Huawei unveils larger iron KunLun server at CeBIT • The Sign up

Huawei unveils larger iron KunLun server at CeBIT • The Sign up

Huawei has unveiled a more impressive variation of its best-close KunLun server at CeBIT, amongst a raft of other major iron-ish hardware and computer software bulletins.

KunLun is Huawei’s major freaking box of a server, with up to 32 CPUs in a rack. This V5 edition is generally a Xeon SP refresh with NVMe travel support. This is the newest transfer in Huawei’s Skylake refresh of its server line.

The new hardware is intended for the best close niche of the X86 server sector, the so-called mission-essential little bit, and this scalable box is created from four SCE sub-enclosures, each and every with from two to 8 processors and 12TB of DDR4 memory (96 DIMMs), and one CME central administration chassis. There are 3 types, the 9008 with one 8-CPU chassis, the 9016 with two and the 9032 with 4. The 9008 and 9016 are upgradable to 32 CPUs.


KunLun 9008 (still left), 9016 (center) and 9032 (suitable)

These are both Xeon Platinum 8100 or Gold 6100 processors, with up to 768 cores (with 24-main CPUs) and 32TB memory for each node.

The program can be physically partitioned into four-CPU groups or there can be up to forty x 1-96 CPU main logical partitions for each host. There are PCIe growth slots, two x 10GbitE SFP+ ports and two x GE RJ45 ports.

The local storage for each SCE box is up to forty eight x two.5-inch SAS/SATA disk drives or forty x two.5-inch NVMe SSDs. Capacities are not offered in the datasheet. RAID stages of ,1,ten, 5, 50, 6 and 60 are supported. Supported operating methods are RedHat and SuSE Linux, Home windows Server and VMware ESX.


Huawei dunks server triplets in Skylake for a v5 refresh

Read Far more

Huawei suggests the KunLun v5 has a multi-layer fault-tolerant architecture with fault-tolerant chips, firmware, and OSs. It is totally redundant with no one place of failure. We’re instructed it provides forty for each cent increased effectiveness than RISC servers, but with no facts.

Why have a 32-CPU rack server as an alternative of 32 individual rack shelf servers? The limited respond to is that you can team the CPUs and cores into compute models sized to workloads.

If you have a set of variably sized workloads, which you want to operate at differing occasions and in differing mixtures, then a one 8, 16 or 32-CPU server which you can flexibly and dynamically modify into appropriately-sized compute models, through bodily and logical partitioning, can make superior feeling.

This does not make it a composable server in the HPE Synergy/Attala/DriveScale/Liqid feeling where you can dynamically define compute, memory and storage groupings from resource pools. How does KunLun examine to other suppliers’ 32-socket course servers?

Some other major iron server suppliers

Cisco does not have a server in this course.

There is a Dell m100e 10U blade server chassis which can maintain 16 blades, each and every of which can be a four-socket program using 22-main Xeon E5-4600 v4 CPUs, that means a whole of sixty four CPUs and 1,408 cores.

But this is a pre-Skylake program and there is no centrally-managed, four-chassis program equivalent to the KunLun. The exact same appears to be correct for Fujitsu with its blade server chassis items

Hitachi had a 12U CB 2500 chassis obtainable in the 2016 era with up to fourteen twin-processor server blade modules. There is an 8-socket CB520X blade module with LPAR support. This blade chassis program has not had a Skylake makeover.

HPE has its Integrity rack servers using Itanium processors, and its Superdome Flex which supports 32 Xeon SP sockets and 48TB of memory when 8 x 5U chassis are grouped together.

IBM has, of program, its mainframe methods, and its Power9-centered AC922 (Power9 – 22-main) with two to 6 Tesla V100 CPUs can be scaled out to a entire rack or racks.

Huawei also announced a raft of other new items at its CeBit extravaganza, this kind of as Online of Motor vehicles kit, FusionCloud 6.3 entire-stack offerings with more than forty cloud companies, and the FusionInsight LibrA converged facts warehouse.

Huawei suggests this takes advantage of parallel computing, mixed row-column storage, and vectorized execution technologies to complete affiliation evaluation of trillions of facts records in seconds. Huge iron hardware and computer software rules below, all right. ®

Huawei Server

Leave a Reply

Your email address will not be published.