Director of Power Systems
IBM announced, on Oct 5, 2015, the availability of three new POWER8 based servers that are supported by the Ubuntu Linux operating system, with PowerKVM as the optional virtualization support.
The new “LC” models are the S812LC, the S822LC with integrated NVidia K40 GPU capabilities, and the S822LC for commercial workloads (non-GPU). IBM states that the “LC” designation stands for “Linux Cloud,” although pundits also point out the “LC” could stand for “Low Cost” since these new systems have a much lower entry price point than the existing S812L and S822L models… in some cases, almost 60% less for similar configurations.
The new “LC” models are a direct result of the Open Power Consortium, which was founded in 2013 by IBM, NVidia, Tyan, Mellanox, and Google. The “LC” systems are developed and manufactured in China by Tyan and Wistron, and all incorporate the POWER8 Single Chip Module designed and manufactured by IBM. The “LC” models incorporate commodity components such as memory, adapters, and internal disks, which leads to the new price points. These systems carry the IBM logo, are CAPI enabled, and are fully supported with a 3 year 9×5 warranty from IBM. Although the 3 models can support any Linux workload, they are especially suited for targeted workloads based on the configurations. All configurations can support up to 1 TB of RAM and utilize either the 2.92 GHz (10 core chip) or the 3.32 GHz (8 core chip) processor. The models differ in configuration capabilities after this commonality.
The S812LC supports a single 8 or 10 core POWER8 chip, with up to fourteen 8 TB disks (for a total of 112 TB of internal storage), and 4 PCIe Gen 3 slots. This configuration makes the S812LC especially suited for grid or parallel workload, like Hadoop and Spark, which utilize large internal disk storage.
The S822LC supports two 8 or 10 core POWER8 chips, with two 1 TB disks and 5 PCIe slots. One model (the GTA) supports two NVidia K40 GPUs for high performance computing, and the “commercial computing” model (the GCA) does not have the GPUs included. With the integrated NVidia GPUs, the target workloads are high performance computational workloads like seismic, cryptographic, genome research, etc. The GTA model provides a high performance platform for relational database and no-SQL data bases and transaction processing.
So, what then differentiates the existing “L” models from the “LC” models? The “L” models provide support for PowerVM, Red Hat, and S– USE, where the “LC” models do not. The “L” models have higher GHz cores, and also support more PCIe slots for connectivity. Another difference is that the “L” models can be supported by the HMC (virtual or real) and contain the Flexible Service Processor for added RAS.
Both the “L” and the “LC” models are great solutions for applications needing high processing speeds, high memory bandwidth, and high IO bandwidth. I believe, the choice of which system to use will be based on application capacity needs and company virtualization standards. The “LC” models now introduce a lower entry price into high performance Linux computing needs.