Emulex ExpressLane and CrossLink boost HBAs for flash and Hadoop

Storage card maker Emulex has added QoS prioritisation for flash storage and leveraged RDMA via Fibre Channel to increase server performance in clustered operations such as Hadoop

Network equipment maker Emulex has announced two enhancements to its Fibre Channel host bus adapter (HBA) and Fibre Channel/Ethernet converged network adapter (CNA) products that boost performance for flash storage and between servers in clustered operations such as Hadoop.

The new ExpressLane feature tackles the potential for congestion in storage environments with mixed flash and spinning disk media. ExpressLane applies a tag to traffic to or from a LUN associated with flash storage to give it a higher priority.

ExpressLane boosts performance for high-priority traffic by alleviating congestion, reducing latency and improving throughput on the storage network, according to Emulex. It is a feature that can be switched on or off using Emulex’s OneCommand Manager.

“Things that operate using flash storage, such as the transaction log in SAP Hana, are often the bottleneck to good performance and govern what the server can do at any given time and whether they meet SLAs or not,” said Shaun Walsh, senior vice-president of marketing and corporate development for Emulex.

The other new feature, CrossLink, brings remote direct memory access (RDMA) to bear on server-to-server traffic via in-band Fibre Channel message passing. RDMA effectively replicates data from the memory of one server to another in microsecond timescales. This equates to a huge decrease in latency over existing IP-based transports for such traffic, which are measured in milliseconds.

“The speed of, for example, a Hadoop cluster is driven by whether each node is at full capacity, so it’s important to ensure that is the case," said Walsh. "CrossLink, by using RDMA instead of standard, high-latency IP transport, increases speed and the frequency of CPU cycles.”

Read more on Storage fabric, switches and networks

Data Center
Data Management