Over a year ago, I posed the question, “Does 10Gb Ethernet change the Competitive landscape?” Cisco has been the dominant player in networking, for over a decade no competitor ever captured even ten percent of the market. While Ethernet is continuing its march into new markets and new applications, the market landscape has definitely changed. Fresh off of VMworld, there is a buzz in the networking world around new opportunities and architectures.
The Big Trends
Industry consolidation of networking vendors has left the market with fewer, stronger players. Several of the smaller independent switch vendors have been acquired by server vendors, specifically 3COM (HP), BNT (IBM) and Force10 (Dell), giving the server vendors the ability to create integrated stacks of internal IP as well as pushing Cisco out of pre-configured rack solutions. A quick rundown of the stacks:
- Cisco has compute (UCS) and networking (Nexus) product lines that it integrates with EMC (VCE joint venture) and NetApp (FlexPod reference architectures)
- HP has a number of offerings including Matrix, CloudSystem and VirtualSystem which leverage HP servers, HP storage (LeftHand, 3PAR and IBRIX) and HP networking (primarily Procurve/FlexFabric products in the stack)
- Dell has vStart systems which integrates Dell servers, Dell EqualLogic Storage and Dell PowerConnect networking (the Force10 acquisition recently closed, so expect further integration here)
Server virtualization requirements have rippled into the networking space. Consolidated infrastructure requires higher bandwidth networking and more flexible traffic patterns, which are driving new architectures to flatten the network. Every vendor has a different blueprint for building a fabric, built with various switching products and a collection of standards, software and management. Some of the characteristics that make up these new “fabric” solutions:
- Greater utility of networking by supplementing or eliminating Spanning Tree Protocol (STP). STP limited devices to a single path/connection between each device and eliminating this allows for the “flattening” of the traditional 3-tier architecture (access, aggregation, core) to 2-tier. Core Ethernet switching vendors have various options for solving this problem including various L2 multipathing options (see wiki entry) and switch clustering options (Cisco vPC, HP IRF, and Arista MLAG allow multiple physical switches to be managed as a logical single switch).
- Flexibility of the solutions to integrate with both legacy environments and the ability to support new applications for Ethernet such as storage (reliability), HPC (low latency) and the varying requirements of service providers
- While Ethernet is built for scalability, large scale environments are finding new requirements for manageability and programmability that have led to Arista’s Extensible OS (EOS), Juniper’s Junos and the OpenFlow initiative
- Mobility is a bit of a wildcard in the virtualization space. The promise of moving any application anywhere is a good vision. VXLAN was talked about a lot at VMworld (in addition to videos below, this video discusses both VXLAN and virtual networking with Sujal Das of Broadcom and Jason Nash of Varrow)
While I have yet to hear an end-user ask for “fabric architectures”, there are clear customer requirements for deployment of the traits listed above. [Wikibon note: What features or characteristics make up a network fabric? Comments and suggestions welcome]
Networking Vendor Roundup
While I’m a fan of multi-media, I think that the least interesting part of Gartner’s recent Magic Quadrant for Enterprise LAN is the picture. There are some excellent data points in the research (find the August 2010 report here). I’ll add some of my comments (also see the video below for more of Wikibon’s analysis of this space)
- Cisco still has the broadest portfolio, supporting from virtual switching embedded in VMware and Hyper-V, and numerous rack and core switching options (Gartner sees this as a weakness, while I believe that Cisco is blocking any gaps or niches to parry competiton)
- Despite growth by competitors, Cisco still holds dominant market share (recent reports for Ethernet switching put Cisco between 65-75%). Cisco also claims that margins have stabilized in the last quarter (disputing that competitors will commoditize Cisco out of the market)
- Cisco’s biggest strength is in its broad distribution channel and large number of certified customer and partner engineers
- Cisco’s CEO John Chambers pledged this summer that the company will work to be “easier to do business with”. Cisco is often been attacked for “locking-in” customers, charging high prices and providing poor support – these “soft” issues can be much more dangerous than any technology gaps.
- While competition has laughed at Cisco’s exiting of “adjacent” markets, networking engineers have cheered Cisco returning to greater focus on data center and enterprise needs.
VIDEO: Cisco’s Soni Jiandani at VMworld discussing the networking for a virtual environment.
- HP’s solid #2 position in the marketplace (up to 12% Q2 sales) is made up of a variety of product lines including the embedded FlexFabric solutions, Procurve and 3COM products.
- HP has delivered innovations in its embedded technologies (such as the Flex-10 and FlexFabric Virtual Connect), but has yet to leverage that in other product lines. HP also needs to articulate how the product lines fit into a portfolio.
- HP’s biggest strength in the networking space is aggressive pricing. While cost competition is very effective for embedded and bundled/converged offerings, core networking is driven more by risk-aversion and meeting application requirements than saving money.
- HP is doing a good job converting HP server and storage customers to HP networking. Pickup in non-HP shops has not yet been proven, but even so, HP has plenty of growth opportunity in its base.
- HP is leveraging strong services to ensure that customers that leave Cisco are happy with the decision
VIDEO: HP’s Mike Banic at VMworld discusses how cloud and virtualization impact networking
Juniper and Brocade
Both of these companies are independent providers of full networking offerings from edge to core.
- Similar to Cisco, server partnerships are under pressure
- Juniper still has a strong connection with IBM for core switching. Dell’s Force10 acquisition is expected to impact Juniper sales.
- Brocade is still the market leader in FC with strong server partnerships, but has not gained significant traction with its OEMs with Ethernet (Foundry) solutions
- Both companies are showing strong growth in Ethernet switching with around 2% market share each
- Juniper’s QFabric solution has garnered a lot of buzz in the industry. While the solution clearly articulates the market challenges and lays out a compelling story, it only started shipping this week and therefore it will take time to determine how well the vision translates to the reality of customer environments.
- While Brocade’s VCS fabric architecture has been shipping for over a year and Ethernet sales now account for over a quarter of Brocade’s revenue, the storage (FC) and Ethernet portfolios, customers and sales channels have are don’t intersect as much as might be expected. HP, Cisco and others have integrated FC and Ethernet products more than Brocade.
VIDEO: Brocade CEO Michael Klayko at EMC World discussing the FC/Ethernet dynamic.
VIDEO: Juniper’s Abner Germanow at VMworld discussing the changing dynamic of the networking marketplace.
It a bit odd that Gartner lists a number of smaller players, but Arista does not make the list. In general conversations at VMworld, I heard from both customers and channel partners that Arista is gaining traction.
- Arista’s products have a good mix of high performance, programmability, virtualization affinity and low cost
- While Arista is similar to Brocade and Juniper that it is not being delivered in server converged environments, Arista’s go-to-market goes after Cisco’s traditional channel model.
- Arista had good success in the HPC marketplace and has been moving into Web 2.0 and service providers accounts, which leverage the EOS /vEOS programmable systems.
VIDEO: Arista CEO Jayshree Ullal at VMworld talks about how deploying a fabric architecture is an imperative for CIOs and how virtualization is making the networking market sexy again.
Dell: In storage and now in networking, Dell has been moving from third party solutions to in-house, higher margin options by making strategic acquisitions. See previous articles on the Force10 acquisition and why Brocade was not a good fit. Dell now has a solid networking portfolio to fit the markets that it serves.
IBM: IBM has shown little interest in competing in the “stack wars” and while it does have significant networking revenue, IBM still partners with Cisco, Juniper and Brocade and its BNT products are even used by HP. IBM at its heart is a services organization which allows the company to have a unique position in the alignment of the networking industry.
Others: In addition to some smaller niche players, there are a number of startups that are looking to “do for networking what VMware did for servers” and intersect with the OpenFlow intiative such as Nicira and Big Switch Networks.
While CIOs should begin by investigating new stack and fabric architectures. Stacks and fabrics are both significant investments, in general, stacks are faster to deploy and more mature than fabrics. Business units and IT staffs should be working together to understand the various applications and growth requirements as well as the impact of working with hosting or service providers. Wikibon will continue to share guidance from the IT practitioner community and expand on the requirements and vendor landscape as discussed in this article – feedback and debate is always welcome.
Stu Miniman is a Principal Research Contributor for Wikibon. He focuses on networking, virtualization, Infrastructure 2.0 and cloud technologies. Stu can be reached via email (firstname.lastname@example.org) or twitter (@stu). He will be attending Interop in NYC, October 4-6 where he will be debating with Stephen Foskett on iSCSI vs. FC – details here.