Revolutionising storage networks

Andrew Gilman, SNIA Ethernet Storage Forum Board (Dell) & Eric Schott, SNIA Ethernet Storage Forum Member (Dell) discusses how storage network technology is evolving.

As economic conditions recover in 2010 many organisations will be looking to evolve their IT operations with new data centre technologies while enhancing business processes. Server and storage consolidation through virtualisation technology has been an essential element to this evolution by helping administrators contain operational expenditures (OPEX) and capital expenditures (CAPEX). However, as IT consolidation increases overall utilisation and I/O density of our virtualised data centres, the network is emerging as the new bottleneck and an area primed for innovation. In short, we simply need a larger "pipe" for communications to extend the benefits of IT consolidation through virtualisation to the network.

Fortunately, new networking protocols have been swiftly making their way through the standards committees including 10 Gb Ethernet and Fibre Channel over Ethernet (FCoE). These new standards go by several names: "Converged Enhanced Ethernet," "Data Centre Ethernet," and the industry-standard term "Data Centre Bridging," or DCB. Vendors are increasingly integrating these networking capabilities in new server, switch and Storage Area Network (SAN) designs that will lower overall deployment costs in 2010 and beyond. This article will provide an overview of the reasons DCB is being developed and how it can improve networking in the data centre for applications, servers, and storage.

Why ethernet requires TCP for reliable communications

While Ethernet is generally a reliable technology, data packets can be dropped in transmission due to network congestion, traffic load balancing, and loads on servers and switches. Dropped packets typically don't cause big problems, but they can result in performance variations for applications.

Protocols such as TCP are run over IP on top of Ethernet to ensure seamless communications. TCP understands if packets were dropped during the communication; if so, those packets are re-sent, and if not, confirmation of a completed transmission is returned. With this "handshake," TCP delivers reliable communications. TCP also provides routing capabilities so that communications work seamlessly across different networks.

DCB delivers reliability and predictable performance

DCB extends Ethernet by providing a network infrastructure that minimises packet loss, enabling improved data networking and management within the DCB network environment with features for priority flow control (P802.1Qbb), enhanced transmission selection (P802.1Qaz), congestion notification (P802.1Qau), and discovery. The result is more deterministic network behaviour. DCB is enabled through enhanced switches, server network adapters, and storage network adapters.

One component of the DCB framework is class-based pause (an extension to the IEEEP802.3x™ definition) that makes network performance extremely predictable by delivering "lossless" network traffic. While standard Ethernet performs very well, its performance can vary (see graphic). With DCB, the maximum performance is the same, but performance varies very little. This is extremely beneficial for data centre managers, enabling them to better predict performance levels and deliver smooth traffic flows.

Segregating and prioritising traffic

Storage networking best practice recommendations include physically separating the data network traffic from the storage network traffic. Because today's devices commonly have multiple 1Gb ports, segregating traffic has been easy-for example, two ports can be assigned for storage networks and two for data networks. However, consolidated server environments have a different circumstance. These environments generally use server virtualisation software with large servers, and can simplify connectivity with multiple 10 Gb ports-consolidating bandwidth instead of distributing it among multiple 1 Gb ports and a tangle of wires. These 10 Gb ports handle all the traffic-database, web, management, and storage-improving infrastructure utilisation rates. But with traffic consolidated on fewer larger connections, how does IT segregate the storage and data networks, prioritise traffic, and guarantee service levels? Data Centre Bridging (DCB) includes a prioritisation functionality, which improves management of traffic flowing over fewer, larger pipes. In addition to setting priority queues, DCB can allocate portions of bandwidth. For example, storage traffic can be configured as a higher priority than data traffic-but the administrator can allocate 60 percent of bandwidth to the storage traffic and 40 percent to the data, ensuring operations and predictable performance for all.

Storage networks and "lossless" communications

The two primary storage networking protocols are Internet SCSI (iSCSI) and Fibre Channel (FC). In an iSCSI storage network, SCSI commands are sent via TCP/IP to ensure "lossless" communications. In a Fibre Channel storage network, SCSI commands are sent using the Fibre Channel protocol; however, Fibre Channel sends frames with a "handshake" that does not expect packets to be lost. The FC protocol presumes a "lossless" network infrastructure and therefore requires specialised HBAs and switches to provide reliable physical communications.

Unifying the network: FCoE Requires DCB

For unified networking, iSCSI works well with 10 Gb Ethernet. To bring the FC protocol onto this unified 10 Gb infrastructure, the industry is creating Fibre Channel over Ethernet (FCoE). Because the FC protocol requires a "lossless" network, Data Centre Bridging is required for FCoE deployments. FCoE will work only with end-to-end DCB-enabled components; no intermediate, non-DCB enabled switches or devices can function with FCoE.

Deploying Ethernet and DCB. Devices that support both standard Ethernet and DCB will interoperate easily based on the switch settings on the network ports. For example, an iSCSI storage device that supports DCB, and is connected to a DCB switch, can communicate with all devices on the network without re-configuration or intervention. Components communicate ubiquitously, since iSCSI networks can go into and out of DCB portions of the network without packet conversion. In iSCSI environments administrators can start with a small DCB implementation and expand as desired. Conversely, FCoE environments require DCB, necessitating an end-to-end data centre network conversion.

How DCB brings consolidation benefits to networks

Data centre bridging is a new network framework based on extensions of Ethernet standards, that can bring to networks the same IT consolidation benefits that storage and servers have enjoyed in recent years-higher utilisation rates, simpler management, and lower total cost of ownership. It is reliable, offers predictable performance, and can segregate and prioritise traffic. Administrators can now implement standard Ethernet, Data Centre Bridging, or a combination of both.

The SNIA Ethernet Storage Forum is a great resource available to end-users who have questions about how to integrate these established and emerging technologies. For more information log onto www.snia.org/esf.