English

Data Center Switches: Timing the Move to 400G/800G

Updated on Jun 1, 2022
1.4k

data center switch

Today, the data center switch market is migrating to 400G/800G. The main reason is that the popularity of cloud computing and cloud services has driven the demand for higher bandwidth, faster speed and lower latency performance, which means that high-speed switches are more widely used, such as 400G and 800G switches. Read on to learn how the data center landscape is evolving and what's changing in data center switches.

Changes in Data Center Landscape

Throughout the evolution of the data center, the increasing variation in bandwidth, fiber density, and lane speed has had a significant impact on the data center landscape. The popularization of 400G networks in 2019 has promoted the development of 400 technology, and 400G switches have also begun to occupy part of the data center switch market. The derivation of 800G in 2022 also means the development of hyper-converged data center networks. What are the reasons for these changes? We analyzed it from three aspects.

Ethernet speed roadmap

 

     "Also Check- Data Center Switches

 

Explosive Growth of Global Data

Data centers store all the data of individual consumers, businesses or organizations, and large economies. As the Internet grows, so does data. Data centers need to change their ability to transmit data to adapt to the new network environment. 400G data center switches have been extended to smaller cloud service providers and large enterprises as an integral part of network architecture.

The following list shows some key trends in data usage around the world.

  • Social media enjoys explosive traffic growth.

  • Massive small cell densification enables the rollout of various 5G services.

  • Developments of IoT and IIoT (Industrial IoT) are gaining momentum.

  • Traditional office-based work is shifting to remote options.

The Evolution of Data Center Spine-Leaf and Fabric Architectures

To meet the development requirements of high bandwidth, high availability, and low latency network in data centers, the emergence of spine-leaf and fabric network architectures can simplify data center networks and provide higher capacity. Backbone networks require data center switches with higher bandwidth to carry traffic to meet the growing demands of end users. Changes in network architecture have also enabled data centers with more flexible connectivity, higher fiber-count backbone cables, and improved scalability. Using 400G switches or 800G switches, data center carriers can upgrade to higher bandwidth networks.

Growth of Hyperscale and Edge Data Centers

The data center landscape as a whole is significantly impacted by the hyper-scale and edge data centers. Hyperscale service providers are working to bring cloud computing services closer to the end user and their devices as demand for lower latency performance grows. Edge data center switches are also gaining traction for this reason.

Due to the shorter distance between a user and the processing network in edge data centers, data can be transferred faster, providing instantaneous and optimal application and data access at lower latency. At the same time, the expansion of hyper-scale and edge data centers forces owners and operators of smaller data centers to modify their infrastructures to accommodate the ever-increasing demands of traffic.

 

     "Also Check- How 400G Has Transformed Data Centers

 

How the New Landscape Affects Data Center Switches

For data centers, performance is actually a matter of checks and balances among servers, switches and connectivity. Data center switches are definitely affected by the changing landscape of data centers. All network operators are looking for switching solutions that are scalable, dependable, and efficient.

Architecture Shifts From ToR to MoR/EoR

The switch from a top-of-rack (ToR) architecture to a middle-of-row (MoR) or end-of-row (EoR) configuration is one significant modification. MoR and EoR require fewer rack-mounted data center switches than ToR architecture. It means less rack space, less work to maintain, and less power and cooling system use.

Additionally, rather than replacing the longer switch-to-switch links, upgrading to higher speeds only requires replacing the server patch cables. However, in order to facilitate connections between the in-row servers and the MoR/EoR switches, the shift necessitates a structured cabling strategy.

Higher Switch-Port Densities

Closely related to the shift from ToR to MoR/EoR is the increase in data center switch speed. When IEEE802.3ck becomes a ratified standard, serializers/deserializers (SerDes) that provide the electrical I/O for switch application-specific integrated circuits (ASICs) are anticipated to reach 100G. This indicates that switch ASICs are also increasing the density of I/O ports.

It is possible to reduce the number of top-of-rack (ToR) switches by supporting more network device connections with higher switch port densities. As a result, the number of switches required for the data center architecture decreases overall.

The use of QSFP-DD transceiver modules, 400G fiber optic cables, and 200G/400G switches in data centers is becoming increasingly common. Due to the availability of 800 Gb/s optics, Google is anticipating switch shipments of 800G switch this year.

Ever-Increasing Market Demand

The data center switch market largely fulfilled Dell'Oro Group's expectations in 2021. Sales of data center switches in the cloud sector increased by double digits, while sales in the non-cloud sector increased by mid-single digits. By 2026, the global market for data center switches is expected to be worth USD 19.9 billion, growing at a CAGR of 5.6%. The market appears to have survived the COVID pandemic.

global data center switch market

The cloud segment of the data center switch market is expected to expand at almost twice the rate of the non-cloud segment in 2022, according to Dell'Oro Group's forecast. The following are major factors that contributed to such a robust market forecast:

  • Impulse consumption driven by ongoing supply issues

  • Faster speed of digital transformation

  • Expansion cycle driven by new AI workloads

 

     "Also Check- 400G Optics in Hyperscale Data Centers

 

Data Center Switch Market: Opportunities & Challenges

The rosy picture for data center switch market certainly brings along great opportunities, but there are also challenges lying ahead in the transition to 400G/800G.

Opportunities

Silicon diversity: Silicon diversity has been a theme in the data center industry for the past few years. Diversity also puts some pressure on Broadcom, a semiconductor giant.

In 2022, it is anticipated that industry-wide supply constraints and an increase in the number of viable merchant silicon suppliers like Cisco and Marvell/Innovium will further accelerate silicon diversity. At OCP 2021, Cisco said that it would provide Silicon One chips on Wedge400C for top of rack applications to Meta, which used to be Facebook.

Smart Devices: Technological advancements in smart devices drive the need for sophisticated connectivity and enhanced networking solutions. This surge is expected to drive the integration of chips in data center servers, which will provide lucrative growth opportunities for the global data center switches market.

Challenges

Data center operational cost: Data centers need to take into account the local energy prices because energy costs make up a significant portion of the overall operational cost. The cost of energy on its own can be prohibitive for cloud service providers and hyperscale data centers. The market's expansion is also stifled by additional operational expenses like machine upkeep and labor.

Complex architecture: Data center architectures are becoming more complicated due to cloud computing, server virtualization, computing, and storage technologies. Although high-performance and higher-bandwidth data center switches can handle enormous workloads, implementing high-bandwidth solutions across various architectures still presents numerous challenges.

In addition, it becomes difficult to establish compatibility between various technologies used in data centers, which can result in significant expenditures and hold up new deployments.

     

     "Also Check- 400G Data Center Deployment Challenges and Solutions

 

Finding the Right Data Center Switch

The emergence of 400G data center switches and optical modules can help data centers upgrade from the original 100G to 400G. Although the 400G switch is an upgrade of bandwidth, at the same time, it is an overall improvement from hardware bandwidth, deployment, investment protection to SDN solutions and software programming.

If you are building a 400G data center and expect to choose a satisfactory 400G data center switch for your data center, here are some recommendations for you. You can see if it meets your needs by checking the specific performance and functions of the device.

FS 400G data center switches are the 4th generation of Spectrum switches, purpose-built for leaf/spine/super-spine data center applications. They come pre-installed with NVIDIA Cumulus Linux (Open Networking) for advanced automation, customization and scalability using web-scale, making them ideal for web-scale IT, cloud, hyper-converged storage, and data analytics applications.

Switches Ports PTP, SyncE, etc. MLAG, BGP-EVPN, VXLAN, MPLS,etc.
MSN4700-WS2RC 32x 400Gb QSFP-DD Support Support
MSN4700-WS2FC 32x 400Gb QSFP-DD Support Support
MSN4410-WS2RC 24x 100Gb QSFP28-DD | 8x 400Gb QSFP-DD Support Support
MSN4410-WS2FC 24x 100Gb QSFP28-DD | 8x 400Gb QSFP-DD Support Support
N9500-32D (Bare-Metal) 32x 400G QSFP-DD / /

You might be interested in