Tackling Network Congestion

Arthur Cole

When it comes to moving your data from place to place, the chief enemy is network congestion. And it doesn't take a genius to figure out that the more processing and storage that gets loaded into the data center, the more traffic you'll have to snarl things up.


But with the arrival of 10-gig interconnects comes a raft of new approaches that some vendors are bold enough to call "congestion-free" technology.


Teak Technologies and Neterion are touting an Ethernet solution that they say brings a 10-fold improvement in application latency by offering full-link utilization that maintains multiple data flows even under sustained congestion conditions. At the NGDC Conference in San Francisco this week, the companies linked an IBM BladeCenter-H chassis, sporting a mix of AMD and Intel blades, with a pair of Tek 10 Gb CFE switches and six 10 Gb CFE expansion cards outfitted with Neterion's Xframe ASICs.


Using a ping-pong test across multiple NetPIPE client/servers to simulate congestion, the setup saw a 100 percent improvement in load-invariant loss-free link utilization and a four-fold decrease in the number of uplinks in multi-hop network configurations.


For HP users, lower latency is available in the form of a new 10 GbE "Green Blade" switch from BLADE Network Technologies. Designed for the c-Class blade, the device provides 400 Gb of Ethernet connectivity, suitable for converged single-fabric networks and the kinds of workloads generated by virtual networks. The company claims a nine-fold latency reduction and double the throughput of chassis-based 10 GbE modules.


Still another combination of technologies from Woven Systems and Chelsio Communications has proven capable of dynamically adjusting to network congestion at the 128-node Sandia National Laboratories in Santa Clara, Calif., recently, according to this Byte & Switch piece. The package consisted of Chelsio's R310E iWARP HBA and Woven's EFX 1000 Ethernet fabric switch. It rated a peak unidirectional bandwidth of 1204 MBps, with virtually no loss in the presence of additional processors. This compared to 960 MBps for single data-rate Infiniband and a 61 percent bandwidth drop as processors were added.


Latency within the data center is one thing. Trying to send out applications and services over the Web is a whole new ball game. But eWEEK says a company called Anagran, founded by Net pioneer Larry Roberts, seeks to boost the Web's performance with the FR-100 Flow Router. Although targeted at time-sensitive applications like voice and video, the data realm should also benefit from the device's Intelligent Flow Discard system that evaluates packets on the input port, rather than the output port, easing congestion caused by large peer-to-peer file transfers. The device is available in a 1 RU form factor in combinations of up to 48 GE ports or four 10 GbE ports.


In a way, network congestion is one of those problems that's good to have. More traffic generally means more business. But it's the kind of thing that requires constant attention because it can bring business to a halt pretty quickly.

Add Comment      Leave a comment on this blog post

Post a comment





(Maximum characters: 1200). You have 1200 characters left.




Subscribe to our Newsletters

Sign up now and get the best business technology insights direct to your inbox.