Two recent developments promise to alter the ways that compute environments are built and managed. The first, white box switches, lowers network component costs while increasing the options available to network managers. The second, hyper-convergence, simplifies management by delivering a complete compute environment as a single unit.
White box switches
A white box switch consists of a commodity or bare-metal switch with a network operating system installed. Bare-metal switches offer an open, standardized interface. A variety of network operating systems have been designed for that interface.
White box switches offer these benefits: Prices are reduced and flexibility is increased in the same way that the open, Intel-based architecture for servers reduced server cost compared to earlier proprietary products. System managers can choose a server with the processor speed and amount of memory needed for the application and then choose to install Windows or any of the Linux distributions to operate it.
Similarly, network managers can choose a switch with the capacity required from among any of the bare-metal switch vendors and then select from among the network operating systems designed to the open switch interface. Options for white box switches range from freely downloadable network operating systems to operating systems supported and maintained by such vendors as Big Switch Networks, Cumulus Networks or Pica8.
Another alternative is software from Open Network Linux (ONL), an initiative from the Open Compute Project. ONL does not include all of the components required for a complete operating system, but it provides a foundation that developers can use to create a system to meet their own requirements.
Bare-metal switches vendors such as Accton Technology, Alpha Networks and Quanta Computer incorporate switch chips from Broadcom, Intel, Marvell or Mellanox. Each chip vendor offers a series of chips with differing capacity in terms of number of ports and per-port bandwidth.
Switch vendors incorporate these chips along with a processor, memory and network interface components to create a complete system. Most offer a series of products meeting requirements from the network edge to the core.
Network managers can purchase hardware directly from a bare-metal switch vendor and add software. They can also purchase a fully featured and supported system that includes hardware and software. The variety of hardware and software makes it possible for companies to reduce their costs and benefit from more flexibility in how they construct their networks -- especially compared to earlier proprietary options.
Purchasing compute, network and storage components, and integrating and managing them, has grown too complex for many IT operations. Hyper-converged systems provide everything needed -- compute, storage and network resources -- as a single unit. All components are configured and integrated, so installation is quick. The system is managed from a single console by a single operator. Virtual machines can be brought up quickly since there's no need to issue a series of compatible commands to individual components.
When requirements grow, you can add resources by purchasing additional units. When installed, they will combine with existing units to create a single environment still managed from the same console.
Today, IT groups must react quickly to changing requirements. Hyper-converged systems enable IT resources to be put in place, become productive quickly and keep pace with rapidly shifting business demands. Combined, white box switches and hyper-convergence offer excellent ways to breathe new life into your data center network.
What white box switches mean for open networking
A Who's Who in the world of white box switches
A roadmap to the hyper-convergence market