DATACENTER INFRASTRUCTURE MANAGEMENT SOFTWARE. Monitoring, Managing and Optimizing the Datacenter

(1)

DATACENTER INFRASTRUCTURE MANAGEMENT SOFTWARE

Monitoring, Managing and Optimizing the Datacenter

As datacenters become bigger, denser and more complex, it is clear that the most adaptable, economically sustainable and eco-efficient facilities will be those using advanced infrastructure management software.

4 FINDINGS

• DCIM software is multifunctional, has many components, attempts to address numerous technical and business issues, and may consist of overlapping subsystems. This has made it difficult to define. PAGE 9

• We believe the leading DCIM products will evolve into frameworks or suites, which handle many functions equally well, so that it is no longer possible to identify core or primary functions.

PAGE 10

• The three main drivers of investment in DCIM software are economics (mainly through energy-related savings), improved availability, and improved manageability and flexibility.

PAGE 28

• Overall, we provisionally believe the DCIM market is worth roughly $240m in 2011, and will grow to $1.2bn in 2016. PAGE 34

5 IMPLICATIONS

• DCIM adoption, which is generally low, varies widely across different software subsets and datacenters.

Thus, overall adoption figures cited by vendors and analysts may be misleading. PAGE 22

• We believe it is difficult to achieve the more advanced levels of datacenter maturity, or of datacenter effectiveness generally, without extensive use of DCIM software.

PAGE 24

• Greater use of virtualization and changes in server design will result in increased volatility in power consumption – necessitating better integration of IT and infrastructure, and new investment in management and control software. PAGE 5

• The 451 Group describes 15 inhibitors to DCIM adoption, all of which can be overcome in most cases. The biggest are cost, functionality issues, the difficultly of creating/maintaining asset databases and commitment to simple in-house tools. PAGE 30

• We expect M&A activity to increase and accelerate to the point where a group of leaders emerge, most likely offering suites of products. PAGE 36

1 BOTTOM LINE

• The combined effect of the many structural and technological changes sweeping through the datacenter industry is that there will be much greater use of DCIM systems in the next five years, driving strong sales growth in the space.

Analyzing the Business of Enterprise IT Innovation

MAY 2011

DCT DATACENTER

TECHNOLOGIES

(2)

ABOUT THE 451 GROUP

The 451 Group is a technology analyst company. We publish market analysis focused on innovation in enterprise IT, and support our clients through a range of syndicated research and advisory services. Clients of the company — at vendor, investor, service-provider and end-user organizations — rely on 451 insights to do business better.

ABOUT TIER1 RESEARCH

Tier1 Research covers consumer, enterprise and carrier IT services, particularly hosting, colocation, content delivery, Internet services, software-as-a-service and enterprise services. Tier1’s focus is on the movement of services to the Internet — what they are, how they are delivered and where they are going.

© 2011 The 451 Group, Tier1 Research and/or its Affiliates. All Rights Reserved. Reproduc- tion and distribution of this publication, in whole or in part, in any form without prior written permission is forbidden. The terms of use regarding distribution, both internally and externally, shall be governed by the terms laid out in your Service Agreement with The 451 Group, Tier1 Research and/or its Affiliates. The information contained herein has been obtained from sources believed to be reliable. The 451 Group and Tier1 Research disclaim all warranties as to the accuracy, completeness or adequacy of such information. Although The 451 Group and Tier1 Research may discuss legal issues related to the information technology business, The 451 Group and Tier1 Research do not provide legal advice or services and their research should not be construed or used as such. The 451 Group and Tier1 Research shall have no liability for errors, omissions or inadequacies in the information contained herein or for interpretations thereof. The reader assumes sole responsibility for the selection of these materials to achieve its intended results. The opinions expressed herein are subject to change without notice.

Analyzing the Business of Enterprise IT Innovation

New York

20 West 37th Street, 6th Floor New York, NY 10018

Phone: 212.505.3030 Fax: 212.505.2630

San Francisco

140 Geary Street, 9th Floor San Francisco, CA 94108 Phone: 415.989.1555 Fax: 415.989.1558

London

37-41 Gower Street London, UK WC1E 6HH Phone: +44 (0)20.7299.7765 Fax: +44 (0)20.7299.7799

Boston

125 Broad Street, 4th Floor Boston, MA 02109

Phone: 617.261.0699 Fax: 617.261.0688

Better perspective from the top in independent tech research

(3)

SECTION 1 Executive Overview

1.1 INTRODUCTION

In its relatively short history, the global datacenter industry has never been stable. For most of the time, it has been expanding rapidly, and has adapted to wave after wave of technical innovation and commercial disruption. In the decade ahead, operators of datacenters can expect more of the same – much more.

A series of major technological innovations – coupled with significant external legisla- tive, economic and market disruptions – will increasingly dictate that managers rethink the way they plan, design and operate datacenters. These changes include a rapid increase in demand for datacenter capacity and services; the continued adoption of virtualization, dynamic provisioning and cloud computing; the deployment of modular and pre-configured datacenters; and the increasing integration of facilities and IT.

Datacenters are also more challenging than ever to manage. While business becomes ever more dependent on them, they are becoming more capital-intensive and energy-intensive, and increasingly technical. This has made datacenters more sensitive to operational management: critical, enterprise-endangering failures can be triggered by a faulty setting or a missed maintenance check, while hundreds of thousands of dollars in annual energy costs can be saved by a few small process changes and some light reconfiguration.

It is clear that the most adaptable, economically sustainable and best-managed datacenters will be those where managers have accurate and meaningful information about their datacenter’s assets, resource use and status – ideally from the lowest level of infrastructure up into the middle or higher echelons of the IT stack. They can use this information for planning, forecasting and management, for real-time decision-making, and, if prac- tical, to inform and drive automated systems.

This is the focal point and purpose of datacenter infrastructure management (DCIM) software, the subject of this report. Over the past decade, datacenters have been mostly managed using a loose collection of proprietary monitoring systems, custom-built software, and simple productivity tools such as Excel and Visio. This is all set to change with the advent and widespread adoption of new, powerful datacenter management tools.

(7)

1.2 KEY FINDINGS

• The combined effect of the many structural and technological changes sweeping through the datacenter industry is that there will be much greater use of DCIM software.

• DCIM software is multifunctional, has many components, attempts to address various technical and business issues, and may consist of numerous subsystems that appear to duplicate or overlap with other systems. All of this has made it very difficult to define.

• The 451 Group believes that the leading DCIM products will evolve into frameworks or suites, which carry out many functions equally well, so that it is no longer possible to identify core or primary functions.

• Monitoring and reporting systems will likely evolve to play a major role in the emerging dynamically controlled datacenter, where data from the IT and M&E systems must not only be collected and analyzed, but must be acted upon in near-real-time.

• The 451 Group believes that a full DCIM suite will be underpinned by two closely coupled operational databases – the asset management system, which holds detailed and accurate records about all the equipment, and the (real-time and historical) status reporting database.

• We believe there are substantial opportunities in the development of automated,

optimizing datacenters, but that this ‘control’ element will not begin to enjoy significant adoption until 2013 and beyond. This is largely due to product and market immaturity, issues of trust and a lack of proven reference sites.

• Most suppliers of DCIM software have up to now seen only moderate sales growth, although we believe the overall market is beginning to show more rapid growth.

• DCIM adoption varies widely across different software subsets and across different datacenters. For this reason, overall adoption figures cited by some vendors and analysts may be misleading.

• We believe it is difficult, if not impossible, to achieve the more advanced levels of the Green Grid Data Center Maturity Model – or, indeed, advanced levels of datacenter effectiveness generally – without extensive and committed use of DCIM software.

• Adoption of management products has been held back in the past because many customers have struggled to achieve effective deployment. But these products are now becoming more functional, more complete, easier to use and easier to integrate.

• Energy will rise as a proportion of overall datacenter costs. In some cases, there will also be energy scarcity. This will encourage DCIM investments that help to optimize energy use.

• Greater use of virtualization and changes in server design will result in more volatility in power consumption. This will require increased visibility into M&E by IT administrators, better integration of IT and infrastructure, and new investments in management and control software.

(8)

• Increased legislation and energy/carbon reporting are beginning to create a requirement for software and procedures to ensure compliance.

• The three main drivers of investment in DCIM software are economics – mainly through energy-related cost savings – improved availability, and improved manageability and flexibility.

• The 451 Group has identified 15 inhibitors to DCIM adoption, all of which can be overcome in most cases. The biggest of these inhibitors are cost, functionality issues, the difficultly of creating and maintaining asset databases, and commitment to simple in-house tools.

• Adoption of modular and container datacenters will have both positive and negative effects on the DCIM market. While professional, tight and dynamic management will encourage the use of more software, this software may be designed in and supplied directly to the manufacturers, reducing the market for third-party tools.

• Many datacenter software suppliers are still finding it difficult to identify potential customers, to get access to them, and, when they do, to price their products attractively.

• Overall, The 451 Group believes that the DCIM market is worth around $240m in 2011, and will grow to $1.2bn in 2016. These figures are provisional, and will be updated with more research.

• The 451 Group expects M&A activity in this space to increase and eventually accelerate to the point where a group of leaders emerge, most likely offering a suite of products.

• Unlocking budgets and developing a clear software market for the datacenter remains a challenge for DCIM vendors. The most successful suppliers will be those that can demonstrate a strong return on investment and make their case to senior corporate budget holders.

1.3 METHODOLOGY AND TERMINOLOGY

Scope of This Report

This report focuses on tools for datacenter monitoring, reporting, managing, planning and optimization. We examine the components and taxonomy of these systems, their role in the datacenter, the associated benefits, and the development of the overall market.

We are concerned primarily with independent products (independent of the equipment they are monitoring). This report identifies over 40 suppliers with more than 50 tools in this area. We include in this group a number of innovative companies that have developed power management technology for servers. We exclude, however, some specialist datacenter products that are not used for operational purposes, or are primarily intended for use beyond the datacenter infrastructure layer.

(9)

This report was prepared over several months by The 451 Group’s analyst team special- izing in datacenter technologies and eco-efficient IT. The report is primarily qualitative and is based on extensive interaction between The 451 Group, vendors of datacenter infrastructure management software, and operators and managers of datacenters.

Some of these operators and owners are members of the Uptime Institute’s Site Uptime Network or are subscribers to Tier1 Research services. Uptime and Tier1 are divisions of The 451 Group.

Although we have sent questionnaires to suppliers and have calculated some finan- cial figures based on our knowledge of the market, the data in this report is primarily intended for guidance, and should be treated with caution; figures should not be cited without The 451 Group’s approval. While we are very confident in our qualitative results and our analysis of market trends, we are less certain of quantitative forecasts such as market sizing and specific user adoption numbers.

This report was written by Andy Lawrence, Research Director, Eco-Efficient IT (andy.

[email protected]), with help and support from John Stanley, Analyst, Eco- Efficient IT ([email protected]). Any questions about the methodology of this report should be addressed to Andy Lawrence.

Other Information and Services from The 451 Group

Reports such as this one represent a holistic perspective on key emerging markets in the enterprise IT space. These markets evolve quickly, though, so The 451 Group offers additional services that provide critical marketplace updates. These updated reports and perspectives are presented on a daily basis via the company’s core intelligence service – the 451 Market Insight Service. Perspectives on strategic acquisitions and the liquidity environment for technology companies are updated regularly via the company’s forward-looking M&A analysis service – 451 TechDealmaker – which is backed by the industry-leading 451 M&A KnowledgeBase.

Emerging technologies and markets are also covered in additional 451 practice areas, including our Enterprise Security, Datacenter Technologies, Eco-Efficient IT, Information Management, Commercial Adoption of Open Source (CAOS), Infrastructure Computing for the Enterprise (ICE) and 451 Market Monitor services, as well as CloudScape, an interdisciplinary program from The 451 Group and subsidiary Tier1 Research. All of these 451 services, which are accessible via the Web, provide critical and timely analysis specifically focused on the business of enterprise IT innovation.

For more information about The 451 Group, please go to the company’s website:

www.the451group.com

(10)

SECTION 2 Datacenters in Transition

It is widely acknowledged that the economics, the operational practices and the underlying design principles of datacenters – and of IT service provision – are undergoing some fundamental, disruptive changes. These changes form the backdrop to any discus- sion of the market for datacenter software; we discuss some of them below.

2.1 ALTERNATIVE PATHS TO AVAILABILITY

In general, the biggest single concern of datacenter managers is maintaining availability. In many cases, the need to ensure uninterrupted service has pushed even finan- cial considerations into the background. The importance of availability has dominated datacenter design and led to, for example, the certified Tier Classification System for datacenters developed by Uptime Institute.

Approaches to maintaining availability are now changing, however. Some organizations – Google is among them – are adopting, or planning to adopt, new architectures that involve multiple datacenters that have a less resilient physical architecture. This is being made possible by advances in IT that enable rapid workload shifting between datacenters. Although at an early stage, this development will require datacenters to imple- ment software architecture that can identify problems and react rapidly. This will be especially difficult in datacenters that have mixed loads. (Google’s services are far more homogenous than most, enabling it to move work around relatively easily.)

2.2 ENERGY EFFICIENCY

Energy consumption has become a major issue for datacenters in recent years, and will continue to be so for many years to come. Research¹ has shown that operational energy costs are beginning to match the capital costs of servers. Energy consumption also drives up capital equipment costs as well (e.g., for generators and uninterruptible power supplies) and, furthermore, is a major contributor to carbon emissions. In some countries, carbon laws are adding to the cost of datacenter power, and are mandating reporting on carbon and energy use.

Most medium- to large-scale datacenters have embarked on energy-reduction strategies, leading many operators to label their datacenters as ‘green.’ This effort usually

1. Koomey, Jonathan, et al. 31 Mar 2008. A Simple Model for Determining True Total Cost of Ownership for Data Centers. Santa Fe, NM: Uptime Institute. Version 2.1, March 31, 2008.

Belady, Christian. 2007. “In the Data Center, Power and Cooling Costs More Than the IT Equipment it Supports.” Electronics Cooling. Vol. 23, No. 1, February 2007.

Environmental Protection Agency (EPA). 2 Aug 2007. Report to Congress on Server and Data Center Effi- ciency. Environmental Protection Agency (EPA). Available at <http://www.energystar.gov/ia/partners/prod_

development/downloads/EPA_Datacenter_Report_Congress_Final1.pdf>.

(11)

involves the introduction of technologies such as air economizers and strategies such as higher operating temperatures, which in turn require good environmental monitoring and management reporting. Energy savings is a driver for the introduction of DCIM (see Section 6).

2.3 DATACENTER COSTS

Datacenter costs – whether measured by square foot, by rack, or by kilowatt provided to IT – have risen steadily since the major switch over to standard x86 servers in the early 1990s. These costs have risen to the point where, in recent years, enterprises are examining almost any feasible strategy to avoid further datacenter buildouts – despite continuing and rising demand for their IT services.

Among the cost-containment strategies being adopted are the use of third-party datacenter colocation providers (see Section 2.4); the adoption of cloud services to meet excess demand; the construction of smaller, lightweight datacenters to meet excess demand; the use of containers; greater use of virtualization, coupled with the replacement and consoli- dation of IT equipment with smaller and less energy-hungry equipment; and more aggres- sive management of datacenter demand and capacity. This last strategy will, we believe, be widely adopted and will be partly enabled by much greater use of DCIM tools.

2.4 COLOCATION AND THE CLOUD

Rising costs, as well as growing technical complexity, skills shortages and the threat of ever-increasing compliance requirements, is weighing heavily on companies that are not primarily in the datacenter business. Managing datacenters is increasingly becoming its own specialized industry. This has led to growing adoption of colocation, as a means to avoid running datacenters directly, and to avoid capital outlays and the risks involved.

Those that continue to run their own enterprise datacenters are now being benchmarked against third-party providers whose core business is datacenter operations.

The growing adoption of cloud services owes more to the customer’s need for flexibility, agility and capacity than it does to cost reduction. The adoption of cloud services is one of the clearest and strongest trends in datacenters today. For those datacenters supporting cloud services, both public and private, this trend has introduced a multi- plicity of technical and management issues. The full impact of the cloud model on infrastructure design is not yet fully understood, although it is argued that datacenters need to be more modular, to abandon traditional raised-floor designs and be sensitive to volatile IT and heat loads. Increasingly tight integration between the physical and logical layers will also be required – as will more monitoring and control software.

(12)

2.5 FACILITIES AND IT CONVERGE

Datacenters are usually run by two distinct groups of people: those with responsibility for the M&E (mechanical and electrical) equipment, including building facilities, and those responsible for IT (servers, networking, storage). The two responsibilities come together either at the management level or in a commercial relationship sealed by a service-level agreement.

The need to reduce energy consumption and the need to optimize for cost and availability are forcing these disciplines to work together much more closely, or in some cases, be merged altogether. In this way, strategies to increase availability and reduce costs can be applied across the facilities and IT functions. Also, it becomes possible to optimize for energy consumption and cost by tightly linking demand and supply.

In the best-run datacenters, there is now much greater integration between the IT and M&E equipment and functions (although colocation companies, which may only provide the infrastructure and have no access to or control over the IT itself, do not always have this option). This will increasingly be seen both at the datacenter design and equipment design stages, in pre-configured designs, and in the integration of IT infrastructure software (such as IBM Tivoli) and DCIM software.

2.6 DESIGN CHANGES

Datacenters are always designed to ensure the highest level of redundancy and resilience, given the available budget and the requirements of the services they will support. Even the less resilient Tier 1 and Tier 2 datacenters incorporate some levels of redundancy and operational resilience.

The need to reduce energy consumption and reduce both capital and operating costs is now driving the adoption of new designs. One clear example is the use of modu- larized, granular designs (see Section 2.7). Datacenters are also being built with air economizers and evaporative cooling, and in some cases have no mechanical chillers.

Other technologies being deployed include decentralized UPS, unified computing devices (blade-type systems that incorporate integrated servers, network equipment and storage in one cabinet), and virtualized and power-managed servers. The adoption of many of these technologies will make the datacenter much more efficient, but also potentially more volatile and vulnerable to technical failures – so that closer monitoring and management systems are required.

(13)

2.7 PRE-CONFIGURED, MODULAR AND CONTAINER DATACENTERS

The modular approach to datacenter design and buildout, where each new module is only commissioned when there is sufficient demand, is now firmly established.

Although still at an early stage, many suppliers, designers and operators are taking this a step further, and productizing these modules, so that the datacenter can be constructed and powered in stages, on demand. These suppliers believe that many of the biggest problems in datacenters – such as high commissioning costs – can be resolved by the adoption of pre-configured, optimized designs and integrated datacenter modules, pods and containers.

These pre-configured and pre-tested designs may consist of integrated units of compute, storage and networking on the IT side, and cooling and power on the M&E side. The apparent economic and logistical advantages of pre-configured modular builds has brought many companies into this market (e.g., Panduit, Colt Technology Services Group, AST Modular and i/o Data Centers), which will now compete with or work with established vendors (e.g., Schneider Electric/APC, HP, Dell and IBM).

2.8 DCIM AT LAST

One net effect of all these disruptive changes is that we expect another clear trend in datacenter management to become firmly established: much greater use of automated asset, configuration and workload management, along with monitoring and analytics systems. This report refers to such systems as software for datacenter infrastructure management, or DCIM.

(14)

SECTION 3 Defining DCIM

DCIM is difficult to define precisely. It is multifunctional, has many components, attempts to address various technical and business issues, and may consist of numerous subsystems that appear to duplicate or overlap with other systems. In presenta-

tions, we have described DCIM as “a jigsaw puzzle with too many pieces,” while one supplier² likens DCIM to the Indian elephant, which, as the old story goes, appears to be completely different to six blind men who each felt different parts of the animal.

DCIM also goes by many names: datacenter efficiency software, operational technology, datacenter management (DCM) and datacenter operational management (DCOM). These terms are sometimes proprietary, inconsistently used, or not widely recognized. Several people have attempted to define DCIM, but for clarity in this report and for our future analysis, we apply our own definition – the first paragraph may be viewed as complete, while the following paragraphs serve to broaden the definition.

3.1 DCIM DEFINITION

A datacenter infrastructure management system collects and manages information about a datacenter’s assets, resource use and operational status. This information is then distributed, integrated, analyzed and applied in ways that help managers meet business and service-oriented goals and optimize the datacenter’s performance.

In practice, DCIM systems may vary widely in focus, and complete DCIM offerings are likely to consist of a framework or suite of products, from one or many suppliers, that are designed to interoperate with or complement each other.

The close interworking of IT and mechanical/electrical systems will increasingly lead to the deployment of DCIM systems that span datacenter facility infrastructure, physical IT assets and virtual IT assets.

DCIM systems may be particularly effective at helping managers to adapt to technical and business change more easily; to reduce waste and unnecessary over-provisioning;

to plan investments and new capacity; to reduce risk of failure; and to optimize energy consumption.

2. No Limits Software white paper, “The Datacenter Management Elephant,” by David Cole.

(15)

3.2 DCIM FUNCTIONS AND COMPONENTS

In the descriptions below, we illustrate some of the functional components of a DCIM system. These functions are:

1. Data collection (meters, sensors)

2. Environmental monitoring and reporting 3. Asset, configuration and change management 4. Power/energy measurement and modeling

5. Power management, power scaling, power capping 6. Data management, integration and reporting

7. Capacity planning, forecasting, simulation and analytics 8. Optimization, operational BI, load management

In order to build an effective, functional DCIM system, it is necessary to have several of these functional components, and for them to interoperate. It may also be necessary for these components to integrate with other systems, notably the building management system (or facility environmental system) and one or more IT service management systems.

Of course there are various ways to divide up these functions. It is also clear that commercial software products rarely address only one of these areas – it is more common for them to serve between two and six of these functions.

One supplier, Schneider Electric³, proposes categorizing products by their primary function and their secondary function (of which there may be several). This is a good approach that reflects today’s market. However, in our view, leading DCIM products will evolve into frameworks or suites, which carry out many functions equally well, so that it is no longer possible to identify core or primary functions. In this way, development will follow the example of ERP suites, which initially focused on finan- cial and physical assets, but which are now comprehensive, organization-wide business management suites with no discernible primary or secondary functions.

3.3 THE DCIM STACK

Figure 1 lists 10 functional areas that fall into the DCIM framework. Of the 10, the two that are colored differently (BMS and IT service management) are not part of the core stack, but are likely to be very closely integrated with it. Also, there is some debate over whether power management and power capping should be included in the DCIM stack.

3. “Classification of Data Center Operations Technology Management Tools,” APC Schneider Electric White Paper 104, by Kevin Brown and Dennis Bouley.

(16)

FIGURE 1: THE DCIM STACK

Source: The 451 Group

For an alternative view of the DCIM stack, with closer mapping to the datacenter layout, see the Appendix of this report.

Capacity panning, forecasting,

simulation, analytics Optimization, operational BI, load management

Data management, integration and reporting

Data collection, meters, sensors Cooling

control, BMS, alarms

etc

Environ- mental monitoring

and reporting

Asset configuration

and charge management

Power, energy measuring,

modeling

Power man- agement,

power capping

IT service and systems management,

VM mgt

(17)

SECTION 4 Core Components of a DCIM System

4.1 DATA COLLECTION, METERS AND SENSORS

In order to collect information about the operational status of a datacenter – whether it relates to power, temperature, water use or humidity – or even to verify that a system is where it is supposed to be, it is necessary to have in place sensors or meters designed to capture this data.

One important function of DCIM is to efficiently collect, integrate and convert or normalize the necessary data. This is especially important in datacenters where the deployment of sensors can be expensive or difficult, and where the data itself may not be easily read without protocol conversion or normalization into agreed formats. For example, a device may provide access to real-time power draw, but have no means of delivering power consumption data at set intervals for use by analytics and trending tools. Different meters may use different protocols (Bacnet, SNMP, IP, etc.). Or it may be necessary to have redundant sensors, or to put sensors on power sources that are separate from the IT load.

For all these reasons, the collection of data may require a separate subsystem, possibly using wireless sensors and meters and data-conversion devices that can make sense of a variety of data sources.

Examples of suppliers in this category include SynapSense (wireless sensors), Sentilla (wireless devices and data-collection software) and Modius (protocol and data conversion, and normalization of data). All of these vendors, however, increasingly also play at higher levels in the DCIM stack, where added value – and margins – are greater.

Wireless devices are particularly flexible because they can be easily set up and moved, if necessary. The wireless sensor approach can be very granular, allowing for per- device metering. This can be more useful than, for example, metering an entire circuit supporting many devices.

4.2 ENVIRONMENTAL MONITORING AND REPORTING Environmental monitoring and reporting is closely tied to data collection, since the former cannot work without the latter. Many products address both areas together. In order to be of use, data from the sensors (discussed above) needs to be collected, normal- ized and displayed. It may be used for various other purposes – for example, comparative analytics, trending and recognizing patterns for preventative maintenance.

(18)

Environmental monitoring and reporting systems may be loosely split into two groups, at least in terms of origin. The first group consists of products that are often proprietary (less open), that monitor in real time or with greater frequency, and that are designed for immediate alerting. Such products are likely to be based on a hardware device, and to be directly connected to their own sensors (not over an IP network, for example). The software may lack management-friendly interfaces, but instead focus on technical data. Examples here are Liebert’s SiteScan and Schneider/APC’s NetBotz.

The second group, which encompasses a newer generation of products, is more open and is intended to provide insight into the datacenter by extracting and reporting data, usually at less frequent intervals and, more probably, over an open network. At the heart of these products is a standard method for collecting data, a database and a flexible reporting system.

Among the devices that can be linked into the system are generators, uninterruptible power supplies (UPS), switches, power distribution units (PDUs), cooling equipment, fire- and leak- detection equipment and, increasingly, any IT equipment such as servers or storage.

Monitoring and reporting systems will be increasingly used in modern datacenters as a central integration point alongside the asset management system. Their value lies in their ability to create models and to link multiple types of data (such as heat, humidity and temperature information) from multiple sources. They are particularly important for feeding data into configuration and capacity management systems.

Monitoring and reporting systems do not need two-way interfaces, since they are not control systems. However, it is possible to use these systems for control if standard IT protocols are used. They may, therefore, play a big role in the emerging dynamically controlled datacenter, where data from the IT and M&E systems must not only be collected and analyzed, but also must be acted upon in near-real-time.

These systems are widely used for the collection and analysis of power data. It is, however, unclear if they will be widely used as primary systems for power modeling, measuring and reporting, or if separate, dedicated systems will continue to play an important role.

Among the suppliers in this area are Modius, OSIsoft, FieldView Solutions, Schneider/APC, Emerson Network Power, Sentilla and nlyte Software.

4.3 ASSET, CONFIGURATION AND CHANGE MANAGEMENT IT service managers have long used tools for asset, configuration, capacity and change management. These tools are designed to give administrators an accurate, real-time view of all IT assets so that the data can be used for support, operational management and capacity planning. In larger organizations with mission-critical IT services, such tools are considered essential.

(19)

Despite datacenter managers’ strong focus on service and availability, these software tools are not yet widely used to manage infrastructure equipment in the datacenter, nor are ITIL⁴-like practices widely followed (although new codes of management practice are being introduced⁵). There is, however, growing interest in and adoption of these tools. The key components here are configuration/asset management and change management.

A configuration/asset management tool is a centralized system that stores detailed information about the physical equipment and IT hardware in the datacenter. (It may be linked to a similar system for tracking virtual assets running on the IT equipment, such as virtual machines or application instances). This is a detailed electronic asset registry that may store data on operational characteristics and limitations, power consumption, exact physical location, purchase and maintenance history, dimensions, weight, number and type of connections, heat output, and position in the power chain. In some cases, these systems may be linked to, or include, a maintenance management system (such as IBM’s Maximo, which stores asset data as well as recording the service history of the equipment).

All of this information is extremely valuable, not only for efficient day-to-day running of the datacenter, but also for longer-term capacity planning (see Section 4.7). There are also considerable benefits derived from linking these systems to equivalent IT asset and configuration management systems (see Section 4.8). The systems can help staff, for example, see where equipment is, where it can be placed, and what power and cooling is available at those positions. Such is the value of maintaining asset databases that many DCIM products now include lightweight asset lists, even when this is clearly not their primary focus. Almost all DCIM tools will include some kind of integration point with established products, such as Emerson Network Power’s Aperture.

The more advanced of these asset database systems also provide a workflow or process- flow element, designed to facilitate and track changes, such as the deployment and movement of physical assets – this is usually known as change management. Asset management systems are best accompanied by a change management system, so that information is always up to date. A change management system can schedule and track when changes are made, such as the introduction of a new server, and ensure that the impact on power and cooling are recorded.

Asset management and change management systems may incorporate a capacity management capability. This tool will alert users to issues such as power mismatches or insuffi- cient capacity when they are making equipment changes. Even without the use of real- time data (from the monitoring systems), this feature should be able to identify whether an area of the datacenter is over-cooled or under-cooled, for example, or if power delivered to certain racks exceeds the amount that will be required.

4. ITIL (IT Infrastructure Library) is an internationally accepted, widely used and certifiable methodology for IT service management.

5. The Uptime Institute has introduced Operational Sustainability, a detailed set of practices for running data- centers effectively. Practitioners and datacenters may be certified. See www.uptimeinstitute.org.

(20)

The key suppliers in this area are Emerson/Aperture, with its existing products and emerging Trellis suite; Schneider/APC, with its InfrastruXure capacity/change manager;

and nlyte Software. Several other vendors also have credible products here.

4.4 POWER, ENERGY MEASURING AND MODELING

While the management of power quality and the power chain has always been important to datacenter managers, the close monitoring of power consumption is a more recent activity. This is due to limited power availability at many datacenters, and thus the need to reduce waste; the rising cost of power as a percentage of overall costs;

and growing environmental concerns over carbon emissions.

The monitoring and management of power is a core DCIM component function;

although environmental monitoring and reporting systems are mostly capable of tracking power/energy efficiency, power management may be seen as an entirely separate discipline in the datacenter, meriting its own tools. There are many alternative approaches to DCIM where power tracking is a primary activity, rather than a secondary one attached to environmental monitoring systems.

We see three areas of activity in DCIM power monitoring, none of which are exclu- sive of the others:

1. Power quality monitoring and analytics 2. Independent power monitoring

3. IT-based server power modeling

Power quality monitoring and analytics – Datacenters are mission-critical facilities with a substantial investment in power distribution equipment and infrastructure. A

‘power chain’ in a datacenter will include backup generators, high-voltage switch gear, power surge and harmonics protection, transformers, UPS and PDUs. The main purpose of such products is to ensure the steady availability of sufficient high-quality power, to reduce electrical hazards, and to diagnose or predict problems.

While these systems clearly can play a role in efficient energy management, they have traditionally differed in several ways from those higher-level power monitoring and reporting systems that are primarily intended for management consumption (see below).

Subsystem analytics is focused on the power chain, monitors by the millisecond, and is intended to identify and report existing and potential problems, whereas the latter category uses sampled data and is intended to enable business and policy managers (or automated proxies) to make decisions about loads and power states, and to see trends.

Although it is far from a being a trend at present, some suppliers focused on datacenter power distribution and quality/availability are beginning to recognize the oppor- tunity to provide datacenter managers with efficiency and capacity information as

(21)

well as lower-level technical monitoring. These include power-chain modeling specialist Power Analytics (formerly EDSA), PDU supplier Racktivity and branch circuit monitoring supplier TrendPoint Systems. Products that are intended for real-time power quality monitoring and analytics include PowerLogic from Schneider, Power Xpert from Eaton, Paladin from Power Analytics, and several others.

Independent power monitoring – Most of those suppliers involved in datacenter energy monitoring are primarily concerned with energy efficiency, the amount of power available and cooling. The primary goals are cost savings and compliance, with availability a secondary concern.

The simplest way to capture power data is to measure or model electricity use at a number of key tactical points, and then to feed that data into a simple power reporting system.

This may also be the environmental reporting system. There are many ways and places to measure power: from PDU and power-strip meters, on-board server meters, branch circuit monitoring, and UPS input or output. The best way to meter power will depend on budget, accuracy requirements and specific purpose. Once this data is collected, the reporting software can display power usage effectiveness (PUE) ratios, or show where power is being used against where it might be expected to be used.

Among the suppliers that are involved in this area are nlyte, Emerson/Avocent, Schneider, Sentilla, TrendPoint, ServerTech and many others.

IT-based server power modeling – A more recent approach is to model power consumption based on IT device demand. Using this approach, data on utilization and activity is collected from the IT equipment (servers, storage, networking) at regular intervals to create near-real-time readings as well as historical readings. This data is then used to extrapolate power use by referencing a central, independent, regularly updated repository of power utilization information for various equipment models.

The advantage of this approach is that it is possible to build a relatively quick, tactical and low-cost view of power use in the datacenter – taking hours or days, rather than weeks or months. The data from the modeling approach can be useful for analytics and trending, and for identifying mismatches of server utilization against power use. However, the data may not be accurate enough to diagnose or resolve availability issues.⁶

Suppliers taking this IT-based approach include Viridity Software, 1E and Sentilla.

4.5 POWER MANAGEMENT AND CAPPING

Most analyses of the DCIM market would not include IT power management or power capping functionality. This is for two main reasons: first, it scarcely exists as a market

6. Suppliers such as Viridity and 1E say their energy consumption models are proving as accurate as on-board meters, or even more so. To independently verify this claim, a trusted billing-grade accuracy meter would be required to compare readings.

(22)

sector or a common practice – adoption rates are extremely low at present – and second, almost all the players in this space come from an IT background, rather than a facilities background, and their tools interact directly with the servers rather than the facilities equipment.

Our decision to include these functions within the scope of DCIM (at least for now) is based on three arguments. First, the management of power consumption in the datacenter is a datacenter management or facility function, not an IT management function – and these tools are most definitely used to manage power consumption. Second, DCIM is evolving toward a more holistic, integrated approach that includes controlling both demand and supply of power. IT power management involves controlling demand for power. Third, the objectives of power management and power capping are primarily to reduce datacenter costs and improve datacenter operational efficiency. They do little for IT outside the datacenter, unless there is end-to-end accounting for carbon and energy use at an organization.

Drawing on the definitions in our 2008 Eco-IT report on power management⁷, there are two types of demand-side IT power management: server power management and power capping. Server power management places servers into reduced-power states (i.e., ‘sleep’

states) according to policies and schedules, or in response to a drop in machine utilization – for example, at night when demand is low.

Arguably, the benefits of this are most evident when systems are tightly linked to the underlying M&E system/DCIM system. Cooling systems should be able to ramp down when servers are drawing less power, and they should know to ramp back up when servers begin to run hotter. This is especially important when systems are turned back on, as the latency in the cooling may allow servers to overheat and turn off during reboot. (We know this to have happened in some trials.) Example of suppliers active in server power management are Power Assure, 1E, CA Technologies and VMware.

Meanwhile, power capping is a related function. This involves setting power constraints on individual servers, or groups of servers, so that they cannot exceed power availability or preset business limits. This is usually done to ‘fit’ more servers into an environment with a given amount of power capacity. As with power management, the most effective systems will be those that cap power in response to an M&E issue, rather than work in isolation.

At present, the suppliers active in this area are the big server manufacturers (notably, HP, IBM and Dell), along with processor and operating system suppliers (Intel, AMD and Microsoft) – although we believe there is potential for smaller independent players to make their mark with management products. UK-based Concurrent Thinking and US-based Power Assure are companies that may develop products in this area.

7. Power Management: Monitoring IT Energy Use From the Desktop to the Datacenter

(23)

4.6 DATA MANAGEMENT, INTEGRATION AND REPORTING

A fully instrumented modern datacenter can produce a lot of data. One of the key roles of DCIM software is to filter, consolidate, manage, organize and present this data in such a way that it is actually useful, and so that managers do not become overwhelmed.

This function is not likely to be carried out in isolation; it is clearly bound up with the data- collection and status-monitoring functions (below it in the DCIM stack), and with BI, simulation and analytics, which we view as an advanced application using the consolidated data.

Most DCIM systems will incorporate a SQL database for storing the data, and will need various conversion tools, rules and protocols for collecting the data. Setting these up in an open, interop- erable way is not a trivial task. For example, energy consumption may be collected from dozens or even hundreds of devices, from dozens of different device types, in multiple formats, and at different intervals. Each of these will have to be stored both separately, for granular reporting, and aggregated, for management reporting.

We believe that the DCIM databases can be likened to an enterprise service bus (ESB), used in IT for collecting and normalizing data from multiple sources, mostly using Web services protocols.

Increasingly, it is a requirement of DCIM systems to be able to interoperate with other software, such as IT service management systems (IBM Tivoli, for example), or with open dashboards. This is likely to be best achieved using Web services interfaces.

The value of the data that is collected can be seen in two main areas: 1) for management-level alerting, reporting and tracking, to spot emerging problems and to optimize for efficiency, and 2) for use in forward-looking capacity planning and simulation (see Section 4.7).

Management-level reporting means that the systems gather critical data and spot problems using multiple data sources. (Reporting metrics distilled from multiple data points are often called

‘derived metrics.’) For example, a DCIM system may spot an inconsistency such as high temperatures in an area of the datacenter with low IT load; or it may detect temperature spikes at the same time each day, which may suggest that there is a cooling latency problem. If IT systems are linked in, as they increasingly are, it may spot that utilization rates do not match power consumption used on a circuit or in a rack, suggesting that servers are doing too little.

The ability to provide clear, graphical, dashboard-type views of key performance indicators (such as PUE) is an important function of commercial products in this area. Increasingly, data from the operational reporting side is being integrated with asset management databases, or vice versa, so that the manager is able to both spot and analyze problems, and see exactly which equipment is involved, where it is located, what its operational characteristics are, and which circuit it is connected to. Ideally, it should be possible to view this graphically.

Ultimately, we believe that a full DCIM suite will need to include or be underpinned by two closely coupled operational databases: the asset management system, which holds detailed and accurate records about all the equipment, and the (real-time and historical) status reporting database.

(24)

Key suppliers in the area of data management, integration and reporting include nlyte, Schneider, OSIsoft, Modius and FieldView. Many others, however, are planning similar functions.

4.7 CAPACITY PLANNING, FORECASTING, SIMULATION AND ANALYTICS

DCIM systems today mostly look at the present status of the datacenter, for the purpose of improving operational efficiency and availability. But datacenter managers must also look forward – some of their biggest challenges are in avoiding huge cost overruns by over-provisioning, and avoiding becoming constrained operationally by a shortage of power, cooling or space.

Managers also struggle with assessing the impact of introducing new equipment into their complex and potentially unstable environments. Will replacing a chiller with an economizer really save energy? Will raising the temperature just push the energy consumption from the CRAC to the server fans? Will a forecast 10% increase in electricity prices justify the introduction of server power management?

Although capacity planning and simulation are perhaps not thought of by some as core DCIM functions, there is an emerging set of software tools designed to address these issues, and these will likely be increasingly integrated. These range from specialist tools with built-in algorithms and databases of best practices to add-on simulation tools that can be tested on the asset and reporting databases in DCIM tools.

Suppliers active in this area include nlyte, Schneider, Romonet and Lumina Deci- sion Systems (Analytica). There are also some free tools, such as DC Planner from Schneider, that can be used for basic datacenter planning.

4.8 OPTIMIZATION, OPERATIONAL BI AND LOAD MANAGEMENT

Most datacenters today are at the lower levels of the Green Grid Datacenter Matu- rity Model (explained in Section 5), and their use of software is mostly confined to reporting their current status and understanding where their assets are. The next step, which only some have taken, is for this data to be brought together into management databases, reporting systems and dashboards. Operators may also be able to run simu- lations and plan forward using this data.

The final step is for this information to be brought together to optimize and manage the datacenter in real time. This involves much greater integration with the IT load, and the IT management systems, than has hitherto been the case. It will also involve much more focus on active control, rather than passive data reporting and analysis.

(25)

For example, a DCIM system might be able to spot that a component is failing, and move VMs off affected servers until the problem is resolved; or it might be able to apply power caps on certain servers, knowing which applications will be affected, in order to avoid punitive demand-based power changes. Or it might be able to spot that one part of the datacenter is running cool, while another is hitting capacity, and move some workloads accordingly. The opportunities for doing this are potentially greater in heavily virtualized cloud computing environments, especially if multiple datacenters are involved.

We believe there are substantial opportunities in this area, but that this ‘control’ element will not begin to enjoy significant adoption until 2013 and beyond. This is largely due to product and market immaturity, issues of trust and the lack of proven reference sites.

4.9 RELATED SOFTWARE

In this report, we have focused on datacenter infrastructure management and its subsystems. There are, however, many other types of software that may be used in the datacenter, or that can be integrated with DCIM. Some of these play an important role, while others are marginal – and some may ultimately be subsumed into DCIM. These tools are summarized in Figure 2. We have listed these in order of their immediate connectedness to DCIM systems.

(26)

FIGURE 2: DCIM-RELATED SOFTWARE

CATEGORY FUNCTiON BACKGROUND/COMMENTS EXAMPLE SUPPLiERS/

PRODUCTS Building Management

Systems

Manages the cooling, ventilation, air quality, humidity and overall comfort levels of the building.

May also manage power consumption.

These systems are used to control the building, rather than the IT room, although there is close interaction. May gradually encompass energy management.

Automated Logic, Johnson Controls, Siemens, Honeywell, Schneider Electric

Datacenter Automation / Operations

Handles the day-to-day management of IT, and automation of routine datacenter tasks, such as runbook automation.

Designed to support efficient and reliable IT service management, these products are likely to interface with and extend into the infrastructure more.

BMC Software Remedy, HP BTO, IBM Tivoli, CA Datacenter Automation, Microsoft DCOM

Virtualization, VM Management, Provisioning

Allows workloads to be consolidated or to share underlying physical resources, reducing inefficiencies.

Virtualization enables dynamic provisioning and movement of workloads. By linking to DCIM, workloads can theoretically be moved to ensure availability, and unneeded servers can be powered down to save energy.

VMware (DRS, DPM and VMotion), Citrix Xen, Microsoft Virtual Server and related management products

Facility Security Manages and tracks access to areas of the datacenter, devices, etc.

Will be linked into DCIM dashboards, but is unlikely to be directly integrated.

Pelco, GE Security

Carbon Modeling /

Sustainability Reporting Enables managers to collect and analyze sustainability data, such as energy use, CO2, water, etc., using international reporting standards.

Mostly a function of environmental management systems, but some emerging products may focus on datacenters.

Enviance, Enablon, Hara Software, C3, SAP (Carbon Impact), CA Technologies (ecoGovernance)

Thermal Assessment Software

Software for heat-mapping datacenters and planning efficient use and placement of resources.

Semi-proprietary software used for computational fluid dynamics (CFD) mapping projects. Likely to be used more with sensor-based alternatives.

Future Facilities Ltd, Innovative Research Inc (TileFlow)

Modular and Container

Management Systems Datacenter management software that has been designed especially for pre-engineered or modular datacenters.

Container suppliers install sensors and controls for managing energy use and heat. They may use third- party software, or develop their own, to provide modular systems management.

HP, IBM, Dell, SGI, Colt, i/o Data Centers, Lee Technologies

Branch Circuit

Monitoring Hardware/software

combination that accurately measures power use at a circuit level.

Arguably a subsystem or feed-in component to DCIM, BCM is used by organizations, such as some colocation providers, that need to accurately map power use to circuits or groups of servers.

TrendPoint, Power Distribution Inc, Eaton, ServerTech

Enterprise Power

Management Integrates power use data from multiple sources (not just datacenters, but offices, etc.) and applies high-level controls based on policy, prices and service levels.

Emerging area, more an ambition than a product category.

Gradually extending into and interfacing with datacenter/server management systems.

Cisco EnergyWise, IBM Tivoli (using multiple products), Joulex

(27)

SECTION 5 DCIM and the Datacenter Maturity Model

One of the clear findings from our research into DCIM software is that adoption varies very widely across different subsets of software and across different datacenters. For this reason, overall adoption figures cited by some vendors and analysts may be misleading. This variable adoption is driven partly by business or compliance need (all datacenters need to have some form of environmental alerting); by software maturity (some capabilities and product types are relatively new); and, to a certain extent, by business type (colocation providers, for example, are less motivated to invest in some systems, because the IT is not their responsibility).

Ultimately, as organizations such as The Green Grid (with its Datacenter Maturity Model) have recognized, there is a long-term trend toward more efficient, dynamically managed datacenters. The Uptime Institute’s Operational Sustainability standard and the European Union’s Code of Conduct also embody the understanding that there are graded standards of effectiveness.

5.2 THE GREEN GRID DATACENTER MATURITY MODEL In early 2011, The Green Grid, an industry association promoting efficient datacenters and computing, introduced its Datacenter Maturity Model. This is a well-thought- out model for all aspects of datacenter operations. It has six stages of maturity – from the entry-level ‘minimal/no progress’ datacenter, which is firefighting, has no grasp of metrics and is highly inefficient; up to the advanced, visionary datacenter of the future, which is optimized, service-oriented and highly automated.

The Green Grid looks at the technology and management systems it expects to see at each stage of its model in the following areas: power, cooling, ‘other facility,’ management, compute, storage, network, and ‘other IT.’ The Green Grid’s Maturity Model is detailed and extensive, and it is not our intention to provide all the details here. They can be found at www.greengrid.org.

However, it is clear to us that at each stage of maturity, and in each area of the Green Grid model, more measurement, monitoring, data collection, management and optimization is required. The Green Grid Maturity Model may therefore be seen as a driver for managers to invest in DCIM software, just as ITIL, for example, has encouraged investment in IT service management systems.

By way of example, we reproduce here a small section of the Green Grid Maturity Model, relating specifically to monitoring and metrics. It shows that the requirement to measure, report and analyze increases with the efficiency level of the datacenter.

(28)

FIGURE 3: GREEN GRID MATURITY MODEL – MONITORING AND METRICS

MATURiTY

LEvEL MONiTORiNG XUE/ADDiTiONAL METRiCS*

(FOR EXAMPLE, PUE) 0 Monitoring or manual monitoring

not in place xUE not measured

1 Automated monitoring of key components in the datacenter

Basic xUE measured

2 Centralized and automated monitoring system inclusive of all mechanical, electrical and facility systems

Basic xUE measured, plan and actions in place for improvements

3 Centralized and automated monitoring system inclusive of all mechanical, electrical, facility and key IT systems

Advanced xUE measured, plan and actions in place for improvements

4 ‘Holistic’ monitoring capability across the datacenter – from source of power to chip performance

Advanced xUE measured, plan and actions in place for improvements; manual analysis/reporting of data to identify energy-saving opportunities

5 ‘Holistic’ monitoring capability across the datacenter – from source of power to business benefit of datacenter

Advanced xUE measured, plan and actions in place for improvements; automated analysis/reporting of data to identify energy-saving opportunities

Source: The Green Grid

*xUE is a term to denote usage effectiveness of any resource that a datacenter may use, such as power, water, reuse of heat, waste. PUE levels are described in the appendix.

Self-Optimizing and Autonomic Datacenters

⁸

It is clear that the higher maturity levels in the Green Grid model are beginning to describe a form of ‘self-optimizing, autonomic datacenter.’ This idealized ‘reference’

datacenter will be discussed in future 451 reports. Some of the attributes are as follows:

• The datacenter incorporates idealized virtual models, against which it constantly seeks to optimize itself.

• The datacenter is able to recognize technical problems as they emerge, self-diagnose most issues, and take steps to resolve them.

• The datacenter meets the highest industry standards for facility and IT efficiency and sustainability, most of the time.

8. Autonomic Computing refers to the self-managing characteristics of distributed computing resources, adapting to unpredictable changes while hiding intrinsic complexity to operators and users (Wikipedia).

DATACENTER INFRASTRUCTURE MANAGEMENT SOFTWARE. Monitoring, Managing and Optimizing the Datacenter