Compute Service Planner™ is a SaaS-based inventory policy modeling and optimization application designed for compute pools (e.g., private clouds) defined by high-cost software stacks (HCSS) that are licensed for use by physical processor.
CSP reduces HCSS pool costs by 20-50% by employing mathematical optimization to rationalize and synchronize inventory management policy for both server refresh and software license procurement. Pool operators are able to describe server and software inventory, use-characteristics (e.g., server configuration, license entitlements, etc.), and conditions that signal the need for adding server and license capacity to a pool. CSP then determines the least-cost way to make compositional changes to the pool that minimizes licenses requirements.
HCSS pools are unique. There exist two competing methods to procure and add capacity to a HCSS pool or cluster:
Replace a quantity of servers in the pool with a quantity of new, higher capacity, low-cost servers that minimizes additional license requirements.
Add some incremental quantity of new servers and licenses.
This choice derives from the combination of characteristics that uniquely define HCSS pools:
Software stacks that cost 3 – 10 times more than the physical server.
High rate of processor performance improvement. Over the course of a typical 3-year enterprise license agreement, processor performance will reliably increase by 70 – 170%.
Servers organized into computer resource pools or clusters can be easily and quickly upgraded to capture far higher performance with negligible effects to service levels.
CSP inventory model
CSP is a deterministic, periodic review inventory model with discrete event (i.e., jump events) processing, and enables a wide range of variables and events to describe pool inventory and its behavior. CSP evaluates a HCSS pool’s configuration, server use-duration, and prospective compositional changes, which correlate to factors commonly found in standard inventory policy models such as lot-size, holding costs, and cost of ordering.
Minimization of cost per capacity for a compute pool is the primary cost objective. Because new servers rapidly improve over time, server lot size (configuration and relative performance) are almost never constant from one refresh event to another. CSP therefore evaluates an optimal inventory policy by combining traditional factors related to: “what to order, in what set size, when to order, and how much to order” with the additional dimension of: “what to replace, in what set size, when to replace, and how much to replace.”
Many feasible combinations are evaluated by CSP to determine an optimal solution. Consider the Pool Upgrade Matrix exhibit, which illustrates a very small pool of six servers that yields 36 feasible combinations for a single refresh event. Alternatively, consider a pool with 25, 100, or more servers, capacity additions arising once or twice each year and the need to account for these different combinations through sequences of incremental server refresh events.
How CSP reduces costs
CSP reduces cost by minimizing rapidly increasing holding costs of HCSS pools, which correlate to server use-duration. Holding costs are characterized by the opportunity to acquire newly arriving, market-available servers, which reliably provide exponentially increasing performance (i.e., lower cost per compute capacity), to which licenses may be reassigned. Because software licenses are priced per processor, the cost ratio of software-to-processor, referred to as the k-ratio, acts as opportunity-cost amplifier.
The higher the k-ratio, the more valuable every percent of server performance increase becomes to the HCSS pool operator.
Reducing opportunity-costs is therefore a function of determining when and how to refresh HCSS pools, align license entitlements, and synchronize license procurement. Server refresh now becomes a “value capture” process to control, retain and convert high rates of server performance improvement into higher compute capacity per license.
The k-ratio effect
The k-ratio signals the level of opportunity-cost amplification relative to server use-duration and rate of server performance increase. Whereas servers have traditionally been considered an asset, HCSS compute pool servers behave more like consumable or perishable inventory that have a quantifiable economic-life. And economic-life for HCSS pools is almost always far shorter than the typical 3-5 years of accounting-life or server refresh cycle.
For example, extending the life of servers beyond four years for a pool serving up http pages incurs modest opportunity-costs. However, operating servers that long in HCSS pools roughly doubles the number of required software licenses. Consider further an example for adding capacity to a pool. It is not unusual for the peak average utilization rate to drop from 60% to 20% after server refresh. Refresh a HCSS pool the same way, however, and the number of required licenses is roughly two times, or more, compared to dropping the pool’s utilization rate from 60% to around 45%.
These examples illustrate how “all you can eat” ELAs can act as a costly substitute for not rationalizing, synchronizing, and optimizing server and license inventory policies, and become cost traps when migrating HCSS environments to virtualized compute pools and private cloud solutions.
A better way
Your enterprise software vendors won the “value capture” process by default for non-virtualized, stand-alone server environments with “all you can eat” ELAs – because there wasn’t a compelling alternative. Of course, these vendors want to continue business as usual, regardless of your investment and adoption of virtualization and cloud-based technologies.
You don’t expect to pay higher energy rates to your electric utility simply because your servers generate more compute performance while using less energy. And now with CSP and your cloud infrastructure, your organization has the means to level the playing field to predictably manage and reduce high-cost license expense.
Let Ravello demonstrate how simple and easy it is to leverage your virtualization and cloud investments to reduce your total compute costs.