- A TechNote on The Next Generation
- Jim Metzler
- Distinguished Research Fellow and Co-Founder
- Webtorials Analyst Division
Over the last couple of years, the trade press has devoted a great deal of attention to integration in the enterprise data center. Reading this coverage, you may have wondered whether an integrated data center is in your near-term future. The answer depends on the form of integration at hand.
One type of integration that gets a lot of attention is the convergence of the data center LAN and the storage-area network (SAN). To gauge the interest in this form of integration, attendees at the October 2011 Interop conference were surveyed about the degree to which their companies would likely deploy a converged LAN-SAN fabric over the next two years. Of the 244 responses, slightly more than half indicated that their organizations will undertake at least a moderate degree of LAN-SAN convergence within that period.
However, given some of the challenges associated with converging the LAN and the SAN, those survey results might be somewhat optimistic.
Convergence Deja Vu
A curious aspect of the ongoing trade-press discussion about LAN-SAN convergence is that it looks remarkably like the discussion that took place over a decade ago about the convergence of voice and data networks. For example, LAN-SAN convergence proponents typically state that having a single integrated network costs less than having two separate networks. That same argument was used to justify the convergence of voice and data networks.
The mechanics of adopting a converged LAN-SAN fabric likely will also mirror those of voice-data network convergence. For example, when VoIP was first being deployed, few IT organizations ripped out their existing voice networks just to implement a combined voice and data network. Instead, they implemented VoIP in response to an event such as opening a new office or making a needed major upgrade to their PBX's software. Similarly, most IT organizations won't rip out existing networks solely to implement a converged LAN-SAN fabric; rather, they will implement the converged fabric as part of redesigning their data center LANs.
'Yes' on Data Center LAN Redesign - 'Iffy' on How
A recent Webtorials research report on cloud networking contains a section on "The Emerging Data Center LAN." The report indicates that the majority of IT organizations already have begun the process of redesigning their data center LANs. This would seem to indicate that the adoption of a converged LAN-SAN fabric will quickly become mainstream.
However, the report further points out that there is considerable uncertainty within IT organizations about exactly how they will redesign their data center LANs. They are unsure, for example, whether or not to replace the spanning-tree protocol and, if they do, which technologies to replace it with.
Likewise, most IT organizations aren't sure how many layers their new data center architecture will have. Will it be the traditional three-tier architecture comprising access, distribution and core switching layers? Or do emerging technologies enable implementation of a two-tier - or even a single-tier - architecture?
Technical Issues with LAN-SAN Convergence
Beyond answering these general questions about data center LAN design, IT organizations need to address some of the technical challenges associated with converging the LAN and the SAN. For example, traditional Ethernet's best-effort service allows buffers to overflow during periods of congestion and relies on upper-level protocols such as TCP to manage congestion and recover lost packets through re-transmissions.
In an integrated LAN-SAN, this could result in a level of delay that is unacceptable for storage. There are, however, emerging technologies that enable Ethernet to support a converged fabric, which will be discussed in a future TechNote.
As discussed in the previously referenced research report, there are strong factors driving IT organizations to redesign their data center LANs. However, the fact that some of the enabling technologies are fairly new to the market - and some are even still in development - means that data center LAN redesign will take place slowly over a number of years. As a result, the adoption of a converged LAN-SAN fabric will also be a gradual, years-long process.
One type of integration that gets a lot of attention is the convergence of the data center LAN and the storage-area network (SAN). To gauge the interest in this form of integration, attendees at the October 2011 Interop conference were surveyed about the degree to which their companies would likely deploy a converged LAN-SAN fabric over the next two years. Of the 244 responses, slightly more than half indicated that their organizations will undertake at least a moderate degree of LAN-SAN convergence within that period.
However, given some of the challenges associated with converging the LAN and the SAN, those survey results might be somewhat optimistic.
Convergence Deja Vu
A curious aspect of the ongoing trade-press discussion about LAN-SAN convergence is that it looks remarkably like the discussion that took place over a decade ago about the convergence of voice and data networks. For example, LAN-SAN convergence proponents typically state that having a single integrated network costs less than having two separate networks. That same argument was used to justify the convergence of voice and data networks.The mechanics of adopting a converged LAN-SAN fabric likely will also mirror those of voice-data network convergence. For example, when VoIP was first being deployed, few IT organizations ripped out their existing voice networks just to implement a combined voice and data network. Instead, they implemented VoIP in response to an event such as opening a new office or making a needed major upgrade to their PBX's software. Similarly, most IT organizations won't rip out existing networks solely to implement a converged LAN-SAN fabric; rather, they will implement the converged fabric as part of redesigning their data center LANs.
'Yes' on Data Center LAN Redesign - 'Iffy' on How
A recent Webtorials research report on cloud networking contains a section on "The Emerging Data Center LAN." The report indicates that the majority of IT organizations already have begun the process of redesigning their data center LANs. This would seem to indicate that the adoption of a converged LAN-SAN fabric will quickly become mainstream.
However, the report further points out that there is considerable uncertainty within IT organizations about exactly how they will redesign their data center LANs. They are unsure, for example, whether or not to replace the spanning-tree protocol and, if they do, which technologies to replace it with.
Likewise, most IT organizations aren't sure how many layers their new data center architecture will have. Will it be the traditional three-tier architecture comprising access, distribution and core switching layers? Or do emerging technologies enable implementation of a two-tier - or even a single-tier - architecture?
Technical Issues with LAN-SAN Convergence
Beyond answering these general questions about data center LAN design, IT organizations need to address some of the technical challenges associated with converging the LAN and the SAN. For example, traditional Ethernet's best-effort service allows buffers to overflow during periods of congestion and relies on upper-level protocols such as TCP to manage congestion and recover lost packets through re-transmissions.
In an integrated LAN-SAN, this could result in a level of delay that is unacceptable for storage. There are, however, emerging technologies that enable Ethernet to support a converged fabric, which will be discussed in a future TechNote.
As discussed in the previously referenced research report, there are strong factors driving IT organizations to redesign their data center LANs. However, the fact that some of the enabling technologies are fairly new to the market - and some are even still in development - means that data center LAN redesign will take place slowly over a number of years. As a result, the adoption of a converged LAN-SAN fabric will also be a gradual, years-long process.
While discussing the likelihood of LAN-SAN integration, we cannot ignore the strategic change in the LAN itself. For more than 20 years, the LAN has been the network that linked workstations to servers. That is about to change. The "New LAN" is primarily an Internet Access network, serving stationary and mobile workstations, with smart phones and tablets as the new client devices. More and more often, these devices are selected, paid for and managed by their users rather than the IT department. They typically have Wi-Fi and 3G connections, so Wi-Fi is no longer a special facility for meeting rooms but should cover the premises wall to wall. Since the new clients are designed to access their servers and services in the Internet (the cloud if you wish). Internal servers, that should also be accessible when the client is off-premises, are accessed through a shortcut in the company firewall, linking the LAN and the server network.
LAN-SAN integration will only affect the server network. The traditional LAN was a mix of servers (high link load per node) and workstations (low load per node), which complicated integration of SAN-traffic (very high load per node). Now that the workstations are separated in their access-network, integration of (server)LAN and SAN traffic actually becomes easier to design.
HP Virtual Connect and similar port-reduction techniques help to reduce cabling, but at the same time show how tricky such integration can be from a traffic engineering standpoint. That problem would probably be solved when the common channel can be a 100 GbE link, an order of magnitude faster than each of the tributaries. Yet, server networks and SANs each have their own development paths. SANs may well look very different in 5 years’ time. Tying both network applications to a common implementation may turn out to be less future-proof than expected.