This article is more than 1 year old

Cisco's server CTO says NVMe will shift from speed to capacity tier

Raghunath Nambiar says data centres will be asked to do more with more

Interview NVMe storage is becoming denser, faster, than other forms of storage and will therefore become a capacity tier according to Cisco's chief technology officer for UCS Raghunath Nambiar.

“Right now people are looking at NVMe from a performance point of view,” Nambiar told The Register in Sydney last week, “but the real game changer is going to be capacity.”

Nambiar said 2.5 inch SATA/SAS SSDs will soon hit seven-terabyte capacities, but “NVMe's roadmap will go to 32 terabytes 18 months from now.” That density will mean that even small servers like the UCS B200, Cisco's half-width workhorse, will be able to work with 64 terabytes of data in each server and plenty more across a blade chassis or a fabric.

Nambiar said businesses will put that data to work with more intensive just-in-time analytics.

“Many of the places in the States have a decent knowledge about your buying patterns,” he said, and use that data to power recommendation engines. Consumers, he added, have a tolerance for about one second's wait while an automatic teller machine or online retailer processes their transaction, but business wants their analytics to deliver.

“Today they may look at 100 records,” he said. “In six months, they will look at 1,000 records.” Before long, he thinks organisations will expect that kind of analytics to consider tens of thousands of records in an effort to up-sell, cross-sell or just delight a customer. But those customers will expect the same response time irrespective of the number of records analysed.

To respond, data centre operators will need to get ready to host lots more data in the same racks they use now and Nambiar said 100G Ethernet networks will soon become table stakes. Scaling out by improving throughput will also become critical, in addition to relying on the sheer grunt of new CPUs, GPUs, FPGAs and networks.

Nambiar said Cisco will plans to help in a few ways, including more work on validated designs for emerging workloads in big data and analytics and enhancements to its fabrics. “It is not going to be plug and play,” he warned. “If you want to fully take advantage of these capabilities you need to think deeper, down to data management platforms.”

The recently-announced Starship server automation platform and further enhancements to UCS Director will try to do lots of that in advance. Nambiar said Cisco's aim is to make UCS a hands-off proposition not long after racking and stacking has finished, with OS and applications self-installing and optimising themselves ready for desired workloads.

“We will apply the known 'tunables' based on our best practice,” he said.

“The server is going to be a platform to run applications,” Nambiar concluded. “Hardware itself is much less important. The management layer orchestration, and the application layer is what matters.” ®

More about


Send us news

Other stories you might like