Most data center managers who direct and maintain operations of mid-sized data centers are not quite yet ready to consider the jump to 400G.
While certainly on the near horizon, their attention is set on transitioning from current 10G to 100G infrastructures or to 25G infrastructure as an interim step.
While technology advancements (especially in PAM4 and NRZ schemes) are at the core of the 400G transition, the 100G movement for data centers is not technology driven. 100G technologies intended for data centers were conceived about 15 years ago, indicating 100G should be considered a mature technology ready for widespread adoption.
This is best exemplified in the past couple of years where the IEEE’s P802.3, which governs the standardization of Ethernet, has concerned itself with adding some Physical Layer specs and management parameters for the full range of higher speed Ethernets (including 100G). Further, they've worked on a creating a standard for 100G PHY defining 100G over a single wavelength on a DWDM and with a range of 80 km or more.
Two potential data center upgrade routes
- New Connections at Same Speeds: Some may decide that their existing architecture meets demands for the time being and that additional connections at current transfer speeds will suffice in the meantime. In essence, they can "make do" for now, rather than spring for an update. From a technology perspective, this approach is the "out of sight, out of mind" approach. But, from the financial standpoint, putting off purchases for a couple of years can yield substantial rewards
- New Connections at 25G/100G Speeds: Continuing with existing connection speeds is an option may make sense in the short term, but most centers are migrating to a 25G/100G option to better keep up with rapidly increasing demands while still limiting cost exposure. This transition from 10G is driven by the need for more bandwidth for video streaming or to offer flexibility in the server rack.
What about 40G?
In years past, 100G upgrades involved removing 10G components and replacing them with 100G and/or 40G links. 40G connectivity is rapidly decreasing and being replaced by Nx25G connectivity as it is less costly than 40G links. One 100G port can be comprised of 4x25G, connecting two servers at 50G (2x25G) to each server. 40G requires a run from each server from separate switch ports, but with 25G/100G, only one switch port is used and connectivity at the server is 50G not 40G. In the 25/100G upgrade there are 100G connections between switches and 50G (perhaps just 25G) to the servers.
Price and 100G
Moving forward, we expect to see an increasing number of 100G and 25G transceivers installed in data centers of all sizes. These transceivers are not particularly high priced on the per unit basis, the total transceiver bill can easily add up. This is where third-party sourcing of the necessary transceivers comes into play. With astronomical price benefits (as low as one-third of an OEM transceiver's cost in some cases) while still carrying full feature compatibility, third-party optics in the modern era are a powerful option for maximizing one's budget while preparing for the future.