This report highlights the competitive landscape of the leading vendors in the data center Ethernet switch market. It outlines the competitive situation and trends in this market, and then describes the leading vendors and discusses the future of competition. Competitive profiles of vendors are also included.

This document was revised on 28 July 2010. For more information, see the corrections page on gartner.com.

• The data center Ethernet switch market is an attractive market with $1.9 billion of vendor revenue and shipments of 10.1 million ports of Gigabit Ethernet and 10-Gigabit Ethernet switches during 2009, representing 13.8% of the overall enterprise Ethernet switch market in terms of value.

• The current bulk of Ethernet switches deployed in existing data centers consists of general-purpose Gigabit Ethernet switches. However, several distinct user requirements, such as lossless Ethernet and virtualization support, point to an emerging market that is differentiated by key trends and new technologies that try to address the network challenges in data centers.

• Cisco Systems is the current leader in the data center switch market, with a shipment share of 53.1% and a revenue share of 47.8% during 2009. HP, Juniper Networks and Blade Network Technologies are the key competitors in the market.

• Data-center-specific technologies are currently in several stages of development, with very few actually available commercially. In many cases, standardized offerings are one to two years away, making it important for vendors to give assurance of investment protection to users.

• Refrain from marketing a "campus switch" as a "data center switch." And explain the features and functionalities that make your offering a data-center-optimized Ethernet switch.

• Demonstrate your existing capabilities for the market, and share your complete vision and road map with potential customers, to win their confidence that you understand their requirements and are aligned to the overall key trends in the data center space.

• Those looking for forming alliances and distribution networks should explore all remaining options among server vendors, as well as regional system integrators (SIs), to compete more effectively in the market and remain viable in a highly polarized market.

• Go to market with a complete portfolio and solutions rather than point products, as users are less likely to go in for disparate purchases and point products. Providers that lack key pieces should acquire them internally or through alliance partners.

• As most data center network standards are not ratified yet, vendors need to keep customers aware of their road maps and future product releases to stimulate demand. That should also help to raise vendor mind share.

A data center Ethernet switch traditionally can be defined as an Ethernet switch deployed inside a data center or in an adjoining network room, providing network connectivity and services between users and the server and storage assets. Traditionally, such Ethernet switches have been no different from campus or branch office Ethernet switches. However, in consideration of the new kinds of requirement coming from a data center that is rapidly changing, there is a growing need to define the market for Ethernet switches that are designed specifically for data centers and purpose-built to fit into next-generation data center designs.

During the last few years, a number of major trends have emerged to define the next-generation data center infrastructure. These trends can be grouped under the following key heads:

• Consolidation — Consolidation of IT within a data center, consolidation of many data centers down to few strategic sites, and consolidation of branch office IT to central data center sites

• Virtualization — Virtualization of servers, storage, networks and applications, among others

• Automation — Integrated management of all IT components, including virtualized and nonvirtualized assets, along with associated facilities and energy usage

• Converged fabric — The emergence of Ethernet as the fabric choice for addressing all of the connectivity within a data center, including storage networks

Each of these new trends has had a significant impact on how the network for the next-generation data center should be designed. While the requirements of the data center network are still evolving, certain specific user needs and characteristics distinguish the data center Ethernet market from other Ethernet markets.

Listed below are some of the key user needs that are unique to the data center Ethernet switch market:

• Real and comprehensive need for 10-Gigabit Ethernet switches at all layers of the data center network and backbone needs of 40 gigabit and 100 gigabit.

• Nonblocking switch fabric with one-to-one backplane-to-interface links. Oversubscription is increasingly unacceptable in virtual environments.

• Low network latencies of close to 5 microseconds in general for access switches with small form factor pluggable plus (SFP+) links going into nanoseconds.

• Flat topologies to bring down the number of hops and reduce the overheads in network latency.

• Virtual machine networking capability and scalable management of network services and policies during "v-motion" virtual machine migration and changes.

• Network scaling beyond spanning tree and virtual LAN (VLAN) limits, without adding processing and protocol overheads.

• High availability and resiliency designs with millisecond failover. The spanning tree with its multisecond convergence is unacceptable.

• Benchmark power consumption of less than 10 watts per 10-Gigibit Ethernet port in fixed port rack units. Embedded blade switches even lower. Front to rear airflow for optimal cooling.

Each of the above user needs will be explained in detail in our subsequent research reports.

Data center Ethernet switches, like the overall Ethernet switch, come in two types — fixed-function and chassis-based, but they are deployed in the data center network in varying forms and ways. The different types of data center Ethernet switches and their deployment options are described below.

Fixed-function switches, also known as stackable switches, offer connectivity, management and switching functions in a fixed configuration, although some modular ports may be provided for backbone or server connections. Data center switches are all typically managed switches.

Fixed-function switches are deployed as a blade Ethernet switch embedded within a blade server chassis. They are also increasingly deployed externally but within the rack, typically in pairs, occupying two or more rack units. Depending on the cabling option and the length of a row, these can be deployed in the end rack of a row (end of row) or the midrack of a row (midrow) to provide connectivity to all the racks in the row. With shorter-run cabling options, these switches are increasingly being deployed on the top few units of a rack (top of rack) to address connectivity for every rack or every other rack.

Designwise, they can be implemented as an access switch directly providing connectivity to server units. They can also take the role of an aggregation switch to provide backbone connectivity to blade switches or other access switches. To scale to a growing number of backbone links, they can be either stacked or joined together with multichassis link aggregation and virtual chassis technology to logically perform as a single chassis unit.

Chassis-based switches, also known as modular switches, can support various LAN modules, providing switched LAN ports at various speeds and offering optional features, including network management, Layer 3 routing, security services and Layer 4-7 services.

Chassis-based switches are typically deployed in the main distribution area of the EIA/TIA-942 data center network layout, which can be a dedicated rack in a separate zone in the raised floor area or in a neighboring network/telecom room, sometimes in both. They are also an option for aggregation deployment in the horizontal distribution area which corresponds to a dedicated rack enclosure in the server zone.

These switches are mostly deployed as core switches in the data center, providing connectivity to external users and acting as the root switch in the data center. In many data centers, they also continue to be deployed as aggregation switches, although in new deployments, some of them are being replaced by fixed-function switches.

There are several standards initiatives dedicated to data center Ethernet switching requirements. While some of them are already ratified, others are in the process of development and approval. Most of these are under the aegis of the Institute of Electrical and Electronics Engineers (IEEE) or the Internet Engineering Task Force (IETF). Vendors are best advised to actively support these standards and initiatives. If they are not part of these initiatives currently, a future road map to support them is increasingly becoming mandatory in the market. Below are some of the key standards:

• IEEE Data Center Bridging (DCB) task force under the 802.1 working group provides a framework to create a lossless Ethernet. Described in the past with various terminology such as Converged Enhanced Ethernet (CEE) and Data Center Ethernet, DCB is a collection of evolving standards under development:

  • Congestion Notification provides end-to-end congestion management for protocols that do not already have congestion control mechanisms built in (e.g., Fibre Channel over Ethernet [FCoE]).

  • Priority-based Flow Control (PFC) provides a link-level flow control mechanism that can be controlled for independently for each priority. The goal of this mechanism is to ensure zero loss due to congestion in DCB networks.

  • Enhanced Transmission Selection (ETS) provides a common management framework for assignment of bandwidth to traffic classes.

  • A discovery and capability exchange protocol that is used for conveying capabilities and configuration of the above features between neighbors to ensure consistent configuration across the network, likely to leverage 802.1AB (Link Layer Discovery Protocol).

• IETF's Transparent Interconnection of Lots of Links (TRILL) and IEEE's Shortest Path Bridging (SPB) are separate but somewhat common initiatives that attempt to create a new layer of transport that takes the best of Layer 2 MAC bridging and Layer 3 IP routing. The goal of both these initiatives is to make Ethernet adopt a transport scheme that utilizes multiple paths and the least-cost/shortest-path calculation to forward/route traffic. This achieves larger domains and greater stability and resiliency/faster failover without adding protocol and configuration overheads.

• The Fibre Channel Industry Association's T11 technical committee has adopted a common addressing structure for the FCoE standard. FCoE enables storage area network (SAN) traffic to be natively transported over Ethernet networks, potentially unifying the LAN and SAN in the data center.

• Virtual Ethernet Port Aggregator (VEPA) is an IEEE initiative to allow virtual machines to effectively benefit from the external dedicated switching device for scaling connectivity requirements and consistency in implementing access and security policies to the virtual environment. The objective is to reduce complexity and potentially improve performance by offloading advanced network functions from the hypervisor and virtual machine to an adjacent bridge.

As described above, a distinct breed of Ethernet switches is required by data center users. However, it is to be noted that the data center market is still composed of a large number of traditional Ethernet switches deployed with few or no capabilities that meet the distinct requirements emerging in the market. For our analysis, we take into account both the traditional Ethernet switches deployed in data centers and those that offer differentiated data center functionalities and features. While assessing traditional Ethernet switches deployed in data centers, there were considerations to be made to account for overlaps in core switches common to data center and end-user LANs. In such contexts, we have accounted for only the server-side ports.

Based on analysis of our quarterly market share research and a fresh survey of vendors, the market for Ethernet switches deployed in data center networks was worth vendor revenue of $1.98 billion in 2009, which is 13.8% of the total enterprise Ethernet switch market. However, in terms of ports shipped, data centers contributed only 10.1 million ports, which is 3.3% of the overall number of ports shipped to enterprises — reflecting the fact that the data center market for Ethernet switches is primarily a Gigabit Ethernet and 10-Gigabit Ethernet market, with hardly any shipments of Fast Ethernet, which continues to be the bulk of the overall Ethernet switch market.

Data centers have a distinct and immediate requirement for speeds of 1 gigabit and upward, even at the access layer, due to input/output (I/O) connection requirements multiplying with server consolidation and virtualization. Interserver operations, as well as the emerging SAN connectivity requirement, are primary drivers for the Ethernet LAN to scale and acquire lossless characteristics.

As a result, the data center market contributes significantly to 10-Gigabit Ethernet adoption in general, data center deployment accounting for 38.9% of vendor revenue and 37% of port shipments of the overall enterprise Ethernet switch market.

The market for data-center-optimized Ethernet switches compared with the overall data center market for Ethernet switches is relatively small, given that few products in the market comprehensively address data-center-specific requirements and functionalities as described in the sections above. We expect this situation to remain for at least a year — until standardization brings about interoperability and also results in greater adoption of the emerging technologies by multiple vendors. Table 1 shows the data center Ethernet switch market in 2009.

The data center market for Ethernet switches has traditionally been addressed by campus LAN vendors, the competition among them mainly driven by the ability to provide switches with high-speed performance and greater port density.

Vendors' efforts to adequately respond to new user needs is resulting in an intense competition for the user's mind share, as well as spending. However, other demand- and supply-related trends are shaping the competition in the data center Ethernet switching industry.

Commoditization and market saturation have taken a toll on the Ethernet switching market in North America and Western Europe. Ethernet switches have little differentiation between different vendor offerings. A high level of substitutability between the products has led to increased price and margin erosion for the vendors. In addition, few enterprises are motivated to upgrade to Gigabit Ethernet for end-user connectivity. Bandwidth supply far outstrips actual demand in campus LANs, leading to stagnation of the overall Ethernet switching market.

In such a scenario, for most Ethernet switch vendors, focusing on the data center market is a logical progression. Not only does the data center market have more potential from a future growth perspective, but it is also less commoditized than the general-purpose market — making it an attractive proposition.

Technologies that are being developed to address some of the new data center network challenges, such as lossless Ethernet, Spanning Tree Protocol limitations and virtual machine network management, are still in the development stage. While vendors like Cisco, HP, Blade, Extreme Networks and Force10 Networks have many or parts of these functionalities, what is available consists of first-generation, prestandard offerings in almost all cases. While standardization initiatives are under way, many of them will take significant time to bear fruit.

Vendors leading these innovations are actively focused on deriving maximum mind share and mileage from their initial and limited offerings. Others are trying to compete in the market by offering a road map to their customers and partnering to get the technology capabilities or participating in the standardization process.

While big IT vendors, such as HP and IBM, have had Ethernet switch business or have been big suppliers of networking products along with their own IT solutions, in the past they have been at best involved in the backdrop as far as networking is concerned. In the span of the past 18 months, the equations have drastically changed. The emergence of integrated data center solutions like Cisco's Unified Computing System (UCS) and HP's Converged Infrastructure solutions, and the resulting cross-pollination among IT systems and networks have led vendors to cross hitherto uncrossed lines and launch offerings that compete with erstwhile partners in their primary focus areas. Network vendors like Cisco have entered server and storage business directly, while IT systems and storage vendors like HP, Brocade, IBM, Fujitsu and Dell have shown aggression and significant direct interest in the network business through new product offerings, acquisitions or building new alliances. This has led to the disappearance of many significant traditional alliances, creating a vacuum that needs to be filled by other partners or companies like Dimension Data, whose traditional alignment does not change. This also creates opportunities for network vendors like Juniper, in the form of new relationships that have the potential to transform its "go to market." Table 2 shows market forces driving the data center Ethernet switch market.

The data center Ethernet switching market is a fast-paced market abuzz with frequent vendor announcements and significant hype created by aggressive marketing campaigns and media outreach by vendors. Cisco and HP are the two megaplayers that directly address the market. Operating from their strengths in the networking and server markets, respectively, these two vendors have been actively positioning themselves with strong technology, channel and marketing initiatives targeting the data center market. While the common objective of these vendors is to propagate a vision of the network as a key component of an integrated data center solution under different names like Converged Infrastructure and UCS, these vendors' approaches differ. Not surprisingly, Cisco has a network-centric play in contrast to HP's server-centric worldview. Both these vendors have deep solution portfolios to address the market needs, though Cisco has a lead in this respect. HP has traditionally lacked a solid data center core switch, but it has significantly closed the gap by gaining this and other crucial pieces from its 3Com acquisition.

IBM is the other important megavendor in this market, though it does not have an Ethernet switch of its own. IBM's network component in its data center fabric offering is through relationships with Blade and Juniper. But its influence in the market continues to be quite strong, with involvement in standards initiatives, such as DCB/CEE and VEPA, and also by being a strong route to the market for its alliance partners.

Leveraging the IBM channel, Juniper is already in a good position in the market through its credible end-of-row/top-of-rack Ethernet switches and chassis-based core Layer 3 switches. Coming with its router and network security success story, Juniper, along with IBM and Blade, is today a credible alternative to Cisco and HP, with its compelling flat data center network strategy.

Since the introduction of blade servers, blade switches, which are embedded into the server chassis, have constituted a niche server-OEM-dependent market that has grown consistently over the years. Blade is the leader in this space, and it owns the majority of blade switch shipments in the market. Its traditional OEM partners have been IBM, HP and NEC, though HP has lately come out with its own Virtual Connect offering for this space. The other key player in this segment has been Cisco. With trends like fabric-based computing and virtualization, blade-based switches have become a significant component within the new generation of data center infrastructure.

While Gartner acknowledges the real opportunities for a fabric-based compute offering in the future and the benefits that it brings over traditional compute platforms, its view is that current offerings are somewhat restrictive and proprietary. However, the intensive focus and campaign by the megavendors have brought about greater credibility to fabric-based computing. This can result in faster adoption by users, while leading to vendor lock-ins and deployment of early-to-market proprietary technologies that may not interoperate with future standardized offerings. Gartner has cautioned that the road to true fabric-based compute deployment is unlikely to be short and trivial, with several technical and human factors needed for a matured deployment to emerge. The switched Ethernet fabric is one place where improvements are required to make the concept of fabric-based solutions become more relevant and bring greater value.

The other key group of vendors is the traditional vendors from the Ethernet switching industry with an interest in targeting the data center market. Many of them have significant installed bases in traditional data centers, which have typically gone in for vendors with an end-to-end range of products, especially high-density core switches. These include Brocade, Avaya (Nortel Networks), Force10 Networks, Extreme Networks, Enterasys Networks, Alcatel Lucent and Huawei, among many others. A good number of them have already charted a road map to reach their goals, while others are in the process of putting a strategy in place. Several of the traditional network vendors are also involved in standards initiatives, while also trying to make themselves relevant in the market by emphasizing their historical strengths and offerings that can be brought into play in the context of the emerging needs in data centers. A key challenge that these players face is their limited credibility in the data center market and the lack of significant channel partners to address the data center opportunity.

This market has also seen a number of new entrants with custom-built and merchant-silicon-based fixed-port-configuration Gigabit Ethernet and 10-Gigabit Ethernet switches focused on providing no-frills, high-speed connectivity to servers at very attractive prices. Arista Networks and Voltaire Networks can be put in this category. These players have positioned themselves as vendors dedicated to meeting the emerging data center Ethernet requirements. However, these niche vendors face significant challenges with their limited offerings, as well as a lack of go-to-market. The recent slowdown in network purchases and the challenges of competing with larger market actors have led to an early consolidation, leaving just a few of the smaller startups surviving. Some, like Woven Systems and Teak Technologies, have gone out of business. Those that survive are competing in the market with different market strategies like partner alliances and product expansions.

For the purpose of selecting vendors to profile in this research, we have included in our list the top six players either by vendor revenue or in terms of port shipments in the data center Ethernet switch market. The top six vendors hold 94.4% of the market in port shipments and 95.9% in vendor revenue, as shown in Figures 1 and 2.

Also, we have added the following two vendors because they met the criteria as described below:

• Extreme Networks — Though this vendor doesn't have a major share of the data center switch market, it has reasonably credible offerings in place. It also has a relatively well-developed strategy and road map for its data center Ethernet play.

• Arista Networks — The startup is in our list due to its data-center-optimized switches, which are attractively priced and come with a functionally rich and flexible operating system.

Standardization of data center technologies like DCB, VEPA, TRILL, SPB and 40-Gigabit/100-Gigabit Ethernet will drag out beyond 2010, due to process delays as well as long-standing contentious approaches. However, many of the prestandard offerings are adopted by users, given that there is indeed a distinct and latent demand for networks supporting key data center initiatives like server consolidation, virtualization and fabric convergence that are addressed by these new technologies. This gives an early lead to key vendors like Cisco and HP. First, they benefit from having a relatively comprehensive data-center-optimized Ethernet solution set. Second, they can also leverage the confidence that they command among users due to their installed base in networks and servers, respectively.

The fear of being locked into proprietary offerings continues to exist, but reassurances from the vendors to support future standards are good enough for many users to start deploying some of the existing offerings.

IBM continues to be a significant player, though it faces intense competition for the data center solution market from HP and Cisco. This will make it respond either by working more closely with OEM partners like Juniper and Blade or by acquiring its own Ethernet switch offerings.

There is significant competition in the market, with Cisco leading the market but HP at a strong second position. The gap between the two increasingly gets closer, with HP forming a much-broader offering after the 3Com acquisition.

Fabric-based offerings and preintegrated solutions don't find takers due to fears of getting locked into specific vendor products or technologies. Lack of standardization of Ethernet switch technologies that make up the fabric-based offerings also adds to the fear of not being able to interoperate with future Ethernet switches and other vendors. This causes users to continue making purchases of Ethernet switches from best-of-breed vendors and use SIs to integrate them into the overall data center IT solution.

This is the preferred scenario for most of the traditional network vendors, like Avaya, Force10 Networks, Extreme Networks, Enterasys and Alcatel Lucent. This gives them more space and time to align themselves with independent server OEM vendors and build their distribution network for the data center market. This is also the best scenario for the new data center networking startups, such as Arista Networks and Voltaire, as it opens up more markets and distribution partners for them as well. This is not an ideal situation for those that have placed heavy bets on the fabric-based/converged infrastructure road map, especially HP, Cisco and IBM, They will have to realign their strategy and partners in the short run to stay more relevant to the market need. Vendors like Cisco and Juniper, which have more-established distribution networks for the enterprise market, will emerge less impacted.

Ethernet switches are significantly impacting the cost structures of data center solutions by bringing the economies of scale of Ethernet (the most widely deployed enterprise networking technology) to data center connectivity within the data center LAN.

Data center Ethernet switches, in turn, are likely to be significantly influenced by commoditization, too. With the increased usage of merchant silicon, 10-Gigabit Ethernet switch prices are regularly charting new benchmarks at the lower end of the circuit. This phenomenon is expected to gain even more momentum as the 10-gigabit BASE-T copper cabling technology achieves greater market penetration with LAN-on-motherboard offerings from chipset vendors and lower network latencies and power consumption.

North America and Western Europe are the biggest markets for data center Ethernet switches. Markets like Asia, Central Europe and Latin America are nascent at this point, with general- purpose Ethernet switches mostly deployed in current data centers and server rooms.

However, the situation in these regional markets is likely to change in the medium and long term. Data center consolidation, as well as new data center construction, is picking up with aggressive growth in IT. And this is resulting in the need for modernized data centers with data-center-optimized Ethernet switches to cater to the requirement of a new breed of high-density servers and IT equipment.

As adoption of Ethernet switches in data centers grows with data center consolidation and their new additional roles in storage connectivity, their contribution to energy usage and carbon emission is likely to rise, with a significant share of the server and storage energy consumption being contributed by Ethernet-based I/O interfaces.

Networking vendors will have to develop and adopt technologies like Energy Efficient Ethernet to bring down the power consumption by switches as well as the network adapters in user stations. They will also have to subscribe to data center trends, such as space optimization and power/cooling efficiencies, introducing form factors that align with these needs.

Cisco is the market leader and the main influencer in the LAN switching market, offering an exhaustive portfolio of products to meet a variety of customer requirements.

In 2009, Cisco sold 5.4 million ports of data center switches, receiving associated revenue of $947.1 million. With a market share of 53.1% in port shipments and 47.8% in revenue, Cisco was the top vendor in the data center Ethernet switch market in 2009.

Cisco's portfolio includes two lines of switches that can be deployed in the data center and are available in virtual, rack, blade, chassis-based or fixed form factors:

• The Nexus series, specifically aimed at the data center, was introduced in 2008 starting with the Nexus 7000, and then other versions were subsequently released (Nexus 5000, 4000, 2000 and 1000). These are designed to support virtualization, cloud computing, Web 2.0 applications, and future unified fabric requirements. The Nexus 7000 are chassis-based switches, while the Nexus 5000 are top-of-rack switches. The Nexus 4000 are blade switches for blade servers, the Nexus 2000 are fabric extenders, and the Nexus 1000V are Virtual Ethernet Module. The Nexus 7000 and 5000 share the same operating system (NX-OS). This is a purpose-built operating system for the data center, offering feature flexibility for a variety of deployment scenarios.

• The Catalyst 4900 (released in 2004) and Catalyst 6500 (1999) can be deployed either in the campus or in the data center. The 4900 is a fixed switch, while the 6500 is chassis-based.

Even though the Catalyst has a road map that should take it to at least 2012, the Nexus series is clearly positioned as the preferred choice for future data centers. While the two platforms can coexist, maintaining multiple products with different operating systems and functionalities can lead to cost overheads and management complexities without proper planning. Following are the main differences and similarities between the Catalyst and Nexus series:

• The Nexus series supports 1 gigabit but is optimized for 10 gigabit and will support 40 gigabit/100 gigabit in the future, while the Catalyst is optimized for 1 gigabit, supports 10 gigabit, and will support 40 gigabit in the future.

• Both support virtualization: The Nexus supports virtualization within a single chassis as well as multiple chassis, while the Catalyst supports virtualization of multiple chassis.

• Due to its purpose-built operating system NX-OS, the Nexus supports convergence of networking and storage (or unified fabric) that the Catalyst does not support.

• The Catalyst supports MPLS, L4-7 services and legacy protocols that the Nexus does not support.

Customers who have already deployed lots of Catalyst switches often continue to deploy the same family for existing data centers. However, when they build out new ones, or when they are new customers, Cisco recommends turning to the Nexus line, even though they will still likely need Catalyst switches also for the services currently not available on Nexus.

Cisco has led in several aspects of data center infrastructure, for which it either has been the first vendor to offer those features or is still the only vendor to offer them:

• Network virtualization — Virtual Device Context is a logical virtualization at the device level that allows multiple instances of the device to operate on the same physical switch.

• FCoE support, with both the Nexus 4000 and 5000 series supporting FCoE.

• DCB features, such as PFC (including only two priorities, rather than the eight in the standard), ETS, Data Center Bridging Exchange Protocol and lossless service.

• Overlay Transport Virtualization — This industry solution provides customers with the means of extending Layer 2 networks into Layer 3 networks for applications within and between data centers.

Table 3 outlines Cisco's strengths and weaknesses.

Though HP has 30 years of experience in networking, it has largely had a low profile in this market, focusing primarily on the midmarket enterprise segment. However, all this has changed, with HP rapidly gaining mind share and consideration momentum in the marketplace, in particular over the past 18 months. In 2009, the ProCurve division was integrated into the bigger HP within its HP Enterprise Servers Storage and Networking organization. And to further reinforce its commitment, HP announced in November 2009 the acquisition of 3Com (including H3C and TippingPoint). The deal was finalized on 12 April 2010.

In 2009, HP shipped 824,200 ports of switches to data centers, accounting for 8.1% of the total shipments of data center Ethernet switches. HP's data center Ethernet switching revenue amounted to $657.7 million, or 33.2% market share. Given its acquisition of 3Com and H3C, combined data center shipments reached 943,300 ports, for a market share of 9.3%, and associated revenue amounted to $666 million, for a market share of 33.6%.

Traditionally, HP supports a standards-based approach and offers products with industry-leading total cost of ownership. Most of its networking products were aimed at the midmarket, but with the acquisition of 3Com, it gained a comprehensive portfolio of high-end enterprise products, targeted to the large-enterprise segment. The combined portfolio now comprises a complete range of switching, routing and security offerings able to address the various needs of a large enterprise. Under its new product road map, HP's data center Ethernet switches are grouped in the A-Series line and include:

• H3C A12500 — Large core switch, chassis-based, gigabit and 10-gigabit versions. Its architecture is designed to support 40/100-Gigabit Ethernet and FCoE, safeguarding network infrastructure investment.

• H3C A9500 — Core and aggregation for server/switch with SecBlade modules.

• H3C A58xx — Access layer/top-of-rack server aggregation, server access and aggregation with Open Services Network modules. Can also be used in enterprise core and distribution applications.

• Virtual Connect Flex 10 gigabit.

• A6120 Ethernet blade switch designed specifically for the HP C-class blade systems (3000 and 7000 enclosures), available in gigabit and 10-gigabit versions (Layer 2).

HP also offers the 2408 FCoE Converged Network Switch model from Storageworks, which features eight Fibre Channel ports for 8 gigabits-per-second performance along with 24 Ethernet ports for 10 gigabits-per-second performance. It also plans to introduce its own products with FCoE capabilities.

HP's data center strategy is centered around its FlexFabric strategy which is designed on a data center fabric with intelligence centered at the edge of the server that enables convergence of server, network and storage resources. HP has shown strong innovation in the data center switching space, with Virtual Connect server interconnect offering being a key example.

Table 4 outlines HP's strengths and weaknesses.

Juniper entered the switching market in April 2008, with the first commercial shipments of its EX Series of Ethernet switches (the EX3200 and EX4200 lines).

In 2008, Juniper revenue that was attributable to the data center was $28 million, and it grew by 243% to reach $96 million in 2009. That enabled Juniper to reach the third position among the top vendors in the data center Ethernet switch market during 2009, with a revenue market share of 4.8%. Port shipments amounted to 961,400 ports, accounting for 9.5% of the market.

Juniper's data center Ethernet switch portfolio is made up of various models, including:

• The fixed-function EX4200 line with Virtual Chassis technology and the chassis-based EX8200 line, targeted at both data center and campus applications.

• The EX2500, a fixed-function switch from Blade, designed for top-of-rack data center deployments.

• The EX4500 line of fixed-function 10-Gigabit Ethernet switches designed for data center top-of-rack and aggregation deployments. The EX4500 is also designed to support Virtual Chassis technology.

The EX Series switches utilize the same Junos operating system used by Juniper's security and routing products, potentially reducing complexity and operational costs.

Since February 2009, Juniper has been working with IBM to jointly develop Project Stratus, a flat, single-layer cloud. Project Stratus is a long-term strategy to develop a single data center fabric with the flexibility and performance required to scale to support large data centers, while continuing to drive down the cost and complexity of managing the data center information infrastructure. Project Stratus is planned to begin shipping in 2011.

Juniper's Virtual Chassis technology creates a network fabric that reduces or eliminates switching tiers, improving space, power, cooling and management efficiencies. With Virtual Chassis technology, up to 10 EX4200 switches can be interconnected and managed as a single, logical device. Although collectively the member switches behave as a single platform, each individual device has its own power supply and fan tray, ensuring high availability. The EX4200 can be deployed as a top-of-rack or end-of-row solution in a Virtual Chassis configuration, which connects directly to a pair of chassis-based EX8200 core data center switches, thus collapsing the aggregation and core layers into a single tier.

In the future, the EX4500 and EX8200 also will support Virtual Chassis technology, further strengthening them for the data center.

Table 5 outlines Juniper's strengths and weaknesses.

Blade is a supplier of Gigabit Ethernet and 10-Gigabit Ethernet network infrastructure solutions that reside in blade servers and server and storage racks. Blade sells its own RackSwitch top-of-rack switches and supplies blade server switches to manufacturers such as HP, IBM and NEC. Its value proposition centers on lossless, low-latency, low-cost and low-power Ethernet I/O infrastructure solutions with "virtual-machine-aware" network virtualization that works with hypervisors from VMware, Microsoft, Oracle Citrix and KVM, as well as switch management.

Blade is completely focused on the data center market. In 2009, it accounted for 17.4% of total ports shipped and 4.3% of the revenue in the data center Ethernet switch market, achieving the second position among the top vendors for shipments, and the fourth position in revenue.

Blade's portfolio includes Gigabit Ethernet and 10-Gigabit Ethernet switches for the data center, in both blade server and top-of-rack form factors. It shipped its first blade server switches for IBM BladeCenter in 2003. Blade also sells its own branded products, such as its RackSwitch top-of-rack switches, and the first one started shipping in 2008. Blade's subsidiary, DataCenter Technologies, provides Advanced Management Modules for the IBM BladeCenter.

Blade pioneered switch airflow that matches server airflow, so as to increase cooling efficiency and lowering costs. Its RackSwitches include specialized software to support server virtualization in the network, called Blade VMready. It resides within the switch — meaning that no additional software is needed in the server — and enables network operators to monitor and control traffic from each individual virtual machine, and automatically move policies that each virtual machine requires, such as quality of service, VLANs and access control lists, for security. BLADEHarmony Manager provides easy-to-use switch management that complements management systems, such as IBM Systems Director and HP Systems Insight Manager.

One of its main differentiations is its Unified Fabric Architecture (UFA). UFA provides a virtual-machine-aware networking platform that allows customers to tie together best-of-breed servers, storage, hypervisors and core networking systems. Blade's solutions are open and compatible with existing network infrastructures. For a vendor with a restricted portfolio, it provides a credible alternative to compete with larger end-to-end vendors like Cisco and HP.

Table 6 outlines Blade's strengths and weaknesses.

Force10 Networks offers high-performance solutions to enterprises, data centers and service providers. Its solutions are designed to deliver new economics by virtualizing and automating Ethernet networks.

Force10 Networks has a strong focus on the data center market, with its 2009 data center Ethernet switch revenue standing at $64.8 million, for a market share of 3.3%.

Force10 Networks has been delivering switches for data center deployments for nearly a decade, beginning with the EtherScale platform that was launched in 2002. It has shown strong commitment to the data center market and a good vision, centered around virtualization and open automation framework. The vendor's products are purpose-built for offering standards-based solutions optimized for data center applications, from top of the rack to core.

They include:

• ExaScale E-Series — Chassis-based switch/routers, launched in April 2009. Optimized for core deployment on large data centers as well as end of row in high-density copper server aggregation

• C-Series — Midrange chassis-based switches, with very low latency across converged networks. Can be used for end-of-row deployment or in the core of small to midsize data centers

• S-Series — Small form factor fixed Layer 2 switches, can be deployed as line rate Gigabit Ethernet and 10-Gigabit Ethernet top-of-rack switches for data center, storage or compute facility

Force10 Networks also plans to launch in 2H10 an FCoE/DCB-enabled top-of-rack solution.

Table 7 outlines Force10 Networks' strengths and weaknesses.

Brocade, the market leader in the Fibre Channel space and a pioneer in the SAN market, acquired Foundry Networks in December 2008. Foundry Networks was a market leader in the Gartner Magic Quadrant for the campus LAN market in 2008, thanks to its complete product line, excellent customer support and financial strength. It has enjoyed good traction among large enterprises in the past, with a high-end and high-performance portfolio. Foundry's first data center switches were launched in 1997.

Brocade accounted for 2.8% of total shipments of switches to data centers in 2009, with 285,500 ports, and associated revenue of $51 million, or 2.6% of the total, giving it the sixth position in the revenue ranking for data center Ethernet switches.

Brocade offers data center and campus network solutions that provide open choice and high efficiency at a low total cost of ownership.

Brocade's key differentiation includes price for performance, scale and its software — routing protocol support. On top of its switches, it also boasts the Brocade ServerIron, which is a fixed and modular application delivery controller platform that controls traffic and optimizes application delivery from the data center, which is another main differentiation.

Brocade portfolio includes the following products:

• NetIron MLX Series and FastIron Super X family, which are both chassis-based switches and routers, targeted at the data center, campus and service provider networks.

• TurboIron 24X switch is a compact, high-performance, high-availability and high-density 10-Gigabit Ethernet/1-Gigabit Ethernet dual-speed solution that meets data center top-of-rack requirements.

• FastIron Edge family, targeted at both the data center and campus networks.

• FastIron Edge X family and FCX family, both fixed-function switches targeted at the data center.

Brocade also has an FCoE product: the Brocade FCOE10-24 Blade, which is a Layer 2 blade with cut-through nonblocking architecture designed for use with Brocade DCX and DCX-4S Backbones. It features 24 10 Gbps CEE ports and extends CEE/FCoE capabilities to Brocade DCX Backbones, enabling end-of-row CEE/FCoE deployment. Also, supporting FCoE, the Brocade 8000 is a top-of-rack link layer (Layer 2) DCB/FCoE switch with 24 10-Gigabit Ethernet ports for LAN connections and eight Fibre Channel ports (with up to 8 Gbps speed) for Fibre Channel SAN connections.

Table 8 outlines Brocade's strengths and weaknesses.

Avaya is a player in enterprise communications systems, providing unified communications, contact centers and related services. Avaya re-entered the data networking market through its acquisition of Nortel's Enterprise Solutions business in December 2009. Avaya subsequently confirmed that it would adopt the current road map of Nortel Enterprise Solutions data products in its entirety.

In 2009, Nortel sold 350,000 ports that were deployed in the data center for associated revenue of $43 million. That gives the vendor a share of 3.5% for shipments and 2.2% for revenue in the data center Ethernet switch market.

Avaya's portfolio includes both chassis-based and fixed-function switches:

• ERS 8600, a core or aggregation chassis-based switch, which has been shipping for over 10 years, with the ERS 8800 as an enhancement to the 8600 family.

• VPS 9000, a core chassis-based switch, based on the ERS 8600, was announced in 2009 and will ship in 2H10.

• ERS 8300, an edge chassis-based switch.

• ERS 5600 (introduced in 2009) and 5500 (2007), stackable switches for the edge, that can be deployed as end of row and/or top of rack.

• ERS 4500, stackable switch for the edge.

One of Avaya's main differentiations is its Split Multi-Link Trunking technology. It provides subsecond resiliency across its portfolio from server access to core. It is using this technology in the VSP 9000 with switch clustering, and this platform is designed to be a lossless unified fabric for FCoE/DCB.

At the moment, Nortel doesn't have a data-center-specific offering, though it is in the process of introducing one. Its portfolio still lacks many data-center-specific features.

Table 9 outlines Avaya's strengths and weaknesses.

We chose to present two challengers. Even though they have smaller market shares, they have demonstrated an interesting vision or product road map.

Extreme Networks is a relatively small networking vendor that aims to deliver high-performance and innovative products for both the campus LAN and data center markets.

Its data center portfolio includes top-of-rack, fixed-configuration and chassis-based switches. The top-of-rack switch, the Summit X650, was released in January 2009; the BlackDiamond 8900, which provides high-density 1-gigabit and 10-gigabit chassis, was released in May 2009; and the BlackDiamond 8900-xl series modules and Summit 480 switch were introduced in January 2010.

In the past 18 months, the company has expanded its data center Ethernet switch portfolio significantly, and it now includes:

• BlackDiamond 8800 — A chassis-based switch that can be deployed in the core/aggregation for small and midsize data center, or at the edge for the midsize and large data center. It can also be used as an interconnect switch providing low-latency connections for High Performance Computing Cluster and data center.

• BlackDiamond 8900 — Chassis-based purpose-built network switch for next-generation data centers. The BlackDiamond 8900 modules deliver high scalability for 10-Gigabit Ethernet.

• Summit X450a 48 — A fixed-function switch that provides a high-bandwidth, nonblocking architecture with trispeed copper Gigabit Ethernet ports for top-of-rack data center applications.

• Summit 480 — A fixed-function high-end Ethernet switch for data center, enterprise aggregation and carrier Ethernet deployments.

• Summit X650 — A purpose-built top-of-rack switch designed for emerging 10-Gigabit-Ethernet-enabled servers deployed in enterprise data centers.

By combining the Ethernet switch SummitStack stacking technology with the Summit X650, Extreme has created a high-density 10-Gigabit Ethernet stackable solution. The Summit X650 switch offers a flexible stacking option to optimize multiple deployment scenarios, providing backward compatibility with the 40-gigabit SummitStack and ultrahigh-speed SummitStack256 running at 256 Gbps. With SummitStack, Summit X650 can be stacked up to eight units, providing up to 192 ports of 10-Gigabit Ethernet in an 8-rack-unit height.

In April 2010, Extreme Networks released the Extreme XNV. It delivers life cycle management capabilities, keeping the network in control as virtual machines move from their creation point to their end of life. It is also a component of Extreme's Four Pillar strategy addressing the evolution of the data center, a strategy that enables networks to migrate from physical to virtual to cloud network environments.

Extreme Networks supports the reduction of network tiers in the data center, and it was the first vendor to announce support of VEPA-ready network hardware that will transition customers to VEPA standards once ratified. It also supports network virtualization integration and automation and customization through scripting and XML-based interface. Its platforms are also ready for 40-gigabit and 100-gigabit support. However, it does not support FCoE at the moment, because the company sees it a couple of years away from market adoption.

Table 10 outlines Extreme Networks' strengths and weaknesses.

Arista Networks is a relatively small privately held company focused on the data center switching market, but it is founded and led by Andy Bechtolsheim, co-founder of Sun Microsystems. Arista started shipping its first 1/10-gigabit switch in 2H08. In the past 18 months, Arista has introduced six more data center switches.

Arista has had some initial success with high-end U.S.-based financial institutions, universities, and high-performance-computing customers like scientific organizations, with a bulk of its revenue coming from these segments.

Arista's hardware platforms are purpose-built for the data center, with low latency and very high density in mind. Based on merchant silicon and custom-designed, its switches offer maximum throughput per square foot and optimum functionalities and features at a very low price. Arista's products include the following:

• 7100 series — A family of fixed-function switches with either 24 or 48 ports. It supports a variety of interfaces, including 1-gigabit/10-gigabit optics, SFP+ and 1/10-gigabit BASE-T.

• 7500 series — A family of chassis-based switches designed to support current 10-gigabit requirements and future deployment of 40 gigabit/100 gigabit.

• The 7048 is a top-of-rack wire speed Layer 2/3 switch with 48 1-gigabit ports and four 1/10 Gigabit Ethernet uplinks delivering up to 40 gigabits of interconnect capacity.

At the heart of Arista's offering is its Linux-based EOS operating system, which can be deployed not only across its range of switches but also in an external appliance or a virtual machine. Arista believes that software is where it can innovate and differentiate from others. The vEOS is one of the innovations that places the company in the unique position of having a robust virtualization offering that is extremely flexible in deployment choice and equally capable of topology discovery, visualization and support of the virtual switch environment.

Table 11 outlines Arista's strengths and weaknesses.

The findings and insights in this report are based on several information sources and modes of research. The backbone of this research framework was a detailed survey that Gartner conducted along the length and breadth of the enterprise networking industry and also IT vendors known to have data center Ethernet switch offerings. This survey was conducted during February and March 2010. Subsequently and in parallel, Gartner also conducted one-to-one phone briefings with almost all respondents who replied to our survey to clarify and get further information.

Gartner's evaluation of the market and the competitors is based on the above primary research, as well as our ongoing quarterly enterprise Ethernet switch market statistics and forecasts. They are backed by extensive secondary research drawn from both Gartner and external sources.

The final profiles published have also gone through an additional round of external review in the hands of the vendors considered for profiling, to check for any factual errors. The insights given, however, remain the independent views and comments of Gartner, based on its objective and unbiased research.