Pensando bags big bucks to take on AWS Nitro

MPLS, the renowned Cisco spin-in quartet, are back in town with their latest startup, an edge computing company called Pensando Systems that seeks to out-do Amazon Nitro. Only this time it is HPE that is doing the backing.

Pensando emerged from stealth last week, announcing it had raised $145m in a Series C round led by HPE and Lightspeed Venture Partners. It will use the money to accelerate engineering, operations, and go-to-market activities.

Pensando claims its proprietary accelerator cards can turn existing on-premises architectures into next-generation clouds.

It said cloud service providers can use its product to gain an edge over Amazon’s Nitro for specialised compute instance and said the accelerator cards perform five to nine times better in productivity, performance and scale than AWS Nitro.

Nitro systems are dedicated hardware cards using ASICs that offload networking, storage and management tasks from EC2 host servers. AWS uses Nitro to design and deliver EC2 instance types with a selection of compute, storage, memory, and networking options.

Pensando claims to be able to do the same kind of customisable host server hardware-offload – only better, with no risk of lock-in and in a scale-out fashion.

The company said its developing product has been influenced by input from vendors such as HPE, NetApp and Equinix and it is already used by multiple Fortune 500 customers such as Goldman Sachs, an investor.

Fab four

Pensando was set up in 2017 and has had three funding rounds, totalling $278m – the latest reportedly gives a post-money valuation of $645m.

  • Series A founder-led round of $71m,
  • Series B customer-led round of $62m,
  • Series C round to raise up to $145m.

The founders are Mario Mazzola, CEO Prem Jain, Luca Cafiero and chief business officer Soni Jiandani. They are the MPLS team – the acronym is derived from their first names – who joined Cisco when it bought Crescendo Communications in 1993. They built eight $bn/year spin-in businesses for Cisco and were richly awarded for their labours.

The software-defined networking business was devised at Insieme, a spin-in startup acquired by Cisco for $836m in 2103. SAN switch developer Andiamo and Nexus switch and /UCS server inventor Nuova were two previous spin-ins at Cisco and Mazzola, Cafiero and Jain were involved with both; Jiandani was at Nuova.

The team were close to CEO John Chambers and they all resigned when Chuck Robbins replaced him in 2015. Chambers is Pensando’s chairman and an investor.

Close to the edge

Pensando’s product is a programmable edge computing accelerator that offloads a host server to deliver cloud, compute, networking, storage and security services which can be chained together in any order.

The edge angle is that edge computing will mean more data has to be processed where it is generated. Pensando helps servers to keep up, as its technology can terminate and encrypt one billion IoT connections in half a rack of servers.

The technology is called an Edge Services Platform and is based on a custom, programmable processor designed to execute a software stack delivering the various services, which is managed by a Venice centralised policy and services controller. It comprises Naples 100 and Naples 25 Distributed Services Cards (DSC) which deliver the service functions to a host server. They have:

  • Custom processor chip to accelerate on-premises servers
  • Software-defined customisable distributed infrastructure  services across cloud, compute, networking, storage, security, and virtual appliances
  • Edge-acceleration, scaling out linearly with any server environment and minimising latency and jitter while freeing up host CPU cycles
  • 100Gb/sec wire-speed encryption, and hardware isolation
  • Compatible with virtualized or bare-metal servers as well as containerised workloads
  • APIs for management.

The Venice controller distributes infrastructure services policies to active Naples nodes. It handles lifecycle management such as deploying in-service software upgrades to Naples nodes and delivers always-on telemetry and end-to-end observability.

The DSC-host attachment method is not revealed and nor is the environment required for the Venice controller software.

Known unknowns

Pensando claims the DSCs reduce host CPU utilisation by 20-40 per cent. This implies the DSC card fits between the host server and existing storage and networking resources. Pensando doesn’t say if this is the case, and doesn’t say how the card connects to those resources, but implies 100GbitE. Nor does it say which host server environments are supported. We are not told how the DSC card is programmed, nor how Pensando’s technology provides cloud services.

It stands to reason that potential Pensando customers have large and sustained incoming data flows requiring a workflow that can be implemented using Pensando programmable hardware to assist the host server’s CPU. The freed-up CPU cycles enable it to do more work and enable customers to avoid buying specialised devices or software to provide the functions that Pensando can host. For example, Pensando said its DSCs can replace firewalls, load balancers and encryption devices.

General availability for the DSCs has not been revealed.