FOCUS CONNECTIVITY UPDATE
Issue 14, February/March
FOCUS UPDATE: CONNECTIVITY
applications such as cloud computing, large scale video-on-demand or content storage, this poses a unique challenge in the data center, which must deal with significant increases in storage capacity, power and the ability to move data over the network.
A
These challenges are not just of physical link speed but also of network congestion and the ability of network equipment in the data center to handle the switching of such large amounts of data.
To handle that capacity new technologies at both the physical and data link layers are being created by standards groups such as the IEEE. The 40/100Gb Ethernet standard, or IEEE 802.3ba, is the newest and fastest Ethernet standard and the first standard to include two speeds.
Speeds of 40/100Gb allow Ethernet technology to meet ever-increasing performance and bandwidth demands driven by consumers, service
providers, data centers and high-
performance computing environments. Hence, 40/100Gb Ethernet can replace current solutions for increased bandwidth, such as link aggregation, which can be highly inefficient and difficult to use and maintain.
The 40/100Gb standard is compatible with previous Ethernet standards, allowing the new Ethernet to be installed anywhere in an end-to- end network without necessarily requiring an upgrade to other components of the network.
THE STANDARD ARGUMENT
This new standard is of particular interest to data centers as deployments of 40/100Gb Ethernet devices will allow not only increased bandwidth between systems but decreased density of 10Gb Ethernet connection ports at the top-of-rack and cross connect points. In addition, advancements in congestion avoidance via new IEEE 802.1 Data Center
58
www.datacenterdynamics.com
ccording to analyst firm IDC, the amount of world information is set to rise 44-fold in the next
decade. Whether due to
CHALLENGES WITH 40 AND 100GB ETHERNET University of New Hampshire InterOperability Laboratory senior engineer for enterprise networks Jeff Lapak and data center research and development member Mikkel Hagen explore the challenges that will accompany the rise of 100GbE
InterOperability Laboratory senior engineer Jeff Lapak
Bridging (DCB) standards allow modern Ethernet-based data centers to be truly free of congestion-induced packet loss.
These
advancements further enable the convergence of all data center traffic to one network.
Traditionally, data center traffic consists of inter-processor communication (IPC), storage area networks (SAN) and data connectivity. Ethernet-based IPC today includes two forms of remote direct memory access (RDMA) transport: IETF’s iWARP (RDMA over TCP/ IP) and IBTA’s RDMA over converged Ethernet (RoCE). Ethernet-based SANs include Internet small
computer system interface (iSCSI)
and Fiber Channel over Ethernet (FCoE). As data center needs exceed 10Gbps speeds, the 40/100Gb Ethernet standard allows converged Ethernet to keep pace as a viable option.
The biggest challenge will be the timing of priority flow control (PFC) frames
The University of New Hampshire InterOperability Laboratory (UNH-IOL) has been developing testing suites and supporting test events and demonstrations for both 40/100GbE and data center technologies. The UNH-IOL has worked with industry leaders to develop several testing “consortia” to assist interoperability in the data center market.
A UNH-IOL testing consortium is a unique program that brings together vendors looking to improve their products, and allows them to perform cost-effective interoperability and conformance testing that significantly
A lot of early devices implement what is referred to as the baseline implementation. As the standard has progressed to version 2.0, different vendors have implemented the standard at different points along the way. It is important that data center designers make sure that each device supports the same DCBX version or their network will likely not work at all.
The current DCB standards have been written with future compatibility in mind, but 40G and 100G will face some challenges in implementing DCB. The biggest challenge will be the timing of priority flow control (PFC) frames.
It is critical that every device is able to respond to PFC in a certain time window to ensure that frames are not dropped. As network speeds increase from 10Gb to 40Gb to 100Gb, timing becomes harder to achieve. The standards group has attempted to make a reasonable trade-off on timing versus memory requirements for 40Gb and 100Gb, but it has not been widely tested and validated.
Data center research and development Mikkel Hagan
decreases their time to market. In regard to the data center, consortia exist for 40/100Gb Ethernet, Data Center Bridging (DCB), and Fiber Channel over Ethernet.
The congestion challenges faced in a 40/100Gb data center make interoperability testing more important than ever. The biggest interoperability challenge associated with deployment of DCB networks is the Data Center Bridging Capability Exchange Protocol (DCBX). While DCBX is being defined to make life easier for network administrators, it is causing major issues with interoperability.
Page 1 |
Page 2 |
Page 3 |
Page 4 |
Page 5 |
Page 6 |
Page 7 |
Page 8 |
Page 9 |
Page 10 |
Page 11 |
Page 12 |
Page 13 |
Page 14 |
Page 15 |
Page 16 |
Page 17 |
Page 18 |
Page 19 |
Page 20 |
Page 21 |
Page 22 |
Page 23 |
Page 24 |
Page 25 |
Page 26 |
Page 27 |
Page 28 |
Page 29 |
Page 30 |
Page 31 |
Page 32 |
Page 33 |
Page 34 |
Page 35 |
Page 36 |
Page 37 |
Page 38 |
Page 39 |
Page 40 |
Page 41 |
Page 42 |
Page 43 |
Page 44 |
Page 45 |
Page 46 |
Page 47 |
Page 48 |
Page 49 |
Page 50 |
Page 51 |
Page 52 |
Page 53 |
Page 54 |
Page 55 |
Page 56 |
Page 57 |
Page 58 |
Page 59 |
Page 60 |
Page 61 |
Page 62 |
Page 63 |
Page 64 |
Page 65 |
Page 66 |
Page 67 |
Page 68 |
Page 69 |
Page 70 |
Page 71 |
Page 72