I was recently at an event that Huawei hosted in Latin America for its telecom carrier community, in which Huawei was showing off an impressive range of carrier-related technology, including distributed data center management, advanced analytics and a heavy emphasis on compute and storage in addition to their traditionally strong core carrier technology. Interestingly they chose this venue for the Latin America unveling of the KunLun server, an impressive bit of engineering which clearly shows that innovation in big-iron x86 servers is not dead. There is some confusion about whether the March announcement at CeBIT constituted the official unveiling of the actual machine, but they had a real system on the floor at this event and claimed it was the first public showing of the actual system.
The Kunlun server, named after a mountian range in Quinghai Province, places Huawei squarely up against the highest end servers from HPE, IBM, Oracle, NEC and Fujitsu, with a list of very advanced RAS features, including memory migration, hot memory and CPU swap, predictive failure diagnostics and a host of others, some enabled by the underlying Xeon E7 technology and others added by Huawei through their custom node controller architecture ( essentially a standard feature of all large x86 servers). Partitionable into smaller logical servers, the Kunlun can serve as a core transaction processor for extreme workloads or as a collection of tightly coupled electrically and logically isolated servers.
So why unveil this high-end engine at a telecom carrier show? My read is that since the carriers will be at the center of much of the IoT action, and that the data streams they process will need an ever expanding inventory of processing capacity, so this is a pretty good venue, Plus it reinforces the emerging primacy of analytics, especially in-memory analytics, which it can address extremely well with its current 24TB (32G DIMMs) of DRAM.Read more