Can ibm's fabric really support 10k tx per second?

Read an IBM’s paper and they claim that Fabric can support 10k tx/s. Anyone know how that comes? Like which consensus algorithm they use and under what kind of hardware environment?

bitshares claims 100k tx/s using delegated proof of stake.

EDIT: from

IBM® Fabric Manager (IFM) is a solution that enables you to quickly replace and recover compute nodes in your environment.

sounds like a closed environment to me.

part of the:
Flex System product overview
Lenovo Flex System products are ideally suited for data center environments that require flexible, cost-effective, secure, and energy-efficient hardware.