The Enterprise Network of the Future Will Be Hyperconverged Gartner RAS Core Research Note G00208637, Bjarne Munch, David A. Willis, 18 November 2010, RA9 05012011 In the enterprise network, we are entering the era of hyperconvergence, where all services are delivered by a common Internet Protocol (IP) network not only for voice, video and data, but also for wired and wireless, and for storage and computing. We propose a new set of design and collaboration principles. Key Findings Application traffic has become more unpredictable and more difficult to model. Key vendors, such as Cisco, HP and Oracle, are pursuing a fully integrated stack approach, spanning areas including networking, storage and servers. Taking advantage of these offerings requires a collaborative effort, with clear lines of decision making regarding the architecture. Standard hierarchical network designs inhibit consistently good performance. Recommendations Network designers should focus their designs around end-to-end solutions for (1) broad infrastructure domains, such as the data center and the WAN, and (2) the workplace environment, such as the campus or branch office. Traditional, technology-centric domains, like voice, data, WAN, etc., only serve as barriers. Managers must also change the organization and collaborative structure by, at minimum, opening up the lines of communication between teams, led by cross-disciplinary leaders. Cost-effective support is achieved via a tightly integrated third-level planning team that is supported by teams of versatilists (that is, engineers who span multiple disciplines). Integrate all network domains and user access control mechanisms to enable seamless user connectivity at the network access and application traffic flow within the network by removing centralized control and connectivity hubs. Apply security policy at the first touchpoint to the network, at multiple layers within it.
2 WHAT YOU NEED TO KNOW As cloud computing, virtualization, mobility, unified communications and video drive more application traffic to the network, traditional network design practices will become increasingly significant constraints on the functioning of the business. Traditional silos of server, storage and network are being combined as vendors sell integrated stack solutions. Security policy management capabilities have to exist at every touchpoint to the network. A new approach that brings together these disciplines enables organizations to make the right decisions about architecture, rather than being led by a single infrastructure view. ANALYSIS Driven by the CIO s top priorities in cloud computing, mobility, unified communications and video, planners must think ahead to how networks are designed, how they are built and how they are run. Dumb pipe networks will not support the levels of scale, reliability or security that business leaders need. Therefore, the silos between network infrastructure and applications are breaking down, as evidenced by major trends such as contextaware computing, application-fluent networking and identity-aware networking. Meanwhile, cloud computing and virtualization are collapsing the boundaries between server, storage and network. The user environment is also moving to a mix of wireless and wired capabilities. We see these changes collectively in the hyperconverged network an approach in which compute, storage, content management, communication and application services are consumed over an integrated network, enabled by virtualization, with integrated security, independent of the network access method, defined by policy, and crafted to the specific device and situational need. Corporate and government networks are hitting an inflection point. After years of merely evolving communication systems in a slow reaction to changing needs, a fundamental change in approach is necessary. This is not merely a move to a consolidated network, or to a common set of protocols. Instead, it is a multidisciplinary approach that anticipates changes in the work environment, leverages a mix of consumerized and commoditized technologies, takes advantage of cloud-service capabilities, and maximizes the efficiency of the core backbone. It is easy to see why this change must occur soon; end-user demands are rapidly changing. Collaboration systems will allow users to move seamlessly from traditional phone calling into audioconferencing, shared workspaces and videoconferencing not as separate systems, but as an integrated experience. Voice, video and data can no longer be separate domains. Delivery cannot just be on-premises, thus restricting access to internal employees, but also must integrate to public cloud-based solutions that span multiple organizations once again, with a minimum of user hassle. The resulting traffic flows will be more peer to peer in nature and less hierarchical. Another reason why this change must occur soon is the messy nature of legacy network systems that comprise the typical network, even as more modern systems are thrown into the mix tactically. It is common to many different models of switches and routers, and multiple PBX or key systems with little to no integration, all running dozens of different software versions. Many traditional equipment suppliers have gone out of business or moved on to new product portfolios. Maintenance costs are on the rise. Complexity is the barrier to lowering cost, improving reliability and enhancing security. Old network designs based on old and invalid assumptions are common. Cost is another aspect. Even as the cost of bandwidth from carriers has declined, the operational costs associated with maintaining networks have risen. Network systems and data communications analysts are projected to be the second-fastest-growing occupation in the U.S. (Source: U.S. Bureau of Labor Statistics Occupational Outlook Handbook, 2010-11). Total network costs, as a percentage of IT spending, currently measures 15% at minimum, a percentage that is likely to increase to more than 20% by 2014, driven by more wireless service consolidation into IT and the rapid rise of video. To adapt, planners should focus on removing bottlenecks and complexity in the network so that they are able to adapt to the changing ways that business users access their applications, and to the changing, and even chaotic, traffic flows in their networks. This means that enterprise network design must move beyond the traditional network connectivity focus to include how to connect users to applications, and essentially adapt their network solutions to changing application deployments and the changing business needs of these applications. Enterprises need to plan their network design around the ongoing breakdown of traditional technology boundaries and traditional network constraints, where usually separate solution areas converge into the hyperconverged enterprise, such as: Convergence in the data center of networking, servers and storage, enabling a more cost-effective network design in support of server virtualization Convergence of the LAN and WAN, in support of changing application deployment architectures (in particular, cloud computing, voice, video and collaboration) 2010 Gartner, Inc. and/or its affiliates. All rights reserved. Gartner is a registered trademark of Gartner, Inc. or its affiliates. This publication may not be reproduced or distributed in any form without Gartner s prior written permission. The information contained in this publication has been obtained from sources believed to be reliable. Gartner disclaims all warranties as to the accuracy, completeness or adequacy of such information and shall have no liability for errors, omissions or inadequacies in such information. This publication consists of the opinions of Gartner s research organization and should not be construed as statements of fact. The opinions expressed herein are subject to change without notice. Although Gartner research may include a discussion of related legal issues, Gartner does not provide legal advice or services and its research should not be construed or used as such. Gartner is a public company, and its shareholders may include firms and funds that have financial interests in entities covered in Gartner research. Gartner s Board of Directors may include senior managers of these firms or funds. Gartner research is produced independently by its research organization without input or influence from these firms, funds or their managers. For further information on the independence and integrity of Gartner research, see Guiding Principles on Independence and Objectivity on its website, http://www.gartner.com/technology/about/ombudsman/omb_guide2.jsp
Convergence of the wired and wireless network, enabling the wireless office The Evolution in Network Design Critical network design factors have changed in three major waves since the 1980s, as illustrated in Figure 1. Often, these designs remain in use today, despite their inefficiency and complexity. In the 1980s, hierarchical systems ruled, where all user services came from the next upstream neighbor, and all connections were wired. There was little convergence beyond Systems Network Architecture (SNA) host emulation on the LAN. Wide-area pricing was driven by point-to-point links and distance-sensitive charging. Meanwhile, distributed departmental systems proliferated, setting the stage for the next wave. In the 1990s, distributed systems became more interconnected, but in a highly static way. Distance still mattered, so network designers created traffic hubs that would aggregate traffic to take advantage of capacity pricing. Fixed time division multiplexing (TDM) voice networks remained based on the 1980s model, but data services began to converge around a common protocol (IP). LANs became fully integrated. Meanwhile, mobile (cellular) services ramped up, but were centered on voice. The 2000s saw the first wave of true convergence of voice and data around IP. Distance became immaterial in-country over the WAN as service-specific pricing appeared, and global regional hubs became the common design for multinational enterprises. Internet services became an important element of the corporate network, but usage was tightly controlled and limited. Meanwhile, the mobile enterprise expanded rapidly, and Wi-Fi grew from a vertical market data collection technology to become a common connection method for carpeted-space deployments in the enterprise. These old designs are still found in operational networks today, leading to a need to modernize them. Hyperconvergence: The Fourth Wave in Network Design The nature of the business environment and the application environment is becoming highly distributed, and application traffic is taking on an almost chaotic nature, moving between all offices, external and internal, and with end users connected via wired, wireless and mobile methods. This is driving convergence in several areas of the network into a hyperconverged infrastructure, where new, distributed network designs center on the following factors. 3 Figure 1. The Evolution in Network Design Is Changing Network Topology From Hub and Spoke to Meshed 1980s Host Voice Networking 1990s WAN 2000s Campus Networking First Convergence Central Office Switch Mainframe Wi-Fi Voice PBX Front-End Processor Controller Core Distribution Phones Access Distance Matters Hierarchical Isolated ->Converged HQ: headquarters
4 Cloud computing Enterprise applications will increasingly move into various types of private and public cloud services, often with limited consideration of the network impact. Traditional centralized hub-and-spoke architectures do not handle these application deployments very well, and hybrid Multiprotocol Label Switching (MPLS) and Internet networks are often still centralizing traffic flows for security reasons, thus introducing similar issues as for hub-andspoke architectures (see Figure 2). Enterprises need to remove these bottlenecks by flattening the network (i.e., adding additional touchpoints between all networks, and distributing more-secure Internet access points around the network). This will improve efficiency, eliminate bottlenecks and single points of failure, and provide a better experience for the user. This may be implemented through technologies that can support distributed Internet access with centralized security policy administration, and service providers that can offload this administrative burden. IP communication for all data types Enterprise communication is migrating to IP-based platforms and being integrated with unified communications and collaboration (UCC), which will increasingly be deployed both internally (within the enterprise) and externally (in the providers clouds; see Figure 3). These will not be deployed as separate solutions, but will be based on integrated hybrid designs. IP telephony and UCC systems based on Session Initiation Protocol (SIP) will allow UCC functions to be distributed around the network, while still remaining functionally integrated. In addition, this will enable highly changeable applications to leverage elements in the voice network that change very little. Enterprises should build architectures that enable the distribution of communications and collaboration functions across separate systems. For example, in UCC, standardize around a technology stack (and vendor) for telephony, messaging, conferencing and collaboration systems both in-house and in the public cloud and social-networking-based solutions. The further up the technology stack, the more diversity should be expected and allowed. Data center virtualization The impact of server virtualization on data center networking has already been profound, and will continue to drive significant changes in data center network design. Vendors like Cisco, HP, Oracle and IBM want to own the stack from top to bottom. From a design perspective, data center networking will flatten and virtualize to enable a free and rapid movement of virtual machines and server load within the data center, and this will continue to evolve with distributed virtualization between data centers. Today, it is technically possible to move virtual server instances between remote data centers at the click of a mouse, as demonstrated by companies such as F5 and VMware. This can move gigabytes quickly, but will drive a need for highcapacity network core, where Ethernet WAN-based interconnected data centers will be the preferred solution due to their support of highly scalable bandwidth and low latency (see Figure 4). Enterprises need to determine whether their strategies allow for a fully integrated stack solution from a single vendor. Also, they should anticipate a need for best-of-breed solutions in infrastructure areas like application delivery controllers, where individual vendors still have a significant innovative lead. Figure 2. The Enterprise Network Typically Consists of Several Network Clouds (e.g., MPLS and Internet in the WAN) Figure 3. Communication and Collaboration Is Being Integrated and Deployed Internally and Externally 2010s WAN and Cloud 2010s Voice HQ MPLS Internet Conferencing Telephony Collaboration Messaging
Figure 4. Data Centers Will Be Directly Interconnected to Support Low-Latency and High-Bandwidth Needs Figure 5. There Is Emerging Convergence of Compute and Storage Networking in the Data Center 5 2010s Data Center Interconnect 2010s Data Center Networking DC DC Compute Storage DC DC: data center Data center Ethernet Server virtualization is also a driver behind the ongoing discussion about convergence of data networking and storage networking within the data center. The evolution in computing hardware and its virtualization has led to an explosion of required interfaces and cabling within each server rack, which is the key reason for the trend toward convergence of storage networking onto Ethernet (see Figure 5), Fibre Channel over Ethernet and Converged Enhanced Ethernet. This evolution is being supported by top-of-rack switching and new unified switches that support both standards. Wireless-first access As wireless capacity, affordability and ubiquity are increasing, users will be satisfied and, indeed, happier with wireless-only access for most workloads. Driven by Institute of Electrical and Electronics Engineers (IEEE) 802.11n today and 802.11ac in the coming years increasing wireless LAN (WLAN) bandwidth capacity, enterprises will evolve the WLAN from just being connectivity for convenience into a complete office connectivity solution. The all-wireless office is emerging, in which companies will blanket their organizations with Wi-Fi, which will become the default connection for all services inside the firewall, and the common connection model for mobile workers (see Figure 6). LAN designs will be greatly simplified, moving to a simple twotier approach. Design principles will change. The new network design needs to focus on enabling application traffic flow across the entire network, where end users will expect consistently good performance levels, irrespective of their method of connectivity, and where application processing can be moved to available processing capacity, wherever it is located, for cost optimization. Security mechanisms need to be embedded and distributed within the network to avoid a bottleneck design. Remote monitoring and network traffic control, and proactive network changes based on network policies, become indispensable as more and more application traffic will exist outside the traditional enterprise s boundaries, and increasingly will consist of HTTP and virtual desktop traffic, which are both notoriously difficult application traffic types in the network. Managers must also change the organization and collaborative structure, at minimum opening up the lines of communication between teams led by cross-disciplinary leaders. Some leading organizations have already moved to an organizational structure that separates foundational networking (i.e., transport, switching and routing) versus applications (unified communications, mobility, messaging, endpoints and application servers) versus logical networking (directory and addressing, including Domain Name System [DNS], Dynamic Host Configuration Protocol [DHCP] and Active Directory). Multiple levels of the delivery and support chain are also changing; cost-effective support is achieved via a tightly integrated third-level planning team that is supported by teams of versatilists.
6 Tactical Guidelines Integrate planning of network, storage, server and desktop infrastructure via a cross-disciplinary planning team. Separate designs into core infrastructure domains: data center infrastructure, wide area infrastructure, campus, remote office, etc. Examine whether your current network design enables traffic flow between all users and their applications (e.g., by region or country) so that performance is acceptable. This is especially critical when handling high volumes of collaboration traffic, which is often between peers, and in adopting public cloudcomputing models, which require highly distributed (but still secured) connections to the Internet. Figure 6. Wireless Network Access Is Becoming Pervasive 2010s Campus Networking Blanket Wi-Fi