David Vellante with Phil Little, David Floyer and Dennis Martin
This is a summary of a discussion Wikibon led in collaboration with On Magazine, EMC's CIO publication.
This is Part 3 of a three-part series.
- See Part 1 Technology Convergence: Transforming data center infrastructure.
- See Part 2 Futureproofing Converged Network Architectures
Summary
The hypothetical business impact of converging separate LAN, SAN, and other networks is substantial and could cut organizations’ network costs by up to 40 percent. However IT organizations (ITOs) will encounter substantial friction related to disagreements over the likelihood and timing of realizing costs savings, an aversion to adding more complexity in the network, and concerns over interoperability.
This was the conclusion from a discussion with a business technology professional, Phil Little, Senior Storage Architect at Pacificorp, which is an energy provider to six western states; and two industry analysts.
While the conceptual savings of converging network traffic onto a single pipe are compelling on paper, practitioners are wary of adding more complexity to the network, specifically as it relates to supporting another protocol, namely FCoE. A key issue is interoperability and the barriers this will present to making changes to the network.
Given these adoption challenges, ITOs should expect a slow phase-in for network convergence taking place in stages over many years. Nonetheless, there is strong agreement that Ethernet generally and specifically lossless (Enhanced) 10-Gigabit Ethernet is an ideal technology for converging block and non-block storage assets over time, regardless of protocol choice. It’s getting there that will be the challenge.
The ROI of Convergence
Despite the concerns over converged network adoption, CIO’s and infrastructure executives need to understand and assess the key benefits of network convergence. Three main business drivers, all related to costs, are spurring investments in converged networks:
- Converging separate LAN, SAN, and system-to-system network traffic onto a single 10-Gigabit Ethernet pipe will, over the long run, reduce costs substantially by reducing the number of network connections and saving on power, cooling, and floor space.
- Deploying converged network adapters (CNAs), which merge Ethernet and SAN cards into a single adapter, will reduce costs by halving the number of cards required. Here’s a basic introduction to FCoE and the idea behind CNAs.
- Network convergence will reduce the number of ports required in the switching infrastructure, which will decrease switch costs dramatically and give economies-of-scale by allowing lines to be shared.
While this scenario will take many years to play out due to adoption constraints, from a CIO perspective the potential to save money and improve environmentals will be a substantial driver, especially in large organizations. Timing the initiative remains a key CIO issue. The answers depend on costs associated with convergence and whether convergence is an immediate business imperative or whether the ITO can wait until the ecosystem matures, decreasing risks.
The Costs of Convergence
To be sure, convergence has costs. Many however, such as the move to 10-Gigabit Ethernet, can be absorbed as part of the natural evolution of networking technology, over several years. As well, organizations should not rip and replace existing server network adapter cards with CNAs. Rather they should ensure that new servers can support CNAs and begin conducting low-cost pilots and proof-of-concept (POC) efforts to test the viability of convergence and gain experience.
A key issue for ITOs is does network convergence generally and FCoE specifically add more complexity to the IT operation. The answer in the near term is unquestionably, yes. The main issue here is getting from point A (where you are today) to convergence. Organizations will have to tip-toe through a minefield of complexity in order to achieve nirvana, and a key sticking point will be interoperability of servers, switches, and end-point devices.
In particular, current network infrastructure, particularly SAN infrastructure, is complex. There are equal levels of complexity in LAN infrastructures with completely different processes and operational procedures from SANs. By introducing new technologies such as FCoE and converging SAN and LAN networks, ITOs will introduce cascading dependencies throughout the network that, in the near-to-mid term, will make managing the network trickier. For example, an upgrade to the SAN may now have a direct dependency on components of the LAN that the network team just can’t deal with in near real time.
While these issues can be addressed over time, in the near term, ITOs must carefully consider these complexities and specifically interoperability issues, which have not been fully vetted by the industry.
Considerations for IT Organizations
The storage and networking industries are pouring billions into R&D around convergence to meet the demand to decrease network costs and improve overall network quality and reliability. The ecosystem is lining up behind giants including Cisco, EMC, HP, IBM, Intel and dozens of smaller companies, which increases the probability that convergence will happen eventually, despite the near term complexity challenges.
To manage this complexity, ITOs should consider three strategies, including:
- Begin with pilots, carving out small portions of networks and testing converged technologies (e.g. Enhanced Ethernet and FCoE) in safe environments.
- Initially, keep major portions of networks physically separate during a transition phase. By keeping this physical division in place for broad parts of the network, some cost savings can be realized (e.g. a reduced number of cards in servers and lower port consumption) while minimizing interoperability and management risks across the network.
- Sort out organizational responsibilities. Networking and storage teams often report to different leaders with varying incentives designed to affect behavior. Begin laying out broader organizational goals and include major stakeholders from the LAN and SAN groups. Develop sets of metrics that are relevant to converged environments that are owned by converged teams.
The timing of convergence initiatives will depend on the size of the organization, the complexity of existing networks, a firm’s reliance on FC and the degree to which FCoE adoption can bring returns and the timing of those returns. In any event, the adoption curve for converged networks will take many years but in general will hit a breakthrough once critical mass is achieved (i.e. the base infrastructure is in place and ROI is virtually instantaneous).
Action Item: Economics are driving convergence and while the choice of storage protocol may vary, with some organizations leaning toward iSCSI and others maintaining an FC/FCoE or mixed iSCSI FCoE infrastructure, consolidating network traffic onto a minimized physical infrastructure will, in the long run, reduce complexity and shave substantial costs. Organizations must begin preparing by setting goals, piloting convergence and initiating the organizational changes that will come to fruition.
Footnotes: