subscribe: Daily Newsletter


Huawei leads with mission-critical server


Kathy Gibson reports from CeBit, Hannover – Bringing the compute power needed to drive intelligent systems for Industry 4.0, Huawei today announced the first 32-socket x86-based mission-critical server.
“The new KunLun  server ushers in a new era of mission-critical computing, says Cody Wang, GM: mission critical servers at Huawei Enterprise. “ICT is the driver behind Industry 4.0 and now is the time that we need to start operating intelligent systems.
“Huawei is a big player in the Industry 4.0 space, where everything is connected and everything is intelligent, raking fourth in overall global server shipments.”
New market dynamics are driving a new era of mission-critical computing, Wang says. “In the past, mission-critical computing was run on high-reliability hardware based on RISC, Power or Itanium processors. Now there is a mature and high-performance architecture with high availability based on x86 processors.
“In addition, users no longer need to run these systems on Unix, but can use Linux and also have access to many more applications.”
Wang explains that the name KunLun is derived from a Chinese mountain that gives rise to two rivers. “We hope this system will be as strong as the mountain, and offer customers value like the mountain does to the rivers.”
There are three specifications of KunLun: KunLun 9008, 9016, and 9032. KunLun 9032 server supports up to 32 CPUs, 576 cores, 1152 threads, and 768 DDR4 DIMMs.
The new servers are very reliable, Wang says, offering 99,999% uptime and hot swop capabilities.
“In terms of performance, the KunLun server set a new SPECint world record, and breakthrough tpmC speeds. It has 40% higher performance and 50% lower TCO than an equivalent RISC system.”
Huawei envisages four main application scenarios for KunLun. They are:
* Database/OLTP – where it offers up 50% improved TCO and improved customer experience
* High-performance computing (HPC) fat node – with 18x threads per node and 25-times memory bandwidth
* In-memory computing – a 24Tb Hana singe node has been successfully tested and evaluated
* Cloud and virtualisation – Huawei believes it is the best platform for virtualisation and consolidation.
The new servers are powered by Intel Xeon E7 v3 series central processing units (CPUs) and supports mainstream databases such as SAP Hana; middleware; and operating systems (OS) including Red Hat Linux, Suse Linux and Windows Server.
Peter Gleissner, vice-president: European region at Intel, points out that 2015 saw the 50th anniversary of Moore’s Law last year, which has been one of the primary enablers of the IT industry.
“The E7 platform is designed from the legacy of Itanium, and boasts the highest bandwith and reliability of any processor to date,” he says.
“It was developed with the Internet of Things (IoT) where data needs to be analysed in realtime.”
SAP is one of Huawei’s strategic partners, and KunLun is designed to run the Hana in-memory system.
“Everything we build in SAP is based on Hana,” says Johannes Woehler, vice-president: Hana at SAP. “That is why we put it in the core of our collaboration with Huawei.
“Hana changes the game for our customers who want realtime success, allowing them to run transaction and database data on the same core. We saw that the collaboration with Huawei will help us to bring this to more customers.
“With Hana’s high compression and scalability across multiple cores, customers can leverage large core servers.”
Woehler adds that the SAP-Huawei collaboration began with a global technology partnership in 2012 and by 2015 a number of customers had begun implementing Hana on Huawei systems. “Since 2015 we have a joint collaboration technology centre in Schenzhen where we focus on test and evaluation with SAP Hana.”
This collaboration extends from solutions for the boardroom of the future to design thinking for IoT and big data.”
Linux is central to KunLun’s ability to outperform the previous generation of mission-critical servers, and Suse is a strategic partner in the development and implementation of the platform.
Ralf Flaxa, president of engineering at Suse, explains that Suse is a pioneer in the Linux and open source space. “We specialise in enterprise Linux and have been working to harden it make it mission-critical.”
Huawei’s RAS 2.0 technology enables proactive fault management. Based on firmware and independent of OSs, the fault response and diagnosis mechanism of RAS 2.0 allows fault information collection and analysis, which automatically isolates risky components, executes failover, or informs maintenance staff of risky components for early replacement before critical errors occur. It is also first in the industry to support online CPU and memory module maintenance without shutting down the server, ensuring continuity of mission critical services and providing a strong foundation for stable business growth.
With Huawei’s Node Controller interconnection chips, KunLun leads in enabling high speed interconnection for 32, and up to 64, Xeon E7 v3 CPUs, exceeding the industry standard of interconnection for eight Xeon E7 v3 CPUs.
The server is also first in the industry to provide both physical and logical partitioning of firmware in open-architecture mission critical servers. A partition can contain one physical core or up to 32 CPUs to meet different computing requirements, maximising resource utilisation.