State of server technology and operating systems: 2011

As server technology continues to change, so do the buying habits of IT professionals. In this special report, we examine what type of server technology and operating systems administrators are choosing for their environments and what factors have influenced their choices

Servers and operating systems are the foundation of a modern data center, and IT executives must balance the increasing

demands for computing resources and efficiency against the harsh realities of budget limits and technological shifts. Staying abreast of trends can be a serious challenge for IT professionals.

In early 2011, SearchDataCenter.com released the Data Center Decisions 2011 survey (DCD: 2011) to gauge trends and understand the factors that influence data center evolution in today's enterprise. We received over 1,000 responses from a range of professionals spanning numerous IT roles. This article, the first in a special report series, examines our readers’ decisions regarding server hardware and operating system choices, and compares how the landscape has changed since 2010.

Server hardware choices and deployment considerations
The last few years have seen an enormous emphasis placed on server consolidation, supported by the proliferation of 8-, 10- and 12-core processors. This means smaller servers are providing significantly more computing power than they did a few years ago, allowing IT administrators to operate more workloads efficiently on less hardware.

Server form factors. The choice of physical server form factors deployed in the data center reflects the growing influence of smaller and more powerful systems. In 2011, it's no surprise that rack-style servers compose the vast majority of server hardware. When asked which form factors are currently deployed, 71% of IT professionals use 2U servers, 63% use 1U servers, 51% use blade servers and 45% use rack form factors larger than 2U. Thirty-six percent of respondents deploy more classical form factors such as standalone towers and commodity computers. Mainframes, although relevant for very large organizations, are only used by 26% of respondents.

Server deployment. The number of workloads is trending upward, but the number of physical servers deployed in the principal data center is holding steady. In 2011, 54% of survey respondents reported running fewer than 100 physical servers, another 25% run 100 to 500 physical servers, 10% deploy 500 to 1,000 physical servers, and only 11% use more than 1,000 physical servers.

These physical server numbers are almost identical to 2010 survey results, with only a slight decline in large-scale deployments of physical servers year over year. This dip reflects continued interest in server consolidation and an increased use of infrastructure outsourcing, both of which can reduce the overall number of physical servers in the data center.

Server purchasing drivers. Various factors drive the acquisition of new servers for the data center. Table 1 shows some of the top reasons. In 2011, system lease, server life cycles and increasing computing requirements are clear drivers for new server purchases. Additional or replacement servers are unquestionably required to make up for displaced or decommissioned equipment. Consolidation is another major driver, and powerful new servers are often acquired to support more workloads running on the same box. New applications, such as a demanding new database or email platform, may need more resources than current servers can provide; this need pushes data centers to purchase new servers.

“It’s encouraging to see high demand driven by new applications, which indicate that IT departments are being asked to do new things,” said Michael Cote, a software industry analyst with Seattle-based RedMonk. “Virtualization-led consolidation is still a huge, ongoing project for the IT world, and we’ll see virtualization conversions going on for several years.”

Table 1 – Factors driving new server acquisitions in the data center

Factor 2011 results 2010 results

Existing servers reaching end of life or coming off lease 41% 41%
Normal increase in business computing capacity 41% 44%
Consolidation or enhancing virtualization capabilities 40% 44%
Support for new applications 32% 38%
Reduce floor space or facilitate consolidation 22% 26%
Handling redundancy or disaster recovery needs 18% No data
Standardize on one or more platforms 14% 15%
Reduce administrative workload 12% 13%
Improve server energy efficiency 12% 25%
Building or extending internal cloud computing 11% 9%
Meeting corporate compliance needs 9% No data
Rapid business growth 9% 8%
Building new (or renovating existing) data center 7% No data
Introducing converged hardware or unified computing 5% No data
Need for larger memory footprint 4% 7%
Need for faster I/O provided by PCIe 3% 3%

The top three drivers remain the same from 2010, though end-of-life issues took precedence over business capacity increases in the last year. Table 1 illustrates many other factors that drive server acquisitions, including support for new applications, disaster recovery, redundancy, platform standardization, the desire to improve energy efficiency, moving to private cloud computing, corporate compliance, normal business growth and so on.

Factors that influence server vendor selection
In previous years, IT organizations routinely selected server vendors based on the strength of the vendor’s relationship, pushing factors such as price, performance and support further down the list. Table 2 shows that those factors have undergone a substantial turnaround in 2011 as organizations refocus on price and performance.

Table 2 – Factors that influence the choice of server vendor

Factor 2011 results 2010 results

Reliability

26%

No data

Price and/or TCO

24%

17%

Existing relationship with vendor

15%

29%

Performance

10%

16%

Technical support and service

10%

15%

Executive or management decision

7%

No data

Product features and functions

4%

12%

Hardware is optimized for server virtualization

3%

5%

Client, customer or tenant decision

1%

No data

Financial stability/resistance to potential acquisition

1%

1%

In 2010, 29% of IT respondents cited existing vendor relationships as the biggest consideration in choosing a server vendor, and just 17% of respondents based their decisions on price or total cost of ownership (TCO). Another 16% of respondents pointed to system performance, while 15% cited technical support and service as the top reasons for selecting a vendor.

“We host a lot of our virtual infrastructure on several physical machines,” said Bill Kleyman, director of technology at World Wide Fittings Inc., an industrial supplier headquartered in Niles, Ill. “Partner and vendor relationships will always be important, but we need a stable system that we can rely on. Our IT team focuses on redundancy, reliability and failover capabilities of servers that we depend on daily.”

And DCD: 2011 survey data reflects Kleyman’s attitude. In 2011, 26% of IT respondents cited system reliability as the single biggest consideration when choosing a server vendor, with price and TCO a close second at 24%. Another 15% of respondents said an existing relationship with the vendor was important, 10% call for system performance, and another 10% want technical support and service from a vendor.

Blade server choices follow a similar pattern. In 2010, blade server performance was the most important factor for purchasing a blade server for 38% of respondents, while 26% were more concerned with price and 14% noted a need for energy efficiency. However, the tables turned in 2011, with just 29% of respondents selecting blade servers based on performance, 21% making the choice for reliability and service, 13% looking at price first and only 11% looking for a strong vendor relationship.

In short, IT organizations want servers that will work for the long haul, and there is confidence that even entry-level servers will meet performance demands. They're less concerned with which vendor they buy from, as long as the price matches their budget. This should be a wake-up call to server vendors that rely on relationships or lock-in tactics to generate recurring revenue.

“I see a lot of systems vendors trying to differentiate on hardware; building something into the actual servers that makes you want to pay a premium for them. But that’s been a tough sell because x86 are so highly commoditized,” Cote said, noting that software is increasingly the best way to realize value in IT.

Server budget changes. Companies are loosening their financial belts, freeing up budgets for the most desirable platforms. In 2011, some of the biggest budget increases appear in Windows-based servers; 86% of respondents noted data center spending to stay steady or increase while only 11% of respondents anticipate a decrease in Windows-based server budgets.

Budget changes were more modest for Linux-based servers in 2011, with 73% of respondents expecting no increase or only a slight increase in spending; only 7% of respondents expect a budget decline for Linux-based systems.

Data centers running Unix and mainframe systems, however, don’t anticipate budgets to swell in 2011. Eleven percent or fewer of respondents will see any budget growth, though most do expect to maintain their current budgets.

Choosing the actual server vendors. The economic turmoil of the last few years appears to be abating and certain IT budgets are showing signs of life, so the need to play catch up is evident. Delayed or deferred technology refresh cycles have dramatically shifted vendor selection criteria in 2011, and system pricing has customers rethinking vendor loyalty.

This trend is evident in respondents’ choice of vendors, moving from traditionally “upscale” vendors like Hewlett-Packard (HP) to more “commodity” vendors like Dell. In 2010, 35% of IT respondents selected HP as the primary server hardware vendor, with Dell a close second at 34%. Eighteen percent of respondents use IBM servers and only 5% reported running Oracle/Sun Microsystems. In fact, in the 2010 SearchDataCenter.com Products of the Year, HP servers took Gold and Dell nabbed Silver. Reliable servers with a growing commodity demand are making a bigger splash in the server market than most experts anticipated.

With DCD: 2011, Dell emerged as the largest primary server vendor with 38% of IT respondents, moving ahead of HP with 32% of respondents. IBM maintained its position with 18% of responses, though this is likely due to the company’s strong and steady role in the mainframe market. Interestingly, Oracle/Sun Microsystems slipped to 3% of responses, while Cisco Systems climbed onto the radar in 2011 with 3%.

The decisive gains Dell made in data center server adoption may reflect a combination of factors: aggressive pricing and quality/reliability improvements as the company begins to actively compete in the enterprise data center game.

Moving the data center toward an integrated infrastructure
An integrated infrastructure unites servers, networking and storage into a single cohesive computing system. Vendors like Cisco Systems established an early lead in integrated infrastructures with its Unified Computing System (UCS); however, HP is catching up with its Converged Infrastructure platform. Other vendors appear to be moving toward integration as well.

The promise of an integrated infrastructure is compelling. It means that one vendor provides the entire data center infrastructure, which is easier to scale and manage than a traditional heterogeneous data center. It's one-stop shopping, and creates one throat to choke with regard to support. But it’s not perfect; there are serious interoperability issues when trying to integrate equipment that may fall outside of the vendor's scope. Vendor lock-in is also an issue, Cote added. “The other side of one throat to choke is one big check to write.”

Slow adoption, but notable growth. Integrated infrastructure adoption is slow. DCD: 2011 reports that only 7% of respondents deploy Cisco UCS throughout the entire data center, another 7% report a partial UCS deployment, 15% are evaluating the technology and 11% expect to evaluate Cisco UCS in the coming year. However, 60% of respondents neither use nor are considering to use the technology. The numbers are similar for HP’s Converged Infrastructure as well as other integrated infrastructure technologies on the market.

Organizations that have adopted an integrated infrastructure cite some significant benefits that may increase its adoption in the coming years, as shown in Table 3. In 2011, the majority of IT respondents who have adopted an integrated infrastructure expect it to improve data center performance. Another large group of respondents want to save money or reduce the capital investment in the infrastructure, eliminate integration problems, simplify systems management or ease hardware maintenance and vendor support issues. [There is no comparative data for 2010.]

Table 3 – Benefits of integrated infrastructure technologies

Benefit 2011 results

Improved data center performance

66%

Save money or reduce capital investments in infrastructure

49%

Eliminate integration problems

44%

Save time and trouble with systems management

42%

Save time and trouble with hardware maintenance

39%

Improve vendor support (stop "finger pointing")

23%

Quality of relationship with the UCS vendor

12%

Vendor’s UCS product roadmap meets business’ needs

9%

In spite of some trepidation among enterprise data centers, integrated infrastructure is gaining relevance and deployments are expanding. DCD: 2011 reports that 53% of respondents that use integrated infrastructures today will invest in the same amount in the coming year, while another 35% plan to make additional purchases. Only 12% report a reduction in integrated infrastructure purchases in the next year.

Operating systems and workloads in the data center
Operating systems play a critical role in the data center to manage hardware resources, launch and manage applications, and provide the user interface(s) that configure and control the system. It's a combination of factors—application support, stability, security and maturity—that make OS choices as important as any hardware decision.

Kleyman suggests future OSes in the data center will offer more manageability. New Windows Server iterations will promise improved hardware support and more granular control over infrastructure-necessary features like remote management and redundancy, Kleyman noted.

Workloads in the data center. One way to understand how suitable an OS is for the data center is to think about workloads—current and potential. Table 4 outlines the major application categories for 2011 and 2010. Although the distribution of workload types has not shifted significantly in the last year, the three top application categories: databases, email and Web servers fall largely into Windows OSes. This suggests Microsoft Windows will be the OS of choice in the enterprise.

Table 4 – Application deployments across a data center

Application category 2011 results 2010 results

Database

83%

No data

Email

78%

73%

Web servers

76%

78%

LAN infrastructure services

73%

No data

Accounting

68%

65%

Business intelligence or data warehousing

52%

58%

Human resources

49%

48%

Collaboration software

43%

No data

Application virtualization

37%

37%

CRM-type applications

36%

36%

ERP-type applications

31%

38%

Online transaction processing

31%

47%

E-commerce

31%

32%

Supply chain and logistics

23%

24%

HPC-type applications

16%

No data

Manufacturing

11%

18%

Scientific

9%

No data

Microsoft operating systems. It's no surprise that Microsoft OSes represent the majority of installations in the data center, though organizations are slow to embrace version upgrades. In DCD: 2011 80% of IT respondents reported that Windows Server 2003 was deployed in their data centers, 67% noted Windows Server 2008 R2 and 60% had deployed Windows Server 2008.

The use of Windows Server 2003 remained steady from 2010 where 81% of respondents cited its use in the data center. However, 52% of respondents used Windows Server 2008, and just 45% of respondents used Windows Server 2008 R2. This means that although Windows Server 2003 is still widely deployed in production, the use of Windows Server 2008 R2 is poised to outpace the version prior to R2.

“We are upgrading our entire Citrix farm to XenApp 6 and focusing on the integration with Server 2008 R2, which is running in a virtual environment,” Kleyman said. “The performance from our tests shows … a definite ease of use and increased stability.”

Even though Microsoft adoption isn’t as pervasive for mission-critical workloads, mission-critical Windows Server 2008 R2 use has grown rapidly in the last year. In 2011, 63% of IT professionals noted that Windows Server 2003 was supporting mission-critical workloads, 60% reported Windows Server 2008 R2 and 47% run Windows Server 2008. In 2010, 68% of IT professionals reported using Windows Server 2003, 42% were on Windows Server 2008 and only 35% chose Windows Server 2008 R2.

Linux operating systems. Open-source OSes have a solid presence in many data centers, though adoption remains slow. The popularity of open source can be attributed to its free (or low-cost) availability, stability and relatively lightweight operation due to smaller resource demands on the server.

When DCD: 2011 asked which open-source OSes had a presence in the data center, a resounding 40% of respondents reported Red Hat Enterprise Linux, 16% noted Ubuntu Linux, 14% said CentOS and Novell SUSE Linux equally, and 12% reported using Oracle Enterprise Linux. The increased adoption of CentOS is perhaps the biggest move in open-source deployments in the last 12 months.

Overall, open-source operating system deployments have changed little over the last year, particularly with Red Hat Enterprise Linux. In 2010, 41% of IT staff said they use Red Hat Enterprise Linux, with 16% using SUSE Linux, 14% using Ubuntu Linux, 11% using Oracle Enterprise Linux, and just over 1% using CentOS.

In spite of its reputation for stability and reliability, Linux use drops noticeably in mission-critical environments. This is due, at least in part, to the lack of organized or professional support that users require in mission-critical data centers.

In 2011, just 29% of respondents said they use Red Hat Enterprise Linux for mission-critical workloads, 8% use CentOS, 8% use Novell SUSE Linux, 7% use Ubuntu Linux, and only 7% report using Oracle Enterprise Linux. There was minimal shift from 2010, in that 30% of IT users reported running Red Hat Enterprise Linux for mission-critical workloads, 9% used Oracle Enterprise Linux, 8% used SUSE Linux, 5% used Ubuntu Linux, and only 1% used CentOS.

Other operating systems. Data centers also deploy a smattering of Unix and mainframe OSes, though, according to DCD 2011, this group is not as statistically significant as Windows and Linux users. Table 5 highlights the use of Unix and mainframe OSes in a non-mission-critical environment. Sun Solaris and IBM AIX lead the pack in Unix, though Sun Solaris use declined slightly into 2011. The decline for i/OS was even more noteworthy.

Table 5 – Unix and mainframe operating systems in non-mission-critical data centers

Operating system 2011 results 2010 results

Sun Solaris

24%

27%

IBM AIX

24%

23%

HP-UX

14%

20%

z/OS

12%

13%

i/OS

10%

16%

Many environments using Unix and mainframe OSes are mission-critical, as shown in Table 6. In DCD: 2011’s findings, IBM AIX edged out Sun Solaris as the mission-critical OS of choice in 2011. However, the overall use of both OSes changed little year over year. 

Table 6 – Unix and mainframe operating systems in a mission-critical data center

Operating system 2011 results 2010 results

IBM AIX

20%

18%

Sun Solaris

18%

20%

z/OS

12%

12%

HP-UX

11%

13%

i/OS

10%

16%

This was first published in July 2011

Dig deeper on Blade servers

Pro+

Features

Enjoy the benefits of Pro+ membership, learn more and join.

0 comments

Oldest 

Forgot Password?

No problem! Submit your e-mail address below. We'll send you an email containing your password.

Your password has been sent to:

SearchWindowsServer

SearchEnterpriseLinux

SearchServerVirtualization

SearchCloudComputing

Close