Server fleet management less complex and nebulous with Nebulon

Nebulon is getting customer traction with its server fleet management server control cards and cloud management, as customers realise they can save big bucks by buying servers integrated with the cards.

The company briefed IT Press Tour attendees about its vision and progress on the partner, customer and product development front. Its technology is based around intelligent so-called Storage Processing Units (SPUs), each hooked up to a host server across a PCIe 3 link and providing a means of booting the server, its direct-attached storage, and managing its software and firmware environment. 

The briefing took place under Covid-19 restrictions and here a masked Craig Nunes holds up a Nebulon swag t-shirt.

Co-founder and CEO Siamak Nazari said Nebulon was partnering several server vendors –  Dell, HPE, Lenovo and Supermicro – and customers buy servers plus Nebulon cards from their strategic server supplier.

We asked about Cisco and COO Craig Nunes said “We haven’t spoken to them recently.” He implied Cisco preferred to sell its HyperFlex hyper-converged system rather than a UCS server + Nebulon card combination.

Nebulon has aspirations to partner with global system integrators, such as Capgemini, but doesn’t have contracted resellers – they are contracted to the server vendors and Nebulon’s invite-only smartPartner program aims to help them sell Nebulon card-equipped servers with rebate incentives, training, and demand creation programs, etc.

Nebulon says it provides Nebulon smart infrastructure for cloud repatriation and four products:

  • smartCore for VMware;
  • smartCore for Kubernetes;
  • smartiaaS for CSPs;
  • smartEdge – single-API, centralised fleet automation and monitoring [2-node HA starting footprint – zero-config cloud quorum].

It sees interest from large customers buying lots of servers who are burdened with server fleet management tasks, requiring servers to run them and admin time to update operating systems and firmware.

Nunes talked about a global CSP customer which had 250 VMware servers, and 60 storage arrays with 2PB of capacity, spread across 10 datacentres. This CSP did look at using HCI (Hyper-Converged Infrastructure) but it needed some of the HCI servers to run the infrastructure and thought that overhead was inefficient.

It decided to migrate to a Nebulon-equipped infrastructure and the migration is about a third of the way through. The CSP has about 100 Nebulon-enabled servers and 1PB+ storage across four datacentres in Europe and North America. Storage vMotion was used to move data from the arrays to the Nebulon-equipped servers, which have between 12 and 30TB of capacity.

Coming features

Nazari said “We allow 30TB attached capacity per CPU. It will soon double to 60TB.” He also said Nebulon supports PCIe 3 now and will add PCIe 4 support in 2023.

Tobias Flitsch, head of product, said that, over the next two quarters, Nebulon will add:

  • Server management integration – centralised hardware and firmware inventory and management across heterogeneous server vendors;
  • Nebulon Ansible collection – tools to configure servers post-deployment just with Ansible without needing extra software/services (Ansible connects to Nebulon-loaded server and loads system software items);
  • Instant cluster recovery – instantly recover an entire cluster.

The latter might be needed when network monitoring of IO patterns indicates a malware attack – ie. dedupe ratios change or IO pattern changes from random to sequential. In future Nebulon could add machine learning to detect this and tell the customer it has detect anomalous behaviour on the storage layer.