This also ties back to my analysis on how the Emulex UCNA is basically like trying to build a house on leased land: the leased land comes from ServerEngines in the form of a baseline Ethernet NIC while the recreated FC stack comes from Emulex and the two get bolted-together in a product that is essentially owned by two companies. This raises concerns from a support perspective. If you have Ethernet problems, Emulex escalates them to ServerEngines and waits for ServerEngines to try and address them, then Emulex caries back fixes to the customer. Who's responsible for what, and how is this model scalable?
Also, this extremely high level of CPU utilization for the Emulex UCNA will probably result in a greater amount of memory required, and the impact to Virtual Machine density per physical server needs to be carefully considered. Users will have to carefully choose which adapter vendor to use so as to not negatively impact their virtualization plans due to overwhelming CPU utilization. The business impact of higher CPU utilization rates caused by I/O infrastructure products means a potential impact to application performance, and IT negatively effecting a revenue generating application will not be well tolerated by the business side of the house.
In related UCNA news, Emulex recently announced "OneCommand Vision," an IT management tool that Analytico, Inc. believes competes with management tools from its OEMs, including EMC Ionix, HP Insight Control and various software tools from IBM. I don't expect any of these OEMs to carry this software product from Emulex as it appears to step on their toes from an overall management perspective. Clearly, Emulex looks like it is trying to expand its market opportunity, and in this tough economy, all vendors are trying to do the same; however, in the process the company also appears to be creeping into the front yard of its top customers. IT management software is in high demand this year, and many storage systems vendors are rolling out their own unique solutions.
As always, we encourage users to review vendor benchmark activity and to, if at all possible, conduct testing of their own prior to deploying I/O infrastructure hardware. Also, remember to look at, or ask for, "all" the performance numbers so that a full assessment of overall performance can be well understood before deploying these new technologies.