Two versions of the new AI and machine learning-heavy servers for edge environments were announced, a balanced commercial version and a more rugged military spec version, but different versions for more specific use cases are likely going forward.
Today, Western Digital is announcing the availability of their Ultrastar Edge server family, a new high-performance server option designed to support analytics processing at the edge in demanding environments, including rugged ones like the military and semi-rugged like oil and gas.
Western Digital is not commonly associated with servers, but they have made ones before, for specific use cases under the Ultrastar brand.
“Our Ultrastar branding started with HDD and SSD products, and we leveraged our devices and build specialized platforms around them,” said Scott Hamilton, Senior Director, Product Management and Marketing at Western Digital. “This includes high performance JBODs, NVMe over-Fabric storage, two different products within OpenFlex umbrella – and a few specialized storage servers.”
Hamilton emphasized that these Ultrastar servers have been very specialized designs to fit specific use cases.
“We aren’t in the white box industry,” he said. “We have a variety of OEM customers who leverage us for things like data transport. They wanted to leverage Western Digital’s vertical integration here to create some specialized storage servers.”
The core demand here comes from the increased demand for new, distributed intelligent architectures outside of core data centers to help ingest, analyze and transform data at the edge. This is particularly the case in remote and rugged locations where networks can be expensive, intermittent or nonexistent. These remote edge use cases, which include HCI where compute is also provided, are increasingly being required to get the processing of edge data even closer than edge data centres.
“The remote edge value proposition is to provide a cloud-like experience, but in more local environments, with or without a network connection, and to support AI and machine learning for analytics closer to where data is captured,” Hamilton stressed.
“With this announcement, we increase presence to the device edge with the UltraStar Edge server,” he added. “There are two variants. One is commercial, while one is for much harsher environments including MIL-SPEC ones.”
The more rugged version is the Ultrastar Edge-MR, an extremely rugged, stackable and transportable server for military and specialized field teams working in harsh remote environments. Aside from the military, it also fits more demanding use case environments, like the oil and gas industry. This version comes with 40 CPU cores and 512 GB memory, and over 60 TB of NVMe flash storage. It also has a graphical processor unit for AI and machine learning tasks. It meets MIL-STD-810G-CHG-1 standards for limits of shock and vibration, as well as the MIL-STD-461G standard for electromagnetic interference.
The other version is the Ultrastar Edge, a transportable 2U rack-mountable server with a portable case for colos and edge data centers, which is aimed at more traditional commercial use cases at the edge, like retail and warehousing. Both versions meet FIPS 140-2 Level 2 security standard, and feature the Trusted Platform Module 2.0, a tamper-evident enclosure.
The initial Go-to-Market focus for these servers is select systems integrators like San Diego-based Aeon Computing who specifically address federal markets and harsh environments.
“We think that this will grow over time, because there’s absolutely a growing need that isn’t being addressed,” Hamilton said. “We think that more SIs will be added to this list. While this is the first version of the UltraStar Edge, with a good balance of CPU, GPU and storage, we expect to see a broader range going forward to address new opportunities. This would include smaller form factors, more GPU in certain use cases, or other combinations.”
Both versions of the Ultrastar Edge server solution are now sampling and orderable with general availability beginning CQ4 2021.
