Vmworld Survey On Capacity And Performance Management, October 2008

Managing Virtual Capacity and Performance VMworld 2008 Survey Systar surveyed 205 attendees at VMware’s VMworld conference in September 2008 to understand key drivers for Capacity and Performance Management investments. After VMworld attendees completed the survey, Systar executives spent a few minutes with many of the respondents to learn more about the current state of their VMware environments, management maturity, growth plans, and operational challenges. Elements of these discussions are included in the key findings of the survey results below. Systar is a leading provider of Capacity Management solutions for the world’s largest enterprises. Systar Capacity Management solutions are branded under the OmniVision name with solutions targeted at management of physical distributed servers and virtualized environments, including VMware, PowerVM, and Solaris Zones. OmniVision solutions also assist our clients with Physical-to-Virtual (P2V) and Virtualto-Virtual (V2V) migration projects. Over 70,000 systems are managed worldwide. For more information about Systar’s OmniVision solutions for VMware, please visit: http://www.systar.com/products/omnivision/for_vmware

© 2008 Systar, Inc.

http://www.systar.com/solutions/virtualization_management

1) Characterize the pace of your VMware adoption.

Moderate: deployments are steady, and the environments is under control 63%

Too Slow: density is not meeting expectations, ROI questionable 14%

Too Fast: sprawl is apparent, quality of service may be at risk 23%

Source: Systar VMworld Survey, September 2008 (205 responses)

Key Findings: It was clear from our conversations with Senior VMware administrators and IT Managers at VMworld that nearly every company was planning to continue growing their VMware investments. Our survey results provide new insight into the pace of that growth around the VMware community and the challenges that come along with it. Nearly 1 in 4 respondents feel that their pace of VMware adoption is too fast. In these organizations, Virtual Machine sprawl may introduce new risks to the quality of service IT organizations can provide as visibility and control of the virtualized environment reaches beyond their grasp. For the “too fast” population, their survey results revealed that virtual machine growth was expected to average over 200% in the next 18 months. Populating new VMs on ESX servers may be the easy part of their growth plan, but ensuring quality of service remains high in the environment will be a key challenge. These IT organizations will need to look for solutions that enable clear visibility to the health, performance, and service quality of the virtualized environments where a top-down view of the systems is available. With the flexibility to share allocations of server subsystems (CPU, Memory, etc.), use of vMotion, and other administrative techniques, the role of the individual server has diminished, making way for the importance of clusters inside the data center. If too many VMs are loaded into a cluster, physical and virtual resource capacity might be constrained when workloads reach their peak. In the midst of sprawling VMware environments, constant visibility to available server capacity and its performance over time will be critical to meeting and maintaining quality of service objectives. For the “too slow” population, we found that a greater desire for increased VM density on ESX servers was an important objective. In our conversations with these VMware administrators and IT Managers, we found them reluctant to pile too many VMs onto an ESX server or cluster because they lacked sufficient analysis on the effect the consolidation. Even if the placement of VMs on a server was deemed successful at the start, they were weary of the lack of administrative visibility to the ESX server or cluster’s performance over time. They lacked an efficient way to measure capacity of the planned configurations and the ability to qualify that capacity over time as more VMs were added. The reluctance to add more VM’s was apparent through survey results showing VM growth for this population was only expected to be 50% over the next 18 months, compared to the 200% for the “too fast” population.

© 2008 Systar, Inc.

http://www.systar.com/solutions/virtualization_management

2) Prioritize areas of capacity management for virtual environments that are important for you. (Number from 1 – 5, highest priority =1) Ranking

Category

#1

Sizing VM requirements

#2

Sizing hosts or servers

#3

Ensuring acceptable response times

#4

Sizing the number of VMs per host

#5

Sizing clusters, pools or farms Source: Systar VMworld Survey, September 2008 (205 responses)

Key Findings: Over 38% of the surveyed population ranked sizing VMs, hosts and servers was their top priority for capacity management. These results tell us that most companies are at an early level of VMware management maturity, where the focus is still oriented toward individual servers or virtual machines. Comfort with configuring and managing capacity for individual server elements tends to mimic traditional physical server environments. In physical server environments, enterprise IT organizations have often maintained a policy of one server for one application. Through hundreds of our own conversations with large IT organizations and research findings from leading analysts, the projected the average utilization of today’s physical server environments generally falls between 10 - 20% thus the, general practice of over-sizing the infrastructure in relation to anticipated workload demands. In the virtualized environments, over-sizing a virtual machine may mean allocating resources that could be used by other workloads to increase overall utilization. Where traditional physical server environments enjoyed dedicated subsystem resources, VMware environments rely on sharing of the subsystem resources amongst numerous virtual machines inside an ESX server. Where rule-of-thumb estimates for sizing physical servers may have been the norm, the exercise becomes much more risky in virtual environments. If workloads on the virtual machines are relatively steady, and not “peaky”, capacity risks are minimized and quality of service is more predictable. As more virtual machines are added to an ESX server, behavior of the workloads can become increasingly unpredictable. It is clear from our survey results that the priority given to sizing VMs and ESX servers reflects the oversizing concern described above. Coming in a close third position, managing response times is a top priority. Clearly, priorities of the surveyed IT organizations were not focused on “virtualize and pray” approaches to capacity management. Respondents showed a strong desire to understand when capacity constraints in virtual server environments would begin to introduce risks to quality of service expectations of the end user. Priority rankings for sizing clusters, pools, and VMs per host, fell off dramatically after the top three responses. With the lowest priority focused on VMware clusters and pools, our conversations with VMworld 2008 attendees confirmed that these powerful and complex configuration options from VMware are not yet widely used.

© 2008 Systar, Inc.

http://www.systar.com/solutions/virtualization_management

3) Select your organization’s need for reporting performance and capacity metrics across the virtualized infrastructure.

Low 5%

Very High 29%

Medium 29%

High 37% Source: Systar VMworld Survey, September 2008 (205 responses)

Key Findings: One of the most striking survey results showed that 66% of the VMworld 2008 attendees surveyed considered capacity and performance metrics to be a “high” or “very high” need. The need for capacity management solutions in VMware environments dominates the landscape. As IT organizations flock to VMware for solutions that enable server consolidation programs (the #1 driver behind VMware growth according to the company), capacity management becomes a primary concern. If servers are consolidated too much, a densely populated VMware ESX server or an under-sized VM can quickly become saturated under peak workloads. Making the situation even more difficult, of those survey respondents describing a “high” or “very high” need for capacity and performance metrics, 78% claimed that they did not have access to best-in-breed management tools for VMware and that management processes are still evolving. The imbalance of high needs and lack of access to best-in-breed solutions combines to represent higher anxiety for IT organizations looking to better qualify the results of server consolidation projects or to quantify their company’s return on VMware investments to date. Before many large IT organizations had even formalized a basic capacity management methodology for their traditional physical server environments, they are being thrust into more complex and demanding virtualized environments. One affect of this quick transition is that best-in-class methodologies and skills for capacity management are not available, or do not transfer well to the virtualized server environments. Given the lack of methodologies and skills, IT organizations looking for best-in-breed capacity management solutions should expand their consideration beyond the most technically advanced solutions, to those solutions that are easy to deploy, learn, and use by non-capacity specialists. These organizations should also consider options that include both physical and virtual server environments to make the most of their investment. Although VMware is the dominate force in virtualized server environments today, selection of management solutions for capacity and performance metrics should not be limited to VMware alone.

© 2008 Systar, Inc.

http://www.systar.com/solutions/virtualization_management

4) Approximately how many ESX servers do you have? (round to nearest 10)

64 70

60

50

ESX Servers

35 40

30

20

10

0

# in production

# in test development

Source: Systar VMworld Survey, September 2008 (205 responses)

Key Findings: As many organizations make progress with their VMware investments, they are also interested to know how their peers are progressing. Where survey results from a few years ago would have shown significantly more VMware ESX servers deployed in test/development environments, the tide toward moving ESX servers into production has clearly turned. VMware ESX server populations in production environments outweigh numbers in test/development by nearly 2:1. Through informal conversations with VMworld 2008 attendees, we are told that the production servers are hosting Tier 2 and Tier 3 business applications, but there continues to be limited adoption of VMware for Tier 1 mission-critical applications. As comfort with VMware continues to grow along with the production environments, we would expect to see more Tier 1 applications infiltrating the VMware deployments. Along with the strong migration of VMware into production environments, VM density on ESX servers in production continues to be somewhat constrained. Although the average VMware administrator claimed nearly 20 VMs deployed per ESX server, 42% of those surveyed had deployed less than 10 VMs per ESX server. VM density under 10 per ESX server seemed to reveal itself as a “magic number” for administrators. With 10 VMs or below, IT managers and VM administrators commented that they could easily track and manage their environment’s health using the basic administrative capabilities of VMware’s vCenter. Once VM populations on servers grew beyond 10 or for those organizations looking to increase VM densities over the magic number, desire for capacity and performance metrics became increasingly important. For example, in environments where IT organizations had deployed 25 or more VMs per ESX server, the “high” or “very high” need for capacity and performance metrics increased to 77%. At companies with VM densities of 25 or more, it seemed obvious that the need for capacity and performance metrics would increase; what was more remarkable for these survey respondents was that only 30% had access to best-in-class management tools, skills, and processes.

© 2008 Systar, Inc.

http://www.systar.com/solutions/virtualization_management

5) Access your organization’s maturity to support virtualized servers.

High (best-in-class tools and processes, advanced skills) 22%

Low (some tools and process, novice skills) 19%

Medium (several tools and process, intermediate skills) 59% Source: Systar VMworld Survey, September 2008 (205 responses)

Key Findings: Where best-in-class management tools, skills, and processes have become commonplace in physical server environments, the migration of those elements has clearly not progressed into virtualized server environments. 78% of those surveyed believed they did not have access to best-in-class tools, processes, or skills in their virtualized server environments. Many VMworld 2008 attendees that we spoke with had quickly determined that their physical server management tools did not provide them management capabilities suited toward virtualized environments. When pursuing their need for capacity and performance metrics in physical server environments, many attendees claimed to rely on gathering huge volumes of raw performance metrics and applying unscientific rule-of-thumb or Microsoft Excel-based analytics – a process that was admittedly time and resource intensive, and unfortunately error-prone. When considering this same approach for determining capacity risks in their environments, most IT operations managers or VMware administrators admitted that the data about their VMs was either not available or nearly impossible to compile based on the scale of their virtualized environments.

© 2008 Systar, Inc.

http://www.systar.com/solutions/virtualization_management

Summary It is clear that a lack of capacity and performance management tools has not limited the excitement or significantly delayed VMware deployments. At the same time, VMworld 2008 survey results demonstrate a significant need for capacity and performance management solutions to help IT Managers and VMware administrators gain better control of their growing virtualized infrastructure, allowing them to maintain the expected quality of service for their end-users. With the lack of transferable capacity and performance management methodologies, skills, and tools from the physical server world, companies deploying large virtualized server environments will seek solutions that are easy to adopt, learn, and use. The best solutions will not only be required to deliver the required functionality, but must also incorporate processes that can easily be adopted by the IT organization. Where solutions require new skill sets, special training, or lengthy learning curves, VMware administrators will look for alternatives that are easier to employ. Automation of capacity and performance management initiatives in virtualized server environments will be a key requirement for IT Managers and VMware administrators. Physical server methods of capacity management that require manual collection and analysis of huge data sets will not translate into the virtual server world. In VMware and other virtualized server environments, data collection and analysis must be automated. Additionally, best-in-class solutions will move beyond presentation of basic utilization metrics for CPU, Memory, Disk, and NIC subsystems, and will begin to analyze true capacity saturation. Capacity saturation should be analyzed as a combination of raw metrics that describe not only when a virtualized system is heavily utilized, but where that heavy utilization is resulting in extended wait times for or contention between server resources. Automating this type of analysis will provide IT managers and VMware administrators visibility to where response times and overall quality of service is at risk. Finally, as virtualized server deployments continue to expand and administrators become more sophisticated in their ability to configure the powerful features offered by VMware, understanding of capacity will need to spread beyond the VM and ESX server to virtualized clusters and pools. Today’s most experienced organizations view capacity of their virtualized server environments, not from the bottom up, but from the top down. To recognize where capacity risks and opportunities exist, IT organizations will need to start from above the virtualized cloud and dive in, rather than begin their assessment from points inside the cloud itself.

© 2008 Systar, Inc.

http://www.systar.com/solutions/virtualization_management

About Systar Systar is a leading worldwide provider of performance management software. Systar’s OmniVision product suite enables customers to achieve the optimal alignment between IT resources and business requirements in both distributed and virtualized server environments. Systar’s proven capacity management solutions deliver the full benefits of virtualization by enabling customers to gain visibility into these complex environments, tune for optimal capacity, and move business-critical applications into production with full confidence. United States 8618 Westwood Center Dr. Suite 240 Vienna, VA 22182 Tel. +1 703-556-8400 Fax +1 703-556-8430 [email protected] France 171 bureaux de la Colline 92213 Saint-Cloud Cedex Tel. +33 (0) 1 49 11 45 00 Fax +33 (0) 1 49 11 45 45 [email protected]

Germany Mergenthallerallee 79-81 D-65760 Eschborn Tel. +49 211 598 8520 [email protected] Spain Centro de Negocios Eisenhower C/ Cañada Real de las Merinas, 17 Edificio 5 - 1º D 28042 Madrid Tel. +34 91 747 88 64 Fax +34 91 747 54 35 [email protected]

United Kingdom Systar Ltd Ground Floor Left 3 Dyer’s Buildings London EC1N 2JT Tel. +44 2072 692 799 Fax +44 2072 429 400 [email protected]

Systar, BusinessBridge, OmniVision, BusinessVision, ServiceVision, WideVision and Systar’s logo are registered trademarks of Systar. VMware, ESX Server, and all other brand names, product names and trademarks are the property of their respective owners.

© 2008 Systar, Inc.

http://www.systar.com/solutions/virtualization_management