Networking is the last of the 4 broad data centre equipment types which needs virtualising for the late 1990s vision of utility computing to become a reality. So it was great to meet Vyatta’s Kelly Herrell, CEO and Jon Arnold, EMEA Managing Director in London recently to discuss their revolutionary approach networking.
Vyatta is a small pre-IPO company based in Belmont, CA: founded in 2006, it has secured $45 million in funding to date from HighBAR, Citrix, JPMorgan and ArrowPath. Among its 1,500 customers and 30 distribution partners, 400 and 10 respectively are in EMEA – impressive, given it has only had an office in the region since the end of 2011. It has some big name customers including Dell, EMC, Rabobank, CSC, Honeywell, Raytheon, CSC and Zyco. Most recently Vyatta’s software has been deployed in large public and private Clouds, such as Amazon, Verizon, Dreamhost and CapGemini.
Vyatta’s software is free to download and runs on industry standard x86 servers
Vyatta Network OS software (currently at version 6.4) runs on industry standard x86 servers, providing router, firewall or VPN devices, although its most widely used deployment is for ‘east/west layer 3 routing and firewall functionality’. For an explanation of the technology definitions see Happy Router.
While its software is free to download, Vyatta makes money from a subscription edition by upgrading and supporting the software deployed on bare metal or as virtual machines.
Its latest product is vPlane, announced a month ago: it differs from its regular offering by de-coupling the forwarding plane from the controller: it provides a line rate of around 10Gb/s per core and scales linearly with additional cores. The advantage of its approach is that it runs on industry standard x86 processors rather than proprietary hardware devices.
The rise of x86 network hardware
It is part of Intel’s strategy to extend the use of its processors beyond servers to storage (where it claims x86 is used in 80% of currently shipping devices) and networking: so it wasn’t a surprise that its recent Xeon E5 (Sandybridge EP) announcement Intel OEMs included 12 network OEMs – among them Cisco, Dell, Nokia Siemens Networks and Radisys.
Intel has been working on networking for a long time, but really came into serious consideration when it modified its chips to help address the Advanced Telecommunications Computing Architecture (ATCA) market, where Sun was the dominant supplier. With the introduction of its new processor Intel claims to have improved performance a 100 fold since 2000 – hardly something that can be said of standard networking equipment.
AMD also provides a good x86 platform for networking with advantages over its competitor in price per core – especially at the high-end, although arguably it has been less focused than Intel on the networking area.
Vyatta believes that network products need to be less than half the price of Cisco’s to compete – Kelly claims his solutions are 10x and 100x less for current generation and vPlane respectively.
Is this network hypervising?
Kelly doesn’t mind if Vyatta is considered as a ‘network hypervisor’, although the company doesn’t itself lead with the concept. We can see a number of reasons why we need to consider defining the area carefully. In particular:
- Networking sits under all IT and Communications implementations and so in a way can be considered an enabler of virtualisation without modernising
- As we discovered in storage hypervising, the networking technology issues differ distinctly from those in the server area
- Decoupling software and hardware is not in itself hypervising – witness the many storage arrays in which this is done; there needs to be an element of heterogeneity (which in fact Vyatta has)
We think the concept will become more widely discussed and used – especially as users need to join the dots in the data centre from servers, clients and storage to networks.
As in the other areas there is clearly a profit pool for new suppliers to address – Cisco has a far more dominate position here (48% share in 2011 according to ITCandor) than EMC has in storage systems (21%). It will be interesting to see how it addresses Software Defined Networks and OpenFlow APIs in coming months, but it is highly unlikely to embrace a strategy which will reduce networking costs substantially or lead to commoditisation.
Vyatta in the mean time will be happy to take on these roles and continue in its aims of changing the economics of networking.