Lyatiss isn’t French for IT’s holy grail, but maybe it should be

Lyatiss, a startup that came out of a French research consortium wants to create a new communication layer designed for the cloud and the upcoming world of federated apps. The idea is to use software installed on servers in different clouds — Amazon Web Services (S amzn) to start with — to monitor and then remediate problems associated with network traffic flows.

Lyatiss has raised $4 million in Series A funding from Idinvest Partners and others. It has headquarters in Santa Clara, Calif. and a research office in Lyon France, where the company originally began as a spin out from Inria, the French National Computer Science Research Institute.

Lyatiss combines the ideas associated with software-defined networking –such as monitoring flow-level data and treating the networking layer as an abstraction — and applies them to business and performance rules associated with applications. In short, Lyatiss says it can do what many of the people who are excited about SDN really want — a way for the network to react and deliver what the application needs.

Pascale Vicat-Blanc, CEO of Lyatiss
Pascale Vicat-Blanc, CEO of Lyatiss
Pascale Vicat-Blanc, the CEO and a co-founder of the company, likens it to creating a version of TCP for the cloud. The company calls this application-defined networking, but it’s probably safer to think of it a means to tie the application to the performance of the underlying hardware despite the increasing layers of virtualization in the way. For example, customers using Lyatiss’ CloudWeaver software (delivered as a service, of course) have been able to track their network flows to see where CPU usage was heating up and where network bottlenecks were occurring, thus letting the customers reallocate or size up their virtual machines as needed.

The service differs from a cloud-based networking monitoring program product such as Boundary’s, which tracks individual packets at the network level in that it tracks the entire flow — which includes where the packet is going and what might be stopping it or slowing it down along the way as opposed to just noticing that it has slowed down or stopped. This level of information, which can include details like the performance of the CPU and how that affects the network, is far more detailed.

This brings us back to TCP. This protocol helps define how devices talk to the web, ensuring that all packets sent around the web join up with their buddies at your end device. The protocol helps divide and track the packets that comprise an email, a movie or a digital photograph. In much the same way CloudWeaver’s application-defined networking hopes to use flow information to track how an application performs across multiple virtual machines, web services and eventually clouds.

It has started with an Amazon-based service, and promises to monitor and then let developers tweak their AWS settings when something is running a bit too hot or slowly. Developers could use this to see an outage before it occurs and then take swift action using the same software.

Eventually the plan is for such things to happen automatically if the developer sets it up that way. For those working in the cloud where scale is essential, the ability to monitor such flows and then have the hardware react to the needs of the application is somewhat of a holy grail. Lyatiss hasn’t managed to achieve this yet, but that’s where it is heading.