Welcome to 2015! This is traditionally the time of year when most of us promise to improve our lives in some way, perhaps by starting a new exercise regime or finally kicking that expensive smoking habit. But maybe this year you can also resolve to boost your personal profile and improve your company’s bottom line by taking a long hard look at some of the IT costs that fall within the scope of your job role. Our theme for January is “Reducing Costs”, and throughout the month we’ll be giving you some insights and ideas that could help.
Gartner recently estimated that overall IT spending increased by 2.6% in 2014, to over 3.5 trillion US dollars. I know what you’re thinking: “I’m just a lowly DBA / administrator / sysprog / developer. What can I do that will make a difference?” Well, you might be able to do more than you think. Let’s break down the major categories of IT spending (shown in Figure 1 below), and see how they might apply to your role.
In this article, I’ll present an overview of some of the cost reduction approaches you could consider in each of the spending categories above. We’ll pick up on several of these themes in more detail in other articles later this month.
Accounting for 17% of total IT spend, hardware covers everything from server purchases and mainframe upgrades to telecommunications equipment. While there’s probably little you can do to influence the price that Dell or IBM charge for new rack servers and mainframes, you may be able to help your organisation “sweat the assets” a little longer. Most IT systems have the potential for tuning and optimisation to reduce hardware resource usage, and deferring a major hardware upgrade can have a significant impact on your annual IT spend.
From a DB2 perspective, there are two major aspects to consider:
- Reduce CPU Consumption. Many hardware upgrades are driven by peak CPU demands exceeding the available hardware capacity, which can elongate the elapsed time for critical processes and cause service levels to be missed. If you can identify these peaks, you can analyse the processes responsible for the CPU demand, identify those that use DB2 and take steps to reduce the demand through tuning and optimisation. You may also be able to move workloads from one server to another, or from one time slot to another, in order to reduce the peak demand and defer the server upgrade. In the mainframe world, some software is licenced according to the overall CPU capacity of the server, so avoiding a CPU/server upgrade could also help to keep software costs under control. Finally, you can consider trading CPU for memory: many mainframe and midrange servers are under-configured with memory and upgrading this (or exploiting unused memory capacity in your existing setup) can often be a quick, easy and relatively cost-effective way of reducing CPU demand and thereby prolonging server life. DB2 examples of this include increased use of buffer pools, dynamic statement cache and sort work areas.
- Reduce Disk Utilisation. Although relatively cheap compared to CPU resource, disk space is still an important component of overall IT hardware costs: especially the more expensive SSD (Solid State Disk) technology which is increasingly used for performance-critical applications on both the mainframe and midrange. Exploiting DB2’s various compression technologies can provide a double benefit of reducing overall disk space requirements / costs and (in the case of table compression) improved buffer pool utilisation. In particular, DB2 for LUW has introduced some very effective new compression options in recent releases, including adaptive compression in version 10 and the newer actionable compression for columnar tables in BLU.
Software accounted for 21% of the worldwide Global IT spend in the Gartner report, but here’s where the worlds of mainframe and midrange diverge a little. The majority of licence fees for IBM and ISV mainframe software are calculated and paid monthly, using a system known as MLC (Monthly Licence Charge). There are many flavours of this (WLC – Workload Licence Charge, AWLC – Advanced Workload Licence Charge, etc), but most are based on the peak 4-hour rolling average CPU utilisation observed during the previous calendar month. Therefore, ongoing mainframe software costs can be reduced using exactly the same peak DB2 CPU reduction measures that were briefly outlined in the hardware section above, with the financial benefits being apparent almost immediately.
One new and interesting development in this area is Mobile Workload Pricing, which offers significant MLC discounts for z/OS workloads originating from mobile devices such as mobile phones and tablets. Another related topic is DB2’s increasing exploitation of zIIP speciality engines, which is able to reduce peak mainframe CPU usage (and therefore MLC fees) by redirecting certain types of workload to zIIP engines. We’ll talk more about the possibilities and challenges of both of these in follow-up articles later this month.
On the midrange platforms most vendors (including IBM) base their licence fees on the total capacity of the server platform (or server logical partition) that they are implemented on. A commonly used approach for this is to use Processor Value Units (PVUs). Each type of CPU is given a PVU rating for each processor core (according to its power/capability) and each software product has a published list price per PVU. A one-time charge is then levied based on the product’s PVU cost, the PVU rating of the CPU and the number of processor cores available to it. So, once again the ability to reduce the peak CPU consumption of a given workload may make it possible to reduce the number of processor cores made available to it, which in turn would reduce the cost associated with purchasing the software. Of course, this approach can only really save money on software licencing if you’re able to tune applications before the production PVU entitlement has been purchased: if you’ve already paid for the higher capacity it’s unlikely you’d be able to obtain any kind of refund. However, DB2 workload tuning could allow you to run additional work within that environment without having to purchase additional PVU entitlement, so there may still be a cost reduction angle.
One final point to note regarding software savings. Many large enterprise customers are considering the benefits of large-scale server consolidation using Linux on System z. In many cases the cost savings attributable to software licencing alone are enough to provide a compelling business case for such consolidation projects. This topic was covered extensively within an earlier article entitled When Worlds Collide – Running DB2 for LUW on System z.
The last and largest part of total worldwide IT spending is down to you and me: nearly two-thirds of IT budgets are spent on permanent staff, contractors, consultants and outsourcing. As Moore’s Law continues to drive down the cost of hardware and competitive pressures keep software pricing in check, we are all consuming an ever-increasing proportion of the total IT spend for our organisation. It’s therefore not surprising to see a relentless focus on productivity features in enterprise software such as DB2, with the objective of allowing us to design, develop, test and maintain new systems with fewer resources/skills than would previously have been possible.
Every new release of DB2 for z/OS and DB2 for LUW has many such features, with recent examples including temporal tables, transparent archive tables, automatic clean-up of pseudo-deleted index keys and variable array support in SQL stored procedures.
While many organisations ensure they upgrade DB2 regularly to stay within the currently supported release boundaries, fewer have the luxury of being able to take the time to fully exploit all of the new features in a release. That means there’s often a treasure trove of new functionality waiting to be rolled out to developers and DBAs to make them more productive and reduce elapsed time for application design and development. Taking the time to run an audit of the new functionality delivered within the last few releases of DB2 can pay big dividends. We’ll be looking at a few of the most commonly overlooked productivity features in a later piece.
This article presented a high-level overview of the major cost categories that contribute to worldwide IT spending, and briefly discussed how each one related to DB2 workloads. It also outlined some of the ways in which a DB2 professional (that’s you) could potentially reduce these costs, thereby improving company profitability. We’ll dig into several of these themes in more technical detail in the articles that follow over the new few weeks. Happy New Year!
 See IBM’s Processor Value Unit [PVU] licensing for Distributed Software web page for further details