The IBM Storage Scale System 6000, like all new solutions in the technology world, is optimized for – you’ll never guess it – AI. The device combines high throughput speed with high IOPS.

IBM introduces the IBM Storage Scale System 6000. This is a new storage system designed for demanding workloads at the cloud data center level. Like every competitor in the tech world, IBM is building a bridge to AI with the introduction of this device: the server is intended to meet “the strict AI requirements of companies”.

High performance

The A and the I are now essential ingredients in the marketing sauce of every announcement, but otherwise contain little that is new. This server is intended for intensive data-related workloads such as analytics and therefore also for AI training. The server offers up to 7 million IOPS and throughput of up to 256 GB/s for read workloads. The device requires 4U of space for this. IBM claims this is much better than what competitors can offer today.

Some AI-specific features have been added. For example, the device supports Nvidia Magnum IOTMDirect memory, which provides a direct path from the GPU to memory. In addition, the Storage Scale System has 6000 Nvidia ConnectX-7 NICs on board.

The device can handle parallel workloads and is ideal for multi-tenant environments. Data encryption and compression are integrated. The operating system is the storage scale software, which, among other things, can connect to a wide ecosystem of hyperscalers to be part of a hybrid solution.