English
Deutsch
Espaniol
Cloud data center's rapid growth regarding networking and storage is the thing Cisco has tracked through the Global Clouds Index for many years. Our current forecast indicates that we will soon meet and surpass the 14ZB target for cloud traffic near 2020. It is astounding to consider the world marveling at achieving the 1ZB milestone two years ago.
Consequently, interconnect and networking industries are impelling their technology advancement teams to encounter the persistent and ever excessive bandwidth connectivity requirements. The next significant transit is the general accessibility of 400GbE technology, which will soon undergo trials at several telcos, cloud providers, and other conditions that require excessive bandwidth connectivity.
Nevertheless, there is no doubt that the 400 Gigabit Ethernet will be famous soon. We still have active discussions about which SFP we can best deliver this technology. We can acknowledge that price is the primary concern for technology transition of any kind. So, taking the fiber optic volume high to bring the price down is necessary for an easy change.
With the modern trends, the optic price has become a problem in transition: the decrease of ASIC price for one bit is overtaking optics. This blog will support utilizing modules QSFP-DD as the perfect connectivity choice for the coming generation cloud, including other large-scale data centers.
The necessity to Control Costs:
We have significant expenses associated with advancing modern technology with modern fiber optic form factors. Every change within the form factor obstructs the volume ramping and accelerates associated prices. These are considerable difficulties to smooth perfect conversion from one to another technology.
Preferably, the "final" extreme volume form factor regarding new speed has to be similar to the earlier version (aka backward adaptability) as detours via changing form factors decrease market traction, driving price high. Backward adaptability confirms that customers have an approach to the technologies we have already improved for our designed programs.
System and network requirements like density make fiber optic module demands, and optical component unification capabilities make fiber optic module feasible. Previously, this has caused initial massive form factors conversion to more minor, lower cost, and denser form factors with the movement of technology towards excessive adoption. Earlier, we observed this cycle.
For instance, 10 Gigabit Ethernet fiber optic modules migrated from the huge pin module 300 to XENPAK, X2, XFP, and eventually to SFP+. Likewise, 100 Gigabit Ethernet modules changed from the earliest form factor CFP to CFP2/CPAK, CFP4, and eventually to QSFP28. Regarding Fiber Channel, we had no shifting from SFP all through the speed boosts from 1 Gigabit to 2 Gigabit to 4 Gigabit to 8 Gigabit to 16 Gigabit to 32 Gigabit since there was no change in use cases.
We have observed these changing form factors with Ethernet to appear as the full use case modified over time: at first long-range, low density before proceeding to the shorter range, excessive density, and volume. We have causes for the corporation to connect with a similar form factor. We have continuously miniaturized timely technologies having a more extended range to fit within the extreme volume providing an absolute solution. Nevertheless, the powerful frictionless development ways and the lowest price points always won.
Reuse advantages system, remarkable experience, and modules providers to help reduce prices and accelerate advancements. We also have a common preference to support the same port density of a switch for one rack unit, maintaining the proven fabric styles as we have architecture the networking equipment near them to attain specific densities. Keeping this consistency restricts the impact on our ecosystem. Eventually, the advantage of the use-case for utilizing the similar form factor can be significant because it enables different speed port possibilities on switches. We often require slower fiber optics in excessive speed ports to ideally move to excessive port speeds.
The Conversion from 100 Gigabit Ethernet to 400 Gigabit Ethernet:
The market necessity to attain 400 Gigabit Ethernet is primarily because of the moving market dynamics. After 10GbE's introduction, we deployed the early adapters far ahead of the excessive volume data centers implementations. At 100 Gigabit Ethernet, again, we deployed former service supplier adopters ahead of extreme volume data centers' markets, although the market needed the reduced progress cycle.
With 400 Gigabit Ethernet, we are observing the actual user conditions, telcos, cloud providers, and hyper scaling data canter wanting to accept the technology simultaneously like the bandwidth number, workloads explode, and hungry apps. With higher volume/lower distance use cases (almost
The industry has adopted the form factor demanding vendors to discover the perfect solutions to fulfill all those requirements. We have no luxury of operating through different iterations of the form factor to achieve the lowest price points that attract customers.
QSFP-DD is the earliest and the excessive volume last 400GbE's form factor. We have shown it can attain the different goals of backward compatibility, breadth of range support, and technical feasibility.
The excessive volume shorter range fiber optics are in progress for different fiber kinds, including fiber optic cables, parallel MMF, and duplex and SMF in QSFP-DD.
We regard cooling modules as a system design issue, and here the strength of corporation experience comes out. We expect some earliest modules QSFP-DD to dispel powers to 15 watts, whereas most are below that. Modern thermal designs within cages and systems show no cooling problems despite these excessive power modules.
With coherent technology's adoption, most difficulty shifts within silicon integration away from fiber-optic integration. Complementary metal-oxide-semiconductor (CMOS) integration has already enabled QSFP-DD to build for these "complex optics” within the design. We are working on the Optical Internetworking Forum to describe a 400 Gigabit Ethernet solution for DCI (Data Center Interconnect) which we call 400GZR (~120 kilometers DWDM) for form factor QSFP-DD. The more extended use case is technically feasible within the similar form factor QSFP-DD-technologies' transition signifies that the corporation no more has to hope for unification to overtake.
For network implementations, over-subscription proportions have to remain the same as 25 Gigabit/100 Gigabit shifts to 100 Gigabit/400 Gigabit. We need a similar port density for one RU, and we can't give up on this. System designers, end-users, ASIC, and networking developers all realize the significance of backward adaptability. Enabling networks and products that can adopt lower speed modules within a parallel port has a broad advantage from network enhancement strategies, development price reduction strategies, and price-reduction strategies. Although some network implementation strategies don't possess this as necessary, they still appreciate the lower prices we have enabled from handling the aggregate volume of the remaining corporation that may keep these requirements.
QSFP-DD: Getting Market Success
We have designed QSFP-DD to face the requirements, including an optimized system design for 400 Gigabit Ethernet to make itself a successful and dominant form factor. It begins with backward adaptability to 100 Gigabit Ethernet QSFP28 and 40 Gigabit Ethernet QSFP+.
The form factor QSFP's beauty lies in the corporation has building solutions above many years of experience around the form factor QSFP enabling rapid maturity and innovation. The three main challenges for all form factors are always:
- Will the connector operate at this excessive speed?
- Is there the possibility of cooling the module?
- Are the fiber optic fit?
We have adopted the "DD" approach to increase the other leading form factors (small form-factor pluggable becoming small form factor pluggable=DD) for coming time solutions. And to influence the backward adaptable solutions, we consider a sign of success. Carrying on the guideline of the system (10 Gigabit Ethernet SFP, 25 Gigabit Ethernet SFP, 50 Gigabit Ethernet SFP, 100 Gigabit Ethernet SFP-DD) and connects (40 Gigabit Ethernet QSFP, 100 Gigabit Ethernet QSFP, 200 Gigabit Ethernet QSFP, 400 Gigabit Ethernet QSFP-DD).
Conclusion:
So, we can observe that the industry is proceeding to achieve significant steps toward victory at 400 Gigabit Ethernet. The initially challenging task of creating a solution that rushed the conversion to excessive-volume, low-price solutions to face all market demands. We expect to accept 400 Gigabit Ethernet around a similar time and are making progress. We build the solutions in labs these days, and public deployments and demonstrations will follow shortly.
