As more and more organizations start harnessing artificial intelligence in their analytics projects, the vendor community is stepping up its efforts to address the trend. IBM led the charge last week by adding a new model to its Power LC server line that is specifically geared towards running machine learning algorithms.

The two-socket server is marketed under the name S822LC and can be equipped with up to four of Nvidia Corp.’s Pascal P100 accelerators, which supposedly provide as much as three times better throughput than its previous-generation chips. The GPUs are linked to the machine’s main processors using a technology called NVLink that allows for data to travel 5-12 faster than the PCIe interfaces included in traditional machines. As a result, Big Blue says that the server is able to provide 80 percent more performance per dollar in certain situations.

Processing speed is also a key selling point for ScyllaDB Inc., an emerging database startup trying to make its mark on the fiercely competitive NoSQL space. The outfit last week unveiled a new version of its Cassandra-based columnar store that promises to provide up to 10 times more throughput than the upstream project while maintaining sub-millisecond latency. And ScyllaDB claims that its system also makes more efficient use of hardware resources, which enables organizations to reduce their infrastructure expenses.

The startup released its update against the backdrop of Yahoo Inc. open-sourcing Pulsar, a homegrown message broker designed to quickly shuffle data among applications. Like ScyllaDB, the web giant placed a strong emphasis on latency during the development process: Its system can handle hundreds of millions of records per day with average publish delays of less than 5 milliseconds. This feature should make Pulsar an attractive alternative to the more popular Kafka message broker from LinkedIn Inc., particularly for real-time applications that need to process information while it’s still fresh.

Image via Flickr