Network Services Ramping Up in the Cloud
Not too long ago, network infrastructure appeared permanently rooted in the physical world. While storage and processing received the fruits of virtualization – consolidation, dynamic provisioning and all the rest – it seemed networking would forever limited by switches, routers, and wires.
We all know how that turned out. Software defined networking (SDN) made it possible to shift and twist network architectures just as easily as the rest of the data infrastructure, ushering in the possibility of the end-to-end virtual data environment.
Naturally, then, cloud providers have been quick to leverage the possibilities. Networking is now firmly ensconced on the virtual plane, and networking services aren't far behind. And of late, the type and variety of those services have increased by orders of magnitude through a number of key technology partnerships.
Rackspace, for one, has turned to Brocade’s Vyatta virtual switch platform to enable high specialized traffic management across its network services portfolio. The partnership allows Rackspace to push network segmentation to new levels, providing clients with broader control over the kinds of data allowed on certain pathways, as well as user access to the virtual network and other hosted services. In this way, Rackspace hopes to offer not only stronger multitenancy guarantees for multiple clients, but higher levels of service for clients pushing multiple workloads onto the cloud. And since the Vyatta platform is available as a virtual appliance, no messy hardware configurations are needed.
At the same time, Cisco Systems is looking to leverage its recently acquired Meraki outsourcing platform to allow cloud providers to deliver greater network management capabilities to their clients. Meraki users now have access to the Cisco Meraki Managed Services Dashboard, which can be used to build everything from simple LANs and WLANs to highly complex switching, application, and data management functions. At the moment, the dashboard works only with Meraki devices and virtual solutions, but it isn’t hard to imagine Cisco taking steps to enable broader network control before too long.
All of this activity on the services level does not obscure the fact that physical infrastructure can still bring an environment down if not properly cared for. As network speeds within the datacenter increase, so too does the sensitivity to jitter, given that individual packets have less wiggle room within high-speed data streams. That need is producing a new class of crystal oscillators like Silicon Labs’ Si535 and Si536 devices, targeted at everything from 10G and 40G data center environments to massive Carrier Ethernet platforms. The units provide less than 200 femtoseconds (fs) of RMS jitter between 10 kHz and 1 MHz, largely through the company’s DSPLL system, which provides low-jitter performance at high-speed differential frequencies. The devices also use a common fixed-frequency crystal, rather than a different crystal for each frequency, for improved stability and reliability.
The advent of virtual networking is the final straw that will break the enterprise dependence on in-house data infrastructure, according to Dimension Data’s Phillip Spies. Once freed of its silo-based physical architecture, networking becomes a fungible commodity, ripe for utility-style, pay-as-you-go delivery models. At the same time, many of the security and availability issues that have hampered cloud adoption to date will fade away, as many of these are routed in the static nature of current network infrastructure.
Is it likely, then, that enterprises of all stripes will begin to dismantle their owned infrastructure in favor of an all-service model? Probably not, at least not right away. There is a lot of capital tied up in today’s data center that can still provide valuable service.
But by the end of the decade certainly, the cost benefits of the cloud model should finally start to outweigh the management and security concerns, particularly for new organizations looking to build data architectures from scratch. And although it may take a while, ultimately the sight of even a large enterprise building and maintaining its own data infrastructure will be the exception, rather than the norm.