Reducing IT Operational Costs with IBM Power Systems Cloud Solutions
Annual IT operational costs continue to increase, with labor commanding a larger and largershare. For example, an IBM internal study of its own distributed infrastructure showed labor tobe over 60% of the total operational cost per year1, while industry analysts estimate labor costscan be as high as 80% of overall data center costs2.
Content
Reducing IT Operational Costs with IBM Power Systems Cloud Solutions
October 2010
Table of Contents
Table of Contents............................................................................................................... 2 Executive Summary........................................................................................................... 3 The Challenge of IT Labor Costs....................................................................................... 4 Overview of Private Clouds on Power Systems .............................................................. 7 Virtualization Benefits of the Power Systems Cloud Solutions ..................................... 9 Standardization Benefits of the Power Systems Cloud Solutions ............................... 13 Automation Benefits of the Power Systems Cloud Solutions ...................................... 15 Putting It All Together .................................................................................................... 17 Summary .......................................................................................................................... 18 Acknowledgements ......................................................................................................... 19
2
Executive Summary
Annual IT operational costs continue to increase, with labor commanding a larger and larger share. For example, an IBM internal study of its own distributed infrastructure showed labor to be over 60% of the total operational cost per year1, while industry analysts estimate labor costs can be as high as 80% of overall data center costs2. As a result, many customers are turning to private clouds, implementing such technologies as virtualization and consolidation, standardized workloads, and automation by way of self-service provisioning in an effort to reduce these costs. While only 12% of enterprises currently utilize some of these techniques, this number is expected to rise to 50% by 20123. IBM Power Systems cloud solutions utilize this three-pronged approach to significantly reduce IT labor costs in a private cloud environment. Quantifying the impact that private cloud technologies have on various aspects of labor, however, has proven elusive, resulting in slower adoption rates. Customers want to know, for example, just how much these cloud solutions will affect the labor required to set-up and maintain both the physical as well as virtual infrastructure for a given deployment platform before committing resources to their implementation. This paper describes an approach to help answer this question. We first examined the challenge of IT labor costs. We then introduced the key components of the Power Systems cloud solutions and how each tier of the private cloud stack fits within the virtualization, standardization and automation model. We proceeded to quantify the impact of virtualization by constructing a labor model that calculated the total labor hours required to set-up and maintain the overall infrastructure for both stand-alone x86 environments and virtualized Power Systems private cloud environments. Using actual customer data, the model allowed us to calculate the breakdown between the labor required for the physical and virtual infrastructure for a given number of workloads and specified time period. Next, we looked at the impact of standardization. We adjusted the model with a “clone” factor to reflect the fact that many companies implementing private clouds are using standardized workloads that can be easily copied or cloned to other virtualized servers to reduce labor costs even further. Finally, we took a look at automation by conducting a hands-on study to capture the time it took for an administrator to manually deploy an application onto a single Power Server versus using Tivoli Service Automation Manager in a Power Systems private cloud environment.
1 IBM Internal Consolidation Project 2 Source: Butler Group 2007 and http://www.itmanagement.com/blog/20070129/report-indicates-mainframe-adoption-continuing-to-grow/ 3 Internal IBM cloud study 2009
3
Following this overall approach yielded the following observations: • Private clouds built on IBM Power Systems servers and software, as well as Tivoli integrated service management software can reduce IT labor hours by up to 96%, translating to over $1 million savings over 5 years when compared with stand-alone x86 servers The greater the consolidation you can achieve, the lower you can reduce total physical server labor hours The more images you can standardize and clone, the lower you can reduce software image labor hours The use of Tivoli Service Automation Manager can reduce labor hours for a unique software image by up to 67% compared to manual deployment on an IBM Power Systems server
• •
•
The Challenge of IT Labor Costs
Continued business growth within an enterprise places ever-increasing demands on two key components of its information technology (IT) infrastructure – servers and storage. The costs associated with expanding and improving an IT infrastructure are typically divided into acquisition and operational costs. Acquisition costs include the purchase price and maintenance agreements for hardware and software. Operational costs include labor costs to deploy and manage hardware and software, as well as facilities charges such as power and cooling. While it is fairly straightforward for a business decision maker to compare the acquisition and facilities charges among various infrastructure proposals, judging the merits of an IT solution in terms of labor costs requires further analysis. Labor costs accrue throughout the IT infrastructure provisioning lifecycle. They begin at the procurement stage with planning and sizing. Once obtained, server assets must be configured, starting at the physical level and continuing with virtualization software, operating systems, middleware, applications, user access and security. While IT resources are in production, they must be regularly monitored, maintained and repaired, if necessary. Lastly, when computing resources no longer meet business requirements, they must be torn down and retired. Figure 1 below illustrates the server and storage provisioning lifecycle and focuses on the labor cost components:
4
Figure 1
When IBM examined its own distributed infrastructure, the study estimated that labor costs comprise 60% of all operational costs per year4, while industry analysis suggests that labor accounts for 80% of all datacenter costs5. For Power Systems environments, examples of labor costs include: • • • • • • • Planning for an upgrade to POWER7 processor-based hardware and estimating its effect on application and operating system levels Installing new Power servers, connecting them to Hardware Management Consoles (HMCs) and upgrading firmware Configuring Virtual I/O Servers (VIOS’s) to enable virtualization and I/O sharing Transferring existing or creating new definitions for all virtual servers, or logical partitions (LPARs) Provisioning storage for all virtual servers Installing the operating system (OS), middleware and applications, and configuring user accounts and security Monitoring processor utilization of virtual servers across multiple physical servers, or hosts Taking manual actions to move a virtual server away from an underperforming physical host via Live Partition Mobility (LPM)
•
4 IBM Internal Consolidation Project 5 Source: Butler Group 2007 and http://www.itmanagement.com/blog/20070129/report-indicates-mainframe-adoption-continuing-to-grow/
5
To help reduce IT labor costs, IBM has developed an approach based on virtualization, standardization and automation of the IT infrastructure, as depicted in Figure 2:
Figure 2
This paper will demonstrate how Power Systems cloud solutions use the three-pronged approach above to reduce labor costs in the datacenter. The discussion will estimate the labor cost reduction from managing 100 workloads running in a Power Systems based cloud environment when compared to managing 100 workloads running on individual physical servers, over a 5-year period.
6
Overview of Private Clouds on Power Systems
Let us first examine the key components of the Power Systems cloud solutions and how each one fits within the virtualization, standardization and automation approach, as shown in Figure 3:
Figure 3
A key starting point for a cloud computing environment on Power Systems is the high performance and density offered by POWER7 processor-based hardware and enterprise-grade storage, such as the IBM SAN Volume Controller, IBM System Storage DS5000, IBM System Storage DS8000 Turbo and IBM XIV Storage Systems. Both servers and storage utilize built-in virtualization technologies. The PowerVM Enterprise Edition hypervisor improves return on investment (ROI) and responsiveness to business changes with I/O sharing for disk and networking; dynamic CPU, memory and I/O movement without workload downtime; CPU assignment granularity down to 1/100th of a processor; and shared dedicated CPU capacity. Additionally, with PowerVM clients can leverage the additional benefits of Live Partition Mobility (LPM) for seamless hardware maintenance and migration, and Active Memory Sharing (AMS) and Active Memory Expansion (AME) for maximizing resource utilization and
7
performance across all virtual servers on a system. Storage virtualization with SAN Volume Controller enables consolidation and simplified management of heterogeneous disk subsystems, while meeting the goals of tiered storage, data integrity and high availability. The capabilities of Power Systems and IBM storage enable the necessary virtualization in the three-pronged labor cost reduction model. The next level up in the Power Systems cloud solutions stack consists of IBM Systems Director Enterprise Edition which includes the virtualization management capabilities of the IBM Systems Director VMControl Enterprise Edition plug-in. Systems Director is IBM’s cross-enterprise management tool, providing monitoring and management of Power Systems, System x, BladeCenter and System z physical virtual servers, storage, networking, software updates, energy consumption and many other components of the IT infrastructure. By consolidating many disparate management interfaces for servers and storage into the browser-based Systems Director console, clients can greatly reduce labor costs stemming from maintaining, opening and using multiple interface sessions. Systems Director also helps reduce labor costs by automating physical and virtual server monitoring and taking pre-defined actions when an event monitor is triggered. One example is of an event action triggered by a monitor is executing a remote command to an HMC to increase CPU allocation of an LPAR when CPU utilization in it spikes. VMControl provides the ability to create, monitor and dynamically change virtual servers on multiple platforms from the single Systems Director interface, by communicating with the PowerVM, x86 and z/VM hypervisors. Further, on Power Systems, VMControl automates and greatly reduces the time of new virtual server deployment. In the process of rapidly deploying new AIX and Linux virtual servers on Power, VMControl also automatically provisions the necessary storage for the new LPARs by communicating directly with San Volume Controller, for example. Lastly, VMControl also enables the creation of server and storage system pools. A server system pool comprised of Power servers is a group of physical Power servers and their LPARs, managed as a single computing resource. VMControl automates the placement of new virtual servers, or workloads, within the server system pool and allows the creation of resilience policies for workloads. A resilience policy can automatically relocate a workload to a less utilized physical server in the system pool if a condition specified by an administrator is met, such as a spike in average CPU utilization across a physical server. The capabilities of Systems Director Enterprise Edition with VMControl provide the image management and standardization in the three-pronged labor cost reduction model. The top level of the Power Systems cloud solutions stack includes the integrated services management application suite from IBM Tivoli. Tivoli software provides full service lifecycle management for business and IT including, application server and storage monitoring, usage and accounting management, as well as a self-service interface for end users and service catalog needed to provide a cloud service delivery model. While many Tivoli applications play an important role in the Power Systems cloud solutions stack, this brief overview focuses on the value and cost savings gained by the use of IBM Tivoli Service Automation Manager. Tivoli Service Automation Manager provides a common workflow automation engine for registering, configuring and deploying virtual servers on multiple hardware platforms. In a Power Systems cloud environment, Tivoli Service Automation Manager interacts directly with Systems Director and VMControl by using open, industry-standard Representational State Transfer (REST) interfaces in order to perform the deployment of a new virtual server as a managed service. Tivoli Service Automation Manager allows you to create and maintain a catalog of available
8
services and enables lifecycle management by tracking new service deployment as a project with start and end dates. Tivoli Service Automation Manager also provides a self-service browser-based interface that IT staff can make available to end users for requesting and using the defined managed services, while preserving system security with individual authorities for each user. Tivoli Service Automation Manager enables the automation in the three-pronged labor cost reduction model.
Virtualization Benefits of the Power Systems Cloud Solutions
To quantify the impact of labor, IBM has developed a labor model based on the server provisioning lifecycle, shown in Figure 4:
Figure 4
The formula represents the total labor hours ascribed to the management of a server environment as comprised of the hours spent managing a physical server over its lifetime plus the hours spent managing the software images over their lifetime. Total hardware server labor hours (H) include the set up and deployment hours representing one-time events such as sizing and configuring workloads, and testing of a physical computing element. They also include hours for scrubbing of servers, decommissioning, maintenance and troubleshooting for physical servers over the analysis period. Total software labor hours (S) include both the initial
9
installation labor associated with the software stack or virtual images on the physical server along with ongoing maintenance and troubleshooting over the assessment period. These tasks include periodic patching and upgrades, associated testing functions, analysis of errors, debugging, fixes, testing and reboots. Solving this equation for a stand-alone x86 environment gives us a picture of how much labor was required before virtualization. Similarly, solving the equation for the Power Systems environment gives us insight into the total hours needed after virtualization. Fortunately, we have data from customer case studies that can help us evaluate both equations as shown in Figure 5:
Figure 5
For the stand-alone x86 server case, this works out to be 30.7 servers per administrator or fulltime equivalent (FTE), while the virtualized Power Systems server case turns out to be 55.9 servers per administrator. Let us now calculate the portion of FTE labor needed to manage a single server or virtual workload. • Stand-alone x86 data shows 30.7 servers managed per FTE, 1/30.7= .0326 FTEs needed per server • Virtualized Power Systems data shows 55.9 virtual servers managed per FTE, 1/55.9 = .0179 FTEs needed per server
10
We can extend this analysis to calculate the number of FTE hours needed to manage 100 workloads over a period of 5 years for these two environments:
FTE hours needed to manage 100 workloads for 5 years: Multiply FTEs needed per workload * total hours over 5 years * number of workloads • • .0326 * 10,400 * 100 = 33,904 hours needed for all stand-alone x86 servers .0179 * 10,400 * 100 = 18,616 hours needed for all virtualized Power Systems servers
We assumed 2,080 hours or 52 weeks per year, 8-hour days for 1 year, resulting in 10,400 hours for 5 years. This analysis shows that a virtualized environment on Power Systems requires 45% less total labor hours to manage 100 Linux workloads for 5 years than a stand-alone x86 environment. It is also important to determine how much of the total management time can be attributed to managing the hardware (H) vs. managing the software images (S)? From our earlier analysis, it took 100 x86 servers to handle 100 Linux workloads. For the virtualized Power Systems case, the same 100 workloads could be handled on a single Power server. Thus, we are left with the following equations (over a 5-year period): (1) Stand-alone x86 (100Hi+100S) * 5 = 33,904 (2) Virtualized Power Systems (1Hp+100S) * 5 = 18,616 While the amount of time to install software on either a Power-based or x86 server is about the same, our own hands-on usage of a virtualized Power-based server showed that it took roughly twice the amount of hours to administer as a stand-alone x86 server. Thus, substituting Hp = 2Hi allows us to solve the equations for their respective H and S values: • • • Hi = 32 hours per year Hp = 64 hours per year S = 36 hours per year
Therefore, over a 1-year planning horizon, the total hardware labor (Hi) to manage one x86 server is 32 hours while the Power Systems server labor hours (Hp) requires twice that (64 hours). Over a 5-year period, the total hardware labor (Hi) to manage one x86 server is 160 hours while the Power Systems server labor hours (Hp) requires twice that (320 hours).The cost to manage a single software image (S) is 36 hours over 1 year and 180 hours over 5 years. We can use the above results to translate the reduction in labor hours through the use of virtualization into reduction of IT labor costs over a 5-year period. According to a May 2009
11
report from the US Department of Labor, Bureau of Labor Statistics6, the average national hourly wage for Network and Computer Systems Administrators was $34.10. Over a 5-year period, it would cost: • •
33,904 hours x $34.10/hour = $1,156,126.40 to manage a stand-alone x86 environment 18,616 hours x $34.10/hour = $634,805.60 to manage a virtualized Power Systems environment
Through the use of virtualization alone in a Power Systems environment, clients can reduce labor costs by over $0.5 million ($521,320.80) over five years when compared to a stand-alone x86 environment. Therefore, one of the levers in reducing labor costs is to reduce the number of physical servers you have to manage. Put another way, the more workloads you can consolidate on a given platform, the more you can lower your labor costs. This makes larger, more scalable systems like IBM Power Systems an ideal virtualization and consolidation platform for implementing private clouds.
6
http://www.bls.gov/oes/current/oes151071.htm
12
Standardization Benefits of the Power Systems Cloud Solutions
As we discussed in the overview section, a key value of cloud solutions is the ability to maintain and use a catalog of standardized software images, including the operating system, applications and patches. Systems Director with VMControl allows clients to maintain an image repository of virtual appliances, reducing the variability in the software images IT staff must manage. To quantify the benefits of standardization, we applied a cloning factor to our original equation as shown below in Figure 6:
Figure 6
Solving this equation for the virtualized Power environment discussed earlier in the paper yields the following: • 1Hp + 100(S/C) = total labor hours over a 1-year period
Since we already know Hp and S from our previous calculations, we can substitute those values, resulting in the following: • 1(64) + 100(36)/C = total labor hours over a 1-year period
Expressing the formula this way allows us to play some “what if” games with the clone factor (C) to gauge the impact of standardization on total labor hours. For example, applying a clone factor of five would mean that out of 100 workloads there are 100/5 or 20 unique images deployed, of which the rest are duplicates of the original twenty unique templates. When calculated over 5 years, this reduces the overall labor hours from the original virtualized Power Systems case of
13
18,616 to 3,920. Therefore, in this example, standardization of software images yields an additional labor cost reduction of 79% over a 5-year period. Taking standardization of images into account, we can make the following observations, as show in Figure 7:
Figure 7
Using the same cost model as in the previous section, we can calculate the additional IT labor cost savings through the use of standardization with a modest clone factor of 5 over a 5-year period: • • 18,616 hours x $34.10/hour = $634,805.60 to manage a virtualized Power Systems environment without standardization 3,920 hours x $34.10/hour = $133,672 to manage the same environment with software image standardization
Clients can further reduce labor costs by another $0.5 million ($501,133.60) over five years by using standardization. Another lever in reducing IT labor costs is the degree to which you can use workload standardization and cloning in your environment. Simply stated, the higher the clone factor, the greater the reduction in labor costs associated with deploying and maintaining software virtual images.
14
Automation Benefits of the Power Systems Cloud Solutions
As we have demonstrated, the virtualization and standardization benefits of Power Systems cloud solutions can significantly lower IT labor costs over a 5-year period. However, a key component of private clouds whose effect we have not taken into account yet is automation. Consider the typical scenario of a department in a company beginning a new project and therefore requiring one or more new workloads to be provisioned. For most clients the standard process involves submitting a ticket to the IT staff, which proceeds to manually deploy the workload. Automation with Power Systems cloud solutions can further reduce the labor costs associated with both of these steps through a user self-service portal and automatic image deployment. As we discussed in the overview section, those capabilities are provided by Tivoli Service Automation Manager. To help assess the extent to which the use of Tivoli Service Automation Manager can reduce labor hours, IBM has conducted a hands-on study, which tracked the time it took to deploy and instantiate an IBM WebSphere-based application on a virtual server using PowerVM.7 The results for a workload are shown in Figure 8:
7
A Study on Reducing Labor Costs Through the Use of WebSphere Cloudburst Appliance. http://www.ibm.com/software/webservers/cloudburst/cloudburst-library/
15
Figure 8
To summarize, leveraging automation provided by Tivoli Service Automation Manager, as part of the Power Systems cloud solutions, further reduces labor costs by 67%. Using the same cost model as in the previous sections, we can calculate the additional IT labor cost savings through the use of automation over a 5-year period: • • 3,920 hours x $34.10/hour = $133,672 to manage a virtualized Power Systems environment using standardization with a clone factor of 5 1,294 hours x $34.10/hour = $44125.40 with the additional use of automation with Tivoli Service Automation Manager as part of the Power Systems cloud solutions
Therefore, clients can further reduce labor costs by $89,546.60 over five years by using automation with Tivoli Service Automation Manager as part of the Power Systems cloud solutions.
16
Putting It All Together
As our analysis demonstrates, using virtualization, standardization and automation as part of a cloud solution on Power Systems can significantly reduce IT labor costs. Over a 5-year period, the benefits of virtualization can lower management hours from 33,904 to 18,616 (a 45% reduction), translated into an IT admin labor costs reduction of $521,320.80. Over the same period, the benefits of standardization can further lower management hours to 3,920 (a 79% reduction), with an IT admin labor cost savings of $501,133.60, while adding automation brings the total number of management hours to 1,294 (a 67% reduction), with an additional IT admin labor cost savings estimate of $89,546.60. The combined estimated labor cost reduction of managing 100 workloads with cloud solutions on Power Systems when compared with 100 physical servers over 5 years is 96% (33,904 to 1,294 hours) with estimate IT admin labor costs reduced a total of $1,112,001, as shown in Figure 9:
Figure 9
17
Summary
Escalating business requirements will continue to drive companies toward datacenter transformation. This includes pursuing ways to take costs out of their existing infrastructure through the use of virtualization, standardization, and automation. The labor model described in this paper can be used to estimate potential savings for a number of different deployment scenarios and technology choices. In our example, we chose to highlight the advantages of using cloud solutions on IBM Power Systems servers. Some of the benefits that can be expected include: • Private clouds built on IBM Power Systems servers and software, as well as Tivoli integrated service management software can reduce IT labor hours by up to 96%, translating to over $1 million savings over 5 years when compared with stand-alone x86 servers The greater the consolidation you can achieve, the lower you can reduce total physical server labor hours The more images you can standardize and clone, the lower you can reduce software image labor hours The use of Tivoli Service Automation Manager can reduce labor hours for a unique software image by up to 67% compared to manual deployment on an IBM Power Systems server
• •
•
18
Acknowledgements
IBM Systems Lab Services and Training
• • • Vess Natchev Mike Gordon Bob Minns
IBM Software Group, Competitive Project Office
• • • Scott Bain Fehmina Merchant John J Thomas
19
© Copyright IBM Corporation 2010
IBM Corporation Software Group Route 100 Somers, NY10589 USA Produced in the United States August 2010 All Rights Reserved IBM, the IBM logo, DB2, Tivoli, WebSphere and POWER7 are trademarks or registered trademarks of International Business Machines Corporation in the United States, other countries, or both. Other company, product or service names may be trademarks or service marks of others. The information contained in this documentation is provided for informational purposes only. While efforts were made to verify the completeness and accuracy of the information contained in this documentation, it is provided “as is” without warranty of any kind, express or implied. In addition, this information is based on IBM’s current product plans and strategy, which are subject to change by IBM without notice. IBM shall not be responsible for any damages arising out of the use of, or otherwise related to, this documentation or any other documentation. Nothing contained in this documentation is intended to, nor shall have the effect of, creating any warranties or representations from IBM (or its suppliers or licensors), or altering the terms and conditions of the applicable license agreement governing the use of IBM software. References in these materials to IBM products, programs, or services do not imply that they will be available in all countries in which IBM operates. Product release dates and/or capabilities referenced in these materials may change at any time at IBM’s sole discretion based on market opportunities or other factors, and are not intended to be a commitment to future product or feature availability in any way.
POW03052USEN-00
20
October 2010
Table of Contents
Table of Contents............................................................................................................... 2 Executive Summary........................................................................................................... 3 The Challenge of IT Labor Costs....................................................................................... 4 Overview of Private Clouds on Power Systems .............................................................. 7 Virtualization Benefits of the Power Systems Cloud Solutions ..................................... 9 Standardization Benefits of the Power Systems Cloud Solutions ............................... 13 Automation Benefits of the Power Systems Cloud Solutions ...................................... 15 Putting It All Together .................................................................................................... 17 Summary .......................................................................................................................... 18 Acknowledgements ......................................................................................................... 19
2
Executive Summary
Annual IT operational costs continue to increase, with labor commanding a larger and larger share. For example, an IBM internal study of its own distributed infrastructure showed labor to be over 60% of the total operational cost per year1, while industry analysts estimate labor costs can be as high as 80% of overall data center costs2. As a result, many customers are turning to private clouds, implementing such technologies as virtualization and consolidation, standardized workloads, and automation by way of self-service provisioning in an effort to reduce these costs. While only 12% of enterprises currently utilize some of these techniques, this number is expected to rise to 50% by 20123. IBM Power Systems cloud solutions utilize this three-pronged approach to significantly reduce IT labor costs in a private cloud environment. Quantifying the impact that private cloud technologies have on various aspects of labor, however, has proven elusive, resulting in slower adoption rates. Customers want to know, for example, just how much these cloud solutions will affect the labor required to set-up and maintain both the physical as well as virtual infrastructure for a given deployment platform before committing resources to their implementation. This paper describes an approach to help answer this question. We first examined the challenge of IT labor costs. We then introduced the key components of the Power Systems cloud solutions and how each tier of the private cloud stack fits within the virtualization, standardization and automation model. We proceeded to quantify the impact of virtualization by constructing a labor model that calculated the total labor hours required to set-up and maintain the overall infrastructure for both stand-alone x86 environments and virtualized Power Systems private cloud environments. Using actual customer data, the model allowed us to calculate the breakdown between the labor required for the physical and virtual infrastructure for a given number of workloads and specified time period. Next, we looked at the impact of standardization. We adjusted the model with a “clone” factor to reflect the fact that many companies implementing private clouds are using standardized workloads that can be easily copied or cloned to other virtualized servers to reduce labor costs even further. Finally, we took a look at automation by conducting a hands-on study to capture the time it took for an administrator to manually deploy an application onto a single Power Server versus using Tivoli Service Automation Manager in a Power Systems private cloud environment.
1 IBM Internal Consolidation Project 2 Source: Butler Group 2007 and http://www.itmanagement.com/blog/20070129/report-indicates-mainframe-adoption-continuing-to-grow/ 3 Internal IBM cloud study 2009
3
Following this overall approach yielded the following observations: • Private clouds built on IBM Power Systems servers and software, as well as Tivoli integrated service management software can reduce IT labor hours by up to 96%, translating to over $1 million savings over 5 years when compared with stand-alone x86 servers The greater the consolidation you can achieve, the lower you can reduce total physical server labor hours The more images you can standardize and clone, the lower you can reduce software image labor hours The use of Tivoli Service Automation Manager can reduce labor hours for a unique software image by up to 67% compared to manual deployment on an IBM Power Systems server
• •
•
The Challenge of IT Labor Costs
Continued business growth within an enterprise places ever-increasing demands on two key components of its information technology (IT) infrastructure – servers and storage. The costs associated with expanding and improving an IT infrastructure are typically divided into acquisition and operational costs. Acquisition costs include the purchase price and maintenance agreements for hardware and software. Operational costs include labor costs to deploy and manage hardware and software, as well as facilities charges such as power and cooling. While it is fairly straightforward for a business decision maker to compare the acquisition and facilities charges among various infrastructure proposals, judging the merits of an IT solution in terms of labor costs requires further analysis. Labor costs accrue throughout the IT infrastructure provisioning lifecycle. They begin at the procurement stage with planning and sizing. Once obtained, server assets must be configured, starting at the physical level and continuing with virtualization software, operating systems, middleware, applications, user access and security. While IT resources are in production, they must be regularly monitored, maintained and repaired, if necessary. Lastly, when computing resources no longer meet business requirements, they must be torn down and retired. Figure 1 below illustrates the server and storage provisioning lifecycle and focuses on the labor cost components:
4
Figure 1
When IBM examined its own distributed infrastructure, the study estimated that labor costs comprise 60% of all operational costs per year4, while industry analysis suggests that labor accounts for 80% of all datacenter costs5. For Power Systems environments, examples of labor costs include: • • • • • • • Planning for an upgrade to POWER7 processor-based hardware and estimating its effect on application and operating system levels Installing new Power servers, connecting them to Hardware Management Consoles (HMCs) and upgrading firmware Configuring Virtual I/O Servers (VIOS’s) to enable virtualization and I/O sharing Transferring existing or creating new definitions for all virtual servers, or logical partitions (LPARs) Provisioning storage for all virtual servers Installing the operating system (OS), middleware and applications, and configuring user accounts and security Monitoring processor utilization of virtual servers across multiple physical servers, or hosts Taking manual actions to move a virtual server away from an underperforming physical host via Live Partition Mobility (LPM)
•
4 IBM Internal Consolidation Project 5 Source: Butler Group 2007 and http://www.itmanagement.com/blog/20070129/report-indicates-mainframe-adoption-continuing-to-grow/
5
To help reduce IT labor costs, IBM has developed an approach based on virtualization, standardization and automation of the IT infrastructure, as depicted in Figure 2:
Figure 2
This paper will demonstrate how Power Systems cloud solutions use the three-pronged approach above to reduce labor costs in the datacenter. The discussion will estimate the labor cost reduction from managing 100 workloads running in a Power Systems based cloud environment when compared to managing 100 workloads running on individual physical servers, over a 5-year period.
6
Overview of Private Clouds on Power Systems
Let us first examine the key components of the Power Systems cloud solutions and how each one fits within the virtualization, standardization and automation approach, as shown in Figure 3:
Figure 3
A key starting point for a cloud computing environment on Power Systems is the high performance and density offered by POWER7 processor-based hardware and enterprise-grade storage, such as the IBM SAN Volume Controller, IBM System Storage DS5000, IBM System Storage DS8000 Turbo and IBM XIV Storage Systems. Both servers and storage utilize built-in virtualization technologies. The PowerVM Enterprise Edition hypervisor improves return on investment (ROI) and responsiveness to business changes with I/O sharing for disk and networking; dynamic CPU, memory and I/O movement without workload downtime; CPU assignment granularity down to 1/100th of a processor; and shared dedicated CPU capacity. Additionally, with PowerVM clients can leverage the additional benefits of Live Partition Mobility (LPM) for seamless hardware maintenance and migration, and Active Memory Sharing (AMS) and Active Memory Expansion (AME) for maximizing resource utilization and
7
performance across all virtual servers on a system. Storage virtualization with SAN Volume Controller enables consolidation and simplified management of heterogeneous disk subsystems, while meeting the goals of tiered storage, data integrity and high availability. The capabilities of Power Systems and IBM storage enable the necessary virtualization in the three-pronged labor cost reduction model. The next level up in the Power Systems cloud solutions stack consists of IBM Systems Director Enterprise Edition which includes the virtualization management capabilities of the IBM Systems Director VMControl Enterprise Edition plug-in. Systems Director is IBM’s cross-enterprise management tool, providing monitoring and management of Power Systems, System x, BladeCenter and System z physical virtual servers, storage, networking, software updates, energy consumption and many other components of the IT infrastructure. By consolidating many disparate management interfaces for servers and storage into the browser-based Systems Director console, clients can greatly reduce labor costs stemming from maintaining, opening and using multiple interface sessions. Systems Director also helps reduce labor costs by automating physical and virtual server monitoring and taking pre-defined actions when an event monitor is triggered. One example is of an event action triggered by a monitor is executing a remote command to an HMC to increase CPU allocation of an LPAR when CPU utilization in it spikes. VMControl provides the ability to create, monitor and dynamically change virtual servers on multiple platforms from the single Systems Director interface, by communicating with the PowerVM, x86 and z/VM hypervisors. Further, on Power Systems, VMControl automates and greatly reduces the time of new virtual server deployment. In the process of rapidly deploying new AIX and Linux virtual servers on Power, VMControl also automatically provisions the necessary storage for the new LPARs by communicating directly with San Volume Controller, for example. Lastly, VMControl also enables the creation of server and storage system pools. A server system pool comprised of Power servers is a group of physical Power servers and their LPARs, managed as a single computing resource. VMControl automates the placement of new virtual servers, or workloads, within the server system pool and allows the creation of resilience policies for workloads. A resilience policy can automatically relocate a workload to a less utilized physical server in the system pool if a condition specified by an administrator is met, such as a spike in average CPU utilization across a physical server. The capabilities of Systems Director Enterprise Edition with VMControl provide the image management and standardization in the three-pronged labor cost reduction model. The top level of the Power Systems cloud solutions stack includes the integrated services management application suite from IBM Tivoli. Tivoli software provides full service lifecycle management for business and IT including, application server and storage monitoring, usage and accounting management, as well as a self-service interface for end users and service catalog needed to provide a cloud service delivery model. While many Tivoli applications play an important role in the Power Systems cloud solutions stack, this brief overview focuses on the value and cost savings gained by the use of IBM Tivoli Service Automation Manager. Tivoli Service Automation Manager provides a common workflow automation engine for registering, configuring and deploying virtual servers on multiple hardware platforms. In a Power Systems cloud environment, Tivoli Service Automation Manager interacts directly with Systems Director and VMControl by using open, industry-standard Representational State Transfer (REST) interfaces in order to perform the deployment of a new virtual server as a managed service. Tivoli Service Automation Manager allows you to create and maintain a catalog of available
8
services and enables lifecycle management by tracking new service deployment as a project with start and end dates. Tivoli Service Automation Manager also provides a self-service browser-based interface that IT staff can make available to end users for requesting and using the defined managed services, while preserving system security with individual authorities for each user. Tivoli Service Automation Manager enables the automation in the three-pronged labor cost reduction model.
Virtualization Benefits of the Power Systems Cloud Solutions
To quantify the impact of labor, IBM has developed a labor model based on the server provisioning lifecycle, shown in Figure 4:
Figure 4
The formula represents the total labor hours ascribed to the management of a server environment as comprised of the hours spent managing a physical server over its lifetime plus the hours spent managing the software images over their lifetime. Total hardware server labor hours (H) include the set up and deployment hours representing one-time events such as sizing and configuring workloads, and testing of a physical computing element. They also include hours for scrubbing of servers, decommissioning, maintenance and troubleshooting for physical servers over the analysis period. Total software labor hours (S) include both the initial
9
installation labor associated with the software stack or virtual images on the physical server along with ongoing maintenance and troubleshooting over the assessment period. These tasks include periodic patching and upgrades, associated testing functions, analysis of errors, debugging, fixes, testing and reboots. Solving this equation for a stand-alone x86 environment gives us a picture of how much labor was required before virtualization. Similarly, solving the equation for the Power Systems environment gives us insight into the total hours needed after virtualization. Fortunately, we have data from customer case studies that can help us evaluate both equations as shown in Figure 5:
Figure 5
For the stand-alone x86 server case, this works out to be 30.7 servers per administrator or fulltime equivalent (FTE), while the virtualized Power Systems server case turns out to be 55.9 servers per administrator. Let us now calculate the portion of FTE labor needed to manage a single server or virtual workload. • Stand-alone x86 data shows 30.7 servers managed per FTE, 1/30.7= .0326 FTEs needed per server • Virtualized Power Systems data shows 55.9 virtual servers managed per FTE, 1/55.9 = .0179 FTEs needed per server
10
We can extend this analysis to calculate the number of FTE hours needed to manage 100 workloads over a period of 5 years for these two environments:
FTE hours needed to manage 100 workloads for 5 years: Multiply FTEs needed per workload * total hours over 5 years * number of workloads • • .0326 * 10,400 * 100 = 33,904 hours needed for all stand-alone x86 servers .0179 * 10,400 * 100 = 18,616 hours needed for all virtualized Power Systems servers
We assumed 2,080 hours or 52 weeks per year, 8-hour days for 1 year, resulting in 10,400 hours for 5 years. This analysis shows that a virtualized environment on Power Systems requires 45% less total labor hours to manage 100 Linux workloads for 5 years than a stand-alone x86 environment. It is also important to determine how much of the total management time can be attributed to managing the hardware (H) vs. managing the software images (S)? From our earlier analysis, it took 100 x86 servers to handle 100 Linux workloads. For the virtualized Power Systems case, the same 100 workloads could be handled on a single Power server. Thus, we are left with the following equations (over a 5-year period): (1) Stand-alone x86 (100Hi+100S) * 5 = 33,904 (2) Virtualized Power Systems (1Hp+100S) * 5 = 18,616 While the amount of time to install software on either a Power-based or x86 server is about the same, our own hands-on usage of a virtualized Power-based server showed that it took roughly twice the amount of hours to administer as a stand-alone x86 server. Thus, substituting Hp = 2Hi allows us to solve the equations for their respective H and S values: • • • Hi = 32 hours per year Hp = 64 hours per year S = 36 hours per year
Therefore, over a 1-year planning horizon, the total hardware labor (Hi) to manage one x86 server is 32 hours while the Power Systems server labor hours (Hp) requires twice that (64 hours). Over a 5-year period, the total hardware labor (Hi) to manage one x86 server is 160 hours while the Power Systems server labor hours (Hp) requires twice that (320 hours).The cost to manage a single software image (S) is 36 hours over 1 year and 180 hours over 5 years. We can use the above results to translate the reduction in labor hours through the use of virtualization into reduction of IT labor costs over a 5-year period. According to a May 2009
11
report from the US Department of Labor, Bureau of Labor Statistics6, the average national hourly wage for Network and Computer Systems Administrators was $34.10. Over a 5-year period, it would cost: • •
33,904 hours x $34.10/hour = $1,156,126.40 to manage a stand-alone x86 environment 18,616 hours x $34.10/hour = $634,805.60 to manage a virtualized Power Systems environment
Through the use of virtualization alone in a Power Systems environment, clients can reduce labor costs by over $0.5 million ($521,320.80) over five years when compared to a stand-alone x86 environment. Therefore, one of the levers in reducing labor costs is to reduce the number of physical servers you have to manage. Put another way, the more workloads you can consolidate on a given platform, the more you can lower your labor costs. This makes larger, more scalable systems like IBM Power Systems an ideal virtualization and consolidation platform for implementing private clouds.
6
http://www.bls.gov/oes/current/oes151071.htm
12
Standardization Benefits of the Power Systems Cloud Solutions
As we discussed in the overview section, a key value of cloud solutions is the ability to maintain and use a catalog of standardized software images, including the operating system, applications and patches. Systems Director with VMControl allows clients to maintain an image repository of virtual appliances, reducing the variability in the software images IT staff must manage. To quantify the benefits of standardization, we applied a cloning factor to our original equation as shown below in Figure 6:
Figure 6
Solving this equation for the virtualized Power environment discussed earlier in the paper yields the following: • 1Hp + 100(S/C) = total labor hours over a 1-year period
Since we already know Hp and S from our previous calculations, we can substitute those values, resulting in the following: • 1(64) + 100(36)/C = total labor hours over a 1-year period
Expressing the formula this way allows us to play some “what if” games with the clone factor (C) to gauge the impact of standardization on total labor hours. For example, applying a clone factor of five would mean that out of 100 workloads there are 100/5 or 20 unique images deployed, of which the rest are duplicates of the original twenty unique templates. When calculated over 5 years, this reduces the overall labor hours from the original virtualized Power Systems case of
13
18,616 to 3,920. Therefore, in this example, standardization of software images yields an additional labor cost reduction of 79% over a 5-year period. Taking standardization of images into account, we can make the following observations, as show in Figure 7:
Figure 7
Using the same cost model as in the previous section, we can calculate the additional IT labor cost savings through the use of standardization with a modest clone factor of 5 over a 5-year period: • • 18,616 hours x $34.10/hour = $634,805.60 to manage a virtualized Power Systems environment without standardization 3,920 hours x $34.10/hour = $133,672 to manage the same environment with software image standardization
Clients can further reduce labor costs by another $0.5 million ($501,133.60) over five years by using standardization. Another lever in reducing IT labor costs is the degree to which you can use workload standardization and cloning in your environment. Simply stated, the higher the clone factor, the greater the reduction in labor costs associated with deploying and maintaining software virtual images.
14
Automation Benefits of the Power Systems Cloud Solutions
As we have demonstrated, the virtualization and standardization benefits of Power Systems cloud solutions can significantly lower IT labor costs over a 5-year period. However, a key component of private clouds whose effect we have not taken into account yet is automation. Consider the typical scenario of a department in a company beginning a new project and therefore requiring one or more new workloads to be provisioned. For most clients the standard process involves submitting a ticket to the IT staff, which proceeds to manually deploy the workload. Automation with Power Systems cloud solutions can further reduce the labor costs associated with both of these steps through a user self-service portal and automatic image deployment. As we discussed in the overview section, those capabilities are provided by Tivoli Service Automation Manager. To help assess the extent to which the use of Tivoli Service Automation Manager can reduce labor hours, IBM has conducted a hands-on study, which tracked the time it took to deploy and instantiate an IBM WebSphere-based application on a virtual server using PowerVM.7 The results for a workload are shown in Figure 8:
7
A Study on Reducing Labor Costs Through the Use of WebSphere Cloudburst Appliance. http://www.ibm.com/software/webservers/cloudburst/cloudburst-library/
15
Figure 8
To summarize, leveraging automation provided by Tivoli Service Automation Manager, as part of the Power Systems cloud solutions, further reduces labor costs by 67%. Using the same cost model as in the previous sections, we can calculate the additional IT labor cost savings through the use of automation over a 5-year period: • • 3,920 hours x $34.10/hour = $133,672 to manage a virtualized Power Systems environment using standardization with a clone factor of 5 1,294 hours x $34.10/hour = $44125.40 with the additional use of automation with Tivoli Service Automation Manager as part of the Power Systems cloud solutions
Therefore, clients can further reduce labor costs by $89,546.60 over five years by using automation with Tivoli Service Automation Manager as part of the Power Systems cloud solutions.
16
Putting It All Together
As our analysis demonstrates, using virtualization, standardization and automation as part of a cloud solution on Power Systems can significantly reduce IT labor costs. Over a 5-year period, the benefits of virtualization can lower management hours from 33,904 to 18,616 (a 45% reduction), translated into an IT admin labor costs reduction of $521,320.80. Over the same period, the benefits of standardization can further lower management hours to 3,920 (a 79% reduction), with an IT admin labor cost savings of $501,133.60, while adding automation brings the total number of management hours to 1,294 (a 67% reduction), with an additional IT admin labor cost savings estimate of $89,546.60. The combined estimated labor cost reduction of managing 100 workloads with cloud solutions on Power Systems when compared with 100 physical servers over 5 years is 96% (33,904 to 1,294 hours) with estimate IT admin labor costs reduced a total of $1,112,001, as shown in Figure 9:
Figure 9
17
Summary
Escalating business requirements will continue to drive companies toward datacenter transformation. This includes pursuing ways to take costs out of their existing infrastructure through the use of virtualization, standardization, and automation. The labor model described in this paper can be used to estimate potential savings for a number of different deployment scenarios and technology choices. In our example, we chose to highlight the advantages of using cloud solutions on IBM Power Systems servers. Some of the benefits that can be expected include: • Private clouds built on IBM Power Systems servers and software, as well as Tivoli integrated service management software can reduce IT labor hours by up to 96%, translating to over $1 million savings over 5 years when compared with stand-alone x86 servers The greater the consolidation you can achieve, the lower you can reduce total physical server labor hours The more images you can standardize and clone, the lower you can reduce software image labor hours The use of Tivoli Service Automation Manager can reduce labor hours for a unique software image by up to 67% compared to manual deployment on an IBM Power Systems server
• •
•
18
Acknowledgements
IBM Systems Lab Services and Training
• • • Vess Natchev Mike Gordon Bob Minns
IBM Software Group, Competitive Project Office
• • • Scott Bain Fehmina Merchant John J Thomas
19
© Copyright IBM Corporation 2010
IBM Corporation Software Group Route 100 Somers, NY10589 USA Produced in the United States August 2010 All Rights Reserved IBM, the IBM logo, DB2, Tivoli, WebSphere and POWER7 are trademarks or registered trademarks of International Business Machines Corporation in the United States, other countries, or both. Other company, product or service names may be trademarks or service marks of others. The information contained in this documentation is provided for informational purposes only. While efforts were made to verify the completeness and accuracy of the information contained in this documentation, it is provided “as is” without warranty of any kind, express or implied. In addition, this information is based on IBM’s current product plans and strategy, which are subject to change by IBM without notice. IBM shall not be responsible for any damages arising out of the use of, or otherwise related to, this documentation or any other documentation. Nothing contained in this documentation is intended to, nor shall have the effect of, creating any warranties or representations from IBM (or its suppliers or licensors), or altering the terms and conditions of the applicable license agreement governing the use of IBM software. References in these materials to IBM products, programs, or services do not imply that they will be available in all countries in which IBM operates. Product release dates and/or capabilities referenced in these materials may change at any time at IBM’s sole discretion based on market opportunities or other factors, and are not intended to be a commitment to future product or feature availability in any way.
POW03052USEN-00
20
