With David Floyer
Wikibon has been covering convergence for several years. While significant and legitimate customer concerns exist about converging disparate networks, we continue to believe that long-term, technology convergence around Ethernet is inevitable, with many customers expressing meaningful interest today in emerging converged infrastructure offerings. The notion of buying logical blocks of compute, storage, and network infrastructure is picking up steam, and the idea of converging discreet networks (e.g. SAN and LAN traffic) continues to make sense over time for many organizations.
In this regard the Wikibon community took note today of Cisco, which announced the delivery of several capabilities that have been on its roadmap for many months if not years. Cisco is adding FCoE support to the core Nexus 7000 (where it is Ethernet only) and MDS (an FC switch which adds an Ethernet module). Specifically, the company announced support for multi-hop FCoE, which enables end-to-end Ethernet (in theory up to seven hops* but in practice more like five) in these high-end enterprise switch offerings. The importance of this announcement from a user perspective is not so much the availability of new switches but more the strategic implications of the architecture vis-a-vie the topic of convergence.
Two things matter most with this announcement in our view:
- Cisco has enabled the delivery of a more expansive end-to-end FCoE with greater flexibility. Previously, FCoE was only available in embedded (UCS) or ToR (Nexus 5000 series). By delivering on the vision of end-to-end FCoE, Cisco is confirming our previous assertions that FCoE is strategic with long-term viability in the data center.
- Most importantly, the existing Cisco functionality of Virtual Device Context on the Nexus 7000 (core switch), which provides the ability to partition a switch, will allow administrators to isolate different traffic flows within a single switch (e.g. SAN vs. LAN traffic).
The implication of this virtual switch functionality means organizations can logically carve out any part of a network as a separate resource and manage it independently of any other part of the network. For example, data people can look after SAN traffic, and networking people can manage the LAN independently using the same switching infrastructure. The addition of FCoE is a key enabler of this feature because the base infrastructure is Ethernet, meaning common pipes can now be shared for SAN, LAN, and server-to-server traffic.
In addition, the emergence and increasing popularity of converged network adaptors (CNAs), which are now into their third generation, significantly adds to the benefits of converged networks. Specifically, the inclusion of CNAs within server infrastructure enables a single connection out of the server box. This means greater server density, better exploitation of server and network real estate, and ultimately lower connection costs.
This capability somewhat addresses the concerns raised by many Wikibon members regarding the need to reorganize teams to exploit converged networks. The concern among many users was that managing converging physical networks using a unified management framework would require merging separate SAN and LAN management organizations. While alluring on paper, this approach would cause organizational dislocation. For example, today, network managers often report up to a different senior management structure than SAN administrators. The converged network vision would potentially require a brand new management hierarchy with revised priorities, incentives, and the like; causing the potential for significant political friction.
Under Cisco’s new approach, SAN and LAN administrators can continue to manage their networks with independence and the latitude to optimize resources according to their own specific priorities. This is a rare case within enterprise technology where having separate administrative functionality is actually desirable. While many vendors often tout the advantages of a single management framework, in this case those benefits, we believe, would be largely illusory.
It is important to note that Cisco has enhanced its Data Center Network Manager (DCNM) to allow a single pane of glass between LAN/SAN to enable a single administrator to view the converged network. This capability in our estimation should primarily be used for strategic resource allocation, not day-to-day problem determination and management. Capabilities such as this should be viewed as a vehicle to facilitate organizational discussions and alignments between different network groups (e.g. data and user network organizations).
Cisco has had a near monopoly on high-end switches for years. However, the company has seen a need to diversify into new markets in order to grow in the enterprise, specifically targeting server and storage switching markets as well as server-to-sever communications. SAN switches in particular have been dominated by Brocade, which has built a multi-billion dollar market. Cisco’s counterpunch strategy has been to push Ethernet as hard as possible and dislocate traditional FC, which has been an uphill battle. FCoE potentially changes the game because rather than fighting FC, Cisco can co-opt it into its Ethernet strategy.
Recently, as a direct result of its aggressive competitive moves, Cisco has been under fire as its unified computing strategy generally and entrance into server markets specifically has created cascading effects throughout enterprise IT and attracted more competition in its core business (e.g. HP/3Com and to a lesser extent Brocade/Foundry). Cisco’s convergence strategy is an attempt to secure its installed base, shore up its defenses against competition and aggressively move into new server and SAN switch markets under the converged network umbrella with Ethernet as the underpinning. Cisco is emphasizing partnerships with the likes of Intel, VMware, EMC, VCE and NetApp among others and prioritizing these over historical relationships with traditional enterprise server vendors (e.g. HP, IBM, Dell and Oracle/Sun). Cisco’s strategy is designed to capitalize on Intel's server chip dominance and VMware momentum. Furthermore, Cisco has successfully persuaded the storage vendors to align with its Ethernet-led converged network strategy, and companies like EMC and NetApp have provided considerable access to their respective installed bases. Meanwhile, the traditional server vendors have no choice but to push convergence via CNAs because they’re out of real estate.
Many Wikibon members have expressed legitimate concern about converging networks. Specific issues have related to performance, cascading network dependencies (e.g. LAN vs. SAN issues), organizational tensions, questionable near-term ROI and the general sense that other IT issues are more pressing than the move to converged networks, at least in the near term.
Specifically as it relates to this announcement, critics will point out that what Cisco is really doing here is delivering on its promises of previous roadmaps. While this is largely true, end-to-end FCoE delivery is important to adoption as thus far proof points have been limited, and this announcement legitimizes the strategy and reduces customer risk within the Cisco installed base. The major customer concern is that organizations are locking themselves into Cisco.
To that end, our advice has consistently been to frame convergence as follows:
- Plan for Ethernet dominance.
- Treat FCoE as an evolutionary technology that should be introduced in phases and ultimately lowers cost and risk while maintaining the FC advantages for the SAN.
- In the near term, there has been no incentive to rush into FCoE although the inclusion of CNAs within server infrastructure is sensible. In the mid-term, as cloud computing and IT-as-a-Service take off, organizations will increasingly be buying blocks of compute, storage, and networking as a single logical resource and FCoE is a key enabler for cloud workloads requiring high performance and availability.
- Cisco and its partners are highly motivated to develop end-to-end proof points. Now is a great time to negotiate, especially for organizations willing to be early references. Consider only entering into substantial agreements with Cisco and its partners that include maximum long-term flexibility and attractive maintenance pricing with multi-year discounts and caps.
- While the push to converged networks is largely today coming from the vendor community, we believe it is inevitable from a technology standpoint, long-term.
Getting on the curve with early pilots will position organizations for fast adoption when the time comes which will accelerate the pace at which organizations can reduce connection costs.
Action Item: Cisco’s recent announcement marks the first time that an end-to-end converged infrastructure based on Ethernet, with multiple switch options, is in play. This is another milestone in the widespread move to convergence and supports mission-critical cloud computing in the enterprise. This announcement is alluring to Cisco customers because it allows them to evolve converged infrastructure with confidence and without a rip-and-replace. Customers that expect to follow the Cisco roadmap should negotiate hard now and lock in long-term flexibility and competitive pricing, especially maintenance pricing. Customers who are reticent to follow the Cisco roadmap at this time should gain visibility and confidence that competitors’ plans will match Cisco’s capabilities.
Footnotes: *Note: Multihop from Cisco is up to 7 hops, but not all OEMs support that (e.g. EMC only supports 5 hops). The hops is the total number of FCFs (the FCoE equivalent of an FC switch) and FC switches, so you might go 2 hops FCoE and 3 hops FC = 5 hops.