Virtualization is an inherent part of cloud computing, and is forcing close examination of network and cabling architectures.
By Patrick McLaughlin
Before the ink is even dry on the TIA-942-A Telecommunications Infrastructure Standard for Data Centers, an addendum … Actually, before the ink is even wet on TIA-942-A, the first addendum to the standard is in the works. Officially approved by the Telecommunications Industry Association (TIA; www.tiaonline.org) in April, 942-A is in the process of formal approval by the American National Standards Institute (ANSI; www.ansi.org) and may be available for purchase in July or August. So in that regard, there’s not even any ink on the standard to be wet or dry at this point. Still, the TIA’s TR-42.1 Committee continues to work on data center infrastructure specifications, including current activity to develop Addendum 1 to 942-A. That addendum will address cabling guidelines for what are known as data center fabrics, and has emerged as a necessity in no small part because of the virtualized environments that pervade cloud-computing data centers.
Standards beat goes on
Jonathan Jew, principal of J&M Consultants (www.j-and-m.com) and a significant contributor to 942-A, its predecessor TIA-942, as well as a number of other TIA and ISO cabling standards, commented on the fact that 942-A was not the proverbial finish line for data center cabling specs. When asked how standards organizations are responding to an ever-changing data center landscape, Jew said they “are updating their standards to incorporate many new industry developments, like energy efficiency and modular data centers. Additionally, we are trying to get ahead of the curve by increasing minimum grades of cable and recommending grades of cable capable of supporting new networks. We are also not restricting ourselves to the five-year cycle of ANSI standard revision and are developing addenda to handle new needs, such as data center fabrics.”
While Addendum 1 to TIA-942-A is in early stages of development and set to undergo initial balloting, the content of the draft ballot makes it clear that the addendum seeks to apply the cabling architectures set forth in 942-A to these emerging data center fabrics. But what is a data center fabric in the first place, and how does a fabric relate to cloud computing and virtualization?
As we have reported previously (“Signs pointing upward for cloud computing,” February 2012), Cisco (www.cisco.com) has explained in its inaugural Global Cloud Index the interrelationship between cloud computing and virtualization. When explaining why the amount of data traffic remaining within the data center will decrease only slightly—from 77 percent to 76 percent between 2010 and 2015—the index said, “The ratio of traffic exiting the data center might be expected to increase over time, because video files are bandwidth-heavy and do not require database or processing traffic commensurate with their file size. However, the ongoing virtualization of data centers offsets this trend. Virtualization of storage, for example, increases traffic within the data center because virtualized storage is no longer local to the rack or server.”
The index also defined and put into perspective the term “workload.” It said, “Traditionally, one server carried one workload. However, with increasing server computing capacity and virtualization, multiple workloads per physical server are common in cloud architectures. Cloud economics, including server cost, resiliency, scalability, and product lifespan, are promoting migration of workloads across servers, both inside the data center and across data centers … Often an end user application can be supported by several workloads distributed across servers. This can generate multiple streams of traffic within and between data centers, in addition to traffic to and from the data center.”
The virtualization of servers, then, has increased the complexity of connectivity among those servers, which is facilitated by the switches within data center networks. A traditional network architecture within a data center includes three tiers of switches—access switches, aggregation switches, and core switches. As such, signal transmission from any one access switch to any other access switch may have to go through three other switches (from the point-of-origin access switch, to an aggregation switch, to a core switch, to another aggregation switch, to the point-of-destination access switch).
An early draft of 942-A Addendum 1 states, “The traditional architecture is well-suited for traffic between servers on the same access switch and from servers to external destinations. However, it isn’t suitable for large virtualized data centers where compute and storage servers may be located anywhere in the data center.” The addendum then provides examples of alternative switch architectures, or fabrics, which have been put forth within the networking industry. More significantly, the addendum addresses the means by which these fabrics can be cabled. We will provide more detailed information on this standard addendum in our next issue, including descriptions of the architectures with names including “fat tree,” “full mesh,” and “centralized switch.”
Data center fabric
As these fabrics have been introduced to data center networking environments, they each have had their proponents and skeptics. In 2010 Lippis Consulting’s (lippisreport.com) president Nicholas John Lippis III authored a paper entitled “A simpler data center fabric emerges for the age of massively scalable data centers.”
Within that paper Lippis describes the so-called “unified fabric”: “The concept of a unified fabric is to virtualize data center resources and connect them through a high bandwidth network that is very scalable, high performance and enables the convergence of multiple protocols onto a single physical network. These resources are compute, storage and applications, which are connected via a network fabric. In short, the network is the unified fabric and the network is Ethernet.
“The industry tends to focus on storage transport over Ethernet as the main concept behind a unified fabric with technologies such as Fibre Channel over Ethernet or FCoE, iSCSI over Ethernet, iWARP over Ethernet and even Infiniband over Ethernet. However, this is a narrow view of a unified fabric, which is being expanded thanks to continual innovation of Ethernet by the vendor community and standards organizations such as the IEEE an IETF. Ethernet innovations such as FCoE, Data Center Bridging or DCB, Cisco’s VN-Link, FEX-Link and virtual PortChannel or vPC have enhanced Ethernet networking to support a wide range of new data center fabric design options.”
The resiliency of cloud services, supported by virtualized computing environments within data centers, emerged as an area of interest when Cabling Installation & Maintenance surveyed its audience about their use of cloud services earlier this year. Conducted in February, the survey aimed to obtain perspectives on cloud computing among professionals in end-user organizations responsible for their networks’ physical layers, and also responsible for evaluating or purchasing cloud services.
One question in the survey asked respondents to rate their motivations for using cloud services on a scale of 1 to 5, in which 1 was “not important” and 5 was “most important.” A score of 3 is “important” and 4 “very important.” Among all respondents—from all geographic regions and with enterprise networks of all sizes—the two most important factors influencing their decision to use cloud services were more network uptime (scoring 3.55) and faster network speeds (scoring 3.29). Among users whose networks include 5,000 or more nodes, those capabilities are even more important. For that group (5,000+ network nodes), more network uptime scored a 3.64 on the importance scale and faster network speeds scored a 3.45.
By those markers, user organizations turning to the cloud have high expectations for robust service from their cloud providers. In one regard, cloud services will be like dial tone; users will expect instant, uninterrupted connectivity at their disposal. For cloud service providers, who will provide these services via virtualized data centers—many of them sprawling—the internal infrastructure is of paramount importance. One facet of that infrastructure is the fabric or architecture. Data center fabrics incorporate highly connected, high-speed series of switches. And of course these fabrics must be supported by structured cabling systems capable of such high speeds.
Patrick McLaughlin is our chief editor.