The latest vSphere 5.0 licensing update caused a lot of commotion in the virtualization space. People are debating the good and bad and pointing out various economical reasons for not adopting the new approach, and either moving to another platform or not upgrading. Some interesting (and valid!) points. The question is WHY? Why does VMware, yet again, change its licensing? One simple answer is “because they can”. A deeper more serious answer is that VMware is struggling with the hardest challenge we all are facing: how to maximize the ROI from our virtualization deployment, i.e. how to assure the performance of the applications while utilizing the infrastructure as efficient as possible.
Performance management is concerned with two fundamentally conflicting goals: (1) maximize resource Utilization; and (2) minimize Application's Delay (response time); the larger the utilization, the higher the delay. This conflict is resolved by redefining the goal: maximize application resource utilization, while guaranteeing QoS.
Figure 1 depicts these delay-utilization tradeoffs. The vertical axis depicts the respective delay of an application server executing at a Virtual Machine (VM). The horizontal axis depicts the utilization of a compute resource by the VM. The operating point moves up/down the operating curve when utilization increases/declines. The left part of the operating-curve is flat; even large increase of utilization results in minor increase in delays. This means that utilization may be efficiently increased to accommodate higher workloads without hurting applications quality of service (QoS). Keeping the operating point too far to the left is "Inefficiency". In contrast, the right side of the curve rises steeply; even a minor increase in utilization results in large delay penalty to QoS. This means that utilization may be best kept from creeping into this "bad-QoS" region.
Figure 2 depicts the fundamentals of tuning, in terms of two thresholds. The "QoS threshold", on the right, represents a bound on utilization beyond which delay becomes unacceptable. The "inefficiency threshold", on the left, is the minimal acceptable target utilization. These two thresholds define the acceptable "Operating Range."
Ideally, it would be best for the operating point to reside at the upper right corner of the operating range, right on at the QoS threshold. This would maximize utilization while still assuring applications QoS. However, such pinpointing of the operating point is ruled out by the fluctuations of utilization. Instead, one may seek to keep the operating point within a target Operating Box, where it maximizes utilization while assuring QoS.
VC 5.0. just added yet another variable to the already complex ROI puzzle you need to solve, as you now need to take into account the cost of vSphere licensing, in order to ensure that your vRAM entitlement is in compliance, and that you still have adequate CPU capacity. Even before this "feature," managing the infrastructure was a challenge. Now you are facing a much tougher task!
So to keep your QoS "in the box", think "outside the box"! :) Look at your environment holistically, across the board, across all the dimensions of the IT stack. VMTurbo's intelligent workload management solution does just that - assuring application performance, while doing so most efficiently, within all the dynamic constraints of the workloads and business priorities.
So, a way to address the new licensing challenges is to use VMTurbo to:
Identify VMs that are overprovisioned with respect to vRAM - and may thus contribute to a vRAM entitlement violation;
Identify the exact amount of vRAM overallocation, based on the VM's profile;
Recommend (and execute) changes to bring the infrastructure within vSphere 5.0 license compliance, while preserving the quality of service (QoS).
We invite you to try VMTurbo, and right-size your environment while assuring application performance!
So - take VMTurbo for a spin - it does all this (and a lot more!)