costs by vendor. Those vendors with the greatest price variation offer the best opportunity for competitive replacement, which will reduce overall software spend.
Another trick of BiC data centers is to reduce core costs by reducing the environment. In software that means reducing the number of different software products deployed. Data centers typically have overlapping products and redundant functionality. A shop, for example, may have 10 database utilities when they really only need three. Similarly, ISAM found ample opportunities among data centers to trim the number of development tools they have without reducing functional capabilities.
And when deciding which products to cull, of course, look at vendor pricing. The difference between an average data centers and BiC data centers can be startling. For a monitoring tool, ISAM found that a small shop, under 1000 MIPS, might pay over $35 per cost unit on average for the product while a comparable size BiC shop would pay just $9 for the same product. The average large shop, one over 5000 MIPS, will pay almost $4 per cost unit for the same monitoring tool compared to $1.60 for a BiC shop.
Finally, avoid buying enterprise licenses, an approach favored by procurement. On the basis of cost per functionality, they are a bad deal. You are paying for convenience and flexibility. You are better off from an overall software licensing cost standpoint to carefully determine the specific capacity you need to license and pay only for that amount. This will require work to understand your exact requirements and usage patterns, but the payoff, as Swanson shows, can be substantial.