By: Bernt Ostergaard, Service Director, telecom & IT Services, Quocirca
Published: 17th September 2013
Copyright Quocirca © 2013
Every week brings fresh exponential growth estimates of data and traffic volumes on the internet. Much of this is driven by fast expanding cloud services, the LTE explosion of mobile data and the internet of things. This forces us to adopt still more efficient network protocols to handle more and more concurrent communication streams.
The main effort is centered on adapting layer-2 Ethernet with its LAN origins to the WAN. Serious traffic bottlenecks are now threatening the inter-data centre traffic where the present VLAN limitation of 4000 concurrent sessions just doesn’t cut it. Virtualisation does provide some relief with the VXLAN and NVGRE virtualized expansions that tunnel layer 2 packets over layer 3 networks with a theoretical max of 16 million concurrent streams. However, individual connections are very difficult to identify (requiring DPI), let alone provide any QoS assurance for. The internet is screaming for more speed and capacity, but service providers are understandably leery at the prospect of driving that fast in the dark.
Enter the The CloudEthernet Forum (CEF). This is a new global industry alliance founded in May 2013 by a group of eleven vendors, internet exchanges and cloud service providers. Their purpose is to address the fast growing global need for scaling and enhancing Ethernet technology to meet the stringent demands of delivering cloud services.
The global cloud services landscape does show regional differences with the US aggressively pursuing innovative new cloud business models, whereas Asia is primarily green field installations, with Europe positioned somwhere between the two. Here competition is not cut throat and managed services are the preferred cloud service delivery vehicles.
Ninety days after inception, CEF today announces key appointments to its board, and a 23rd of September publication of its initial technical committee white paper, addressing cloud types, deterministic traffic performance, end-to-end provisioning, overlay tunnels, and provisioning resources. Later this autumn follows regional meetings and membership recruitment drives in the US, Europe and APAC. Next Spring, CEF president James Walker from Tata Communications hopes to have three times s many members and see the first bilateral inter-DC connection trials.
Getting that far, this fast, demonstrates the speed at which cloud services are evolving with equipment generation of between 4–9 months. The CEF speed of delivery is also due to the ride it is hitching on the coat tails of the MetroEthernet Forum (MEF), a highly successful industry forum with well-established organisational procedures and connectivity standards like ENNI that CEF can build on. The CEF development work builds on the MEF’s Carrier Ethernet for Cloud Services 2.0 that focuses on the multi-tenant cloud service provider (CSP) connection with customers. CEF concentrates on the Ethernet communication between CSP data centres.
Besides evolving higher capacity connections that are secure and manageable between CSP data centres, CEF also addresses the issues of federating data centres around the world to reduce latencies. Broadly speaking, CEF wants to emulate what Google has done with its own global proprietary network and data centres, and make such solutions avaliable across the global WAN CSP community.
According to the CEF president James Walker from Tata Communications, this will be achieved in several stages. Initially the CEF members will develop best practices and design principles. Next step is to enter into bilateral CSP agreements, supported by switching vendors and major internet exchanges with a few truly global refrence customers, in order to gain real world experience before entering early stage standards work, which may involve many different standard bodies (IEEE, ITU, NIST etc.). CEF wants to avoid multiple standard version chaos, as we saw with WiFi, and recognises that the acceleration in cloud developments can only be done seamlessly if a sufficient number of key players join the Forum.
This gets us to the nub of the CEF strategy: there is no time for formal standards, so the Forum needs to get all the major stakeholders on board, and then evolve best practice on the fly. Right now, CEF is dominated by the vendors (Juniper, HP, Avaya, Huawei, Cyan, Citrix, Alcatel-Lucent, Cyan and Ericsson), but there are also representatives from the telcos (Verizon, Tata Comms and PCCW), and internet exchanges (Equinix, Coresite and telX). Notably missing is Cisco, which may be hesitating because of the Huawei and Juniper participation, but Cisco support will be crucial given the company’s heavy presence in transmission protocols, switching and cloud servers. Also missing are the storage players, notably EMC, with solutions for flash cache and storage and faster retrieval of terabytes of data, and in the near future also software defined storage in the NILEproject. Finally, CEF needs to attract the global system integrators. TCS seems an obvious candidate given its close relations with Tata Communications, but also IBM, CSC, Capgemini, Atos and Dimension Data.
So CEF is in a rush, both to stay current with the technology developments in the industry, but also to secure a decisive place in defining the transmission protocols for the intra cloud services space. Supported by MEF with its 200+ members CEF is well positioned, but the next few months may decide the fate of this new forum.
We automatically stop accepting comments 180 days after a post is published. If you would like to know more about this subject, please contact us and we'll try to help.
Published by: electronicdawn Ltd.