|Телеком||ТВ и медиа||Облака||ПО||Кадры|
|ИТ в образовании||ИТ в медицине||Big Data||E-commerce||Спутниковая связь|
|Все новости||World News|
Data Center Network Managers Are Increasingly Being Tasked With Meeting Application Performance Goals
|29 июня 2009|
Polling results from Gartner's Cannes Symposium indicate that attendees face long-standing pressures to prevent network performance problems. Many data center network managers are also being tasked with meeting application performance objectives.
Results of interactive polling at Gartner's November 2008 Symposium Conference in Cannes, France show that data center network managers are shifting their investment priorities for network management tools to meet application performance objectives.
Network managers are being pressured to move up the stack, and not just prevent network performance problems, but meet application performance objectives.
The requirement to manage new unified communication technologies in Europe significantly outpaces the pressure to manage other network technologies, such as wireless, Multiprotocol Label Switching (MPLS), and virtualized environments.
Most respondents have no formal process for releasing applications to the production network.
Invest in network configuration, fault and performance management tools that can be expanded to encompass and manage new network technologies, rather than investing in tools that treat each new network technology as a separate stovepipe.
Learn how to correlate network, unified communication service, and business application data to develop an integrated picture of overall system behavior.
Prior to investing in network configuration and change management (NCCM) tools, establish standard network device configuration policies to reduce complexity and enable more-effective automated change.
Get involved earlier in the application release process, using network emulation, simulation and modeling tools to accurately predict the performance of applications on the production network.
In an interactive survey conducted at Gartner's November 2008 Symposium Conference, attendees to the session titled "Network Management: Tools of the Trade, Trends and Best Practices" were asked about their network management pressures, their top network management tool investment priorities and their processes for releasing applications to the production network. Gartner recognizes that the sample size of approximately 40 respondents does not represent a statistically significant distribution; nonetheless, the results of these polls will interest anyone involved in network management.
The Drivers of Network Management Investment
Presented with a list of pressures — from basic network management disciplines to strategic business alignment to the pressures of dealing with new network technologies — attendees highlighted as the top pressures driving new investment in network management (see Figure 1) the requirements to proactively prevent network performance problems, meet application performance service-level agreements (SLAs) and address the growing presence of unified communications/voice over IP (VoIP) in their environments. (Although network managers clearly face multiple, simultaneous pressures, this polling only allowed them to choose one response as their most important pressure.)
Figure 1. What's the top technology or business pressure that's driving your new investment in network management? (N=37)
Source: Gartner (November 2008)
The highest percentage of respondents (20%) is focused on moving to a more proactive stance in which they're preventing network problems from affecting end users and the business applications that run on the network. As the network evolves into a utility, network administrators are being held immediately responsible for increasingly complex, application-like, network services (for example, unified communications and Internet Protocol television [IPTV]). In this situation, network management tools and processes must not just focus on establishing the root cause for problem resolution, they must also focus on proactively preventing the problem from occurring in the first place; thus, it's a positive sign that the largest percentage of respondents focuses on this goal.
Meeting application performance SLAs and managing unified communications/VoIP follow closely, with 19% each. Unified communications is just the latest in a long line of new technologies and new demands on the enterprise network. Clients will face continuing pressures with each new technology that arrives on the scene. We recommend that clients invest in network configuration, fault and performance management tools that address current pressures and can be expanded to encompass and manage new network technologies, rather than investing in tools that treat each new network technology as a separate stovepipe.
In the end, networks must be managed in the context of the business-relevant IT services they enable. Although knowledge of the status and performance of individual network components will always be important for detailed troubleshooting, it's vital to the business that network managers understand the status and performance of business-oriented, end-to-end IT services. However, only 8% of respondents have the goal of aligning with the business in their sights. A network-specific approach to management that relegates the network management team to the status of peripheral plumbing does little to foster alignment with business units, because the business units are concerned with overall, end-to-end availability and performance. Since the network team is not solely responsible for end-to-end availability and performance, nor should they attempt a coup to seize responsibility for it; they need to work with service delivery managers, application managers, and relationship managers to identify the end-to-end, service-level goals, and translate those into their own goals for alignment.
Network Management Tools for 2009
Attendees were presented with a pyramid model that defined specific types of network management tools, and categorized them under four crucial aspects of network management:
Element management — including NCCM tools
Operations management — including network fault management, protocol analysis and root cause analysis tools
Service management — including network performance-monitoring tools
Business management — including network capacity-planning tools
Attendees were asked where they were placing their network management tool investments in 2009 (see Figure 2). In addition to the above network-specific management tools, we included application performance monitoring (APM) tools as an option. As with the previous question, although network managers may be planning multiple investments, this polling only allowed them to choose their No. 1 priority.
Figure 2. What is your No. 1 priority network management tool investment in 2009? (n=42)
Source: Gartner (November 2008)
Network managers whose top priority investment is in APM tools (42%) outweigh the nearest contender by a factor of two to one. This reflects two realities that network managers face:
Network managers are being tasked with the responsibility for application performance.
No matter what the problem is, the network will be the first target for blame.
More users and applications depend on the network than ever before. It's not just network services such as VoIP that depend on the network, but business applications, such as ERP and CRM. As such, the network becomes a business enabler for an enterprise with utilitylike expectations. The details of how the network works, what's broken or what new capacity must be added are of little concern to business users. Business users view the network through the lens of the performance of their business-critical applications. For the network to be perceived as delivering good service quality, it must be measured in terms of the application performance it enables. As a result, many network management teams are being evaluated based on application performance, not just network performance, and they need the tools to measure and report their application-performance-related results.
Another factor that might have influenced the magnitude of this response is vendor-encouraged confusion between application performance management and application-aware network performance management. Many vendors that monitor network packet flows (typically via NetFlow or some protocol based on NetFlow) have taken the very real insight they provide into the application, or service context of network performance, and amplified it into a claim that they are providing genuine application performance management. Although rarely defensible, this claim has muddied the conceptual waters, and many network administrators think they are buying application performance management when they are, in fact, just buying some rich network performance analytics built on top of a NetFlow data collection engine.
Although networks are often blamed for application performance problems, frequently the underlying issues are poorly written applications, inadequate system resources or simple human error, not to mention unannounced changes in user behavior or usage demands. Our prediction is that, through 2012, more than 80% of application performance and availability failures will be blamed on network problems, but the network will represent less than 20% of the root cause.
APM tools are certainly not exclusive to the network management domain, and are generally more widely used by the application support team and IT operations. It's best when these tools are used collaboratively across IT operations departments, although this doesn't happen as often as it should due to stovepiped IT organizations. However, because the network is usually presumed to be at fault, network managers have long been burdened with the responsibility of diagnosing the actual root cause of an application performance problem. Thus, network managers are wise to have APM tools on hand. Understanding the relationships between application performance and network performance is critical to predicting application performance problems and correctly diagnosing those problems when they occur.
The second priority investment was in network performance management tools (21%), and this category was further broken down into the two basic methods of network performance monitoring:
Network performance reporting — polling network devices to collect standard Simple Network Management Protocol (SNMP) Management Information Base (MIB) data for performance reporting and trend analysis
Network traffic analysis — using specialized network instrumentation (such as probes, appliances and NetFlow) to analyze the makeup of network traffic for performance monitoring and troubleshooting
In the performance management category, 7% of attendees expressed an interest in reporting SNMP MIB data, while 14% placed a priority on network traffic analysis. These results are in line with our typical client inquiry experience. The broader base of Gartner client inquiry calls also reflects stronger interest in the use of NetFlow for network traffic analysis. NetFlow has grown in popularity as a reliable and inexpensive data source that many network managers are using to augment their network intelligence with details about the distribution of protocols and the makeup of application traffic on the network. Ongoing traffic analysis enables reallocation of network resources to higher-priority users or applications, without the need for additional capital investment, using bandwidth allocation, traffic engineering and quality-of-service techniques.
Although it's not a large percentage, the NCCM tools have grown from practically nonexistent to capture 10% of the top priority interest from network management buyers. This growth is driven by corporate compliance and audit requirements. Previously, network managers would have used labor-intensive, manual, command-line efforts; a few tools from their network hardware vendor; and some homegrown, custom scripts. New NCCM tools provide more-rigorous change management and compliance audit capabilities, and, in some cases, reporting on network compliance with Payment Card Industry mandates. Network managers will increasingly be called on to replace manual processes with automated NCCM tools to monitor and control the change process, thus improving staff efficiency, reducing risk and enabling enforcement of compliance policies. Prior to investing in NCCM tools, establish standard network device configuration policies to reduce complexity and enable more-effective automated change.
Process for Releasing Applications to the Production Network
Ideally, production network operations teams work with application development (AD) teams early in the application life cycle to prioritize applications; develop SLAs; specify key thresholds (for example, error, warning and critical); and use network emulation, simulation and modeling tools in the quality assurance (QA) test lab to ensure optimal network design that supports business applications and reliably predicts future performance, based on an understanding of business demands. However, when asked to characterize their process for releasing new applications to the production network, the highest percentage of respondents just "throw it over the wall" from AD and hope for the best (see Figure 3).
Figure 3. How would you characterize your process for releasing applications on the production network? (n=35)
Source: Gartner (November 2008)
The fact that the largest percentage of respondents has no formal process for releasing applications onto the production network is in line with our typical Gartner client inquiry experience. But the percentage of European attendees that require an application to be certified by a separate network management test lab prior to its release on the production network (20%) is significantly larger than what we see from the broader base of Gartner client inquiry calls. European companies often demonstrate more process rigor in all aspects of IT operations, and this kind of certification is evidence of that. However, a separate network management test lab is likely to have largely redundant facilities with the AD QA test lab, reducing the overall cost-effectiveness of the IT department. In addition, the situation could easily degrade into a finger-pointing "blame game" between AD and IT operations, where the separate network test lab is accused of being the "deployment prevention committee."
Although it requires the challenging task of breaking down some of the organizational boundaries found in typical IT shops and using sophisticated network management tools, 9% of the attendees have succeeded in getting network engineers to participate directly in AD QA test labs, using network emulation, simulation and modeling tools to predict the performance of an application once it's released on the production network. This is a more-productive and efficient situation than having a separate network management test lab.
As applications become more modular and distributed, it will become vital for network engineers and network planners to be involved earlier in the application life cycle. We recommend that network management teams start with informal, ad hoc use of Sniffer trace analysis and network emulation/simulation tools for a few application deployments. Profile applications and predict production performance by injecting the knowledge of real-world capacity, slower WAN links and other applications competing for bandwidth. Demonstrate the accuracy of the predictions and the benefits of knowing beforehand the application performance that end users or customers will experience. Use these early pilots and test cases as the foundation for ensuring that network engineers' participation in the AD QA test lab becomes a formal, required part of the application release process.