Cisco just held its PONC 2022 conference, with over 500 attendees registered. We found the most exciting topics at the three-day show were: (a) its progress with RON, (b) PLE availability, and (c) the 400G ZR ramp. Cisco is pushing for integrating optical transceivers with routers, a significant step forward for the industry.
The company indicated that it is seeing significant momentum for uptake of its Routed Optical Networking (RON) products; it now has 20 customers that have deployed RON networks. Additionally, there are 45 Proofs Of Concepts trials underway and more than 150 customers engaged with Cisco in various stages for RON products. Cisco touts that its RON systems result in a 78% decrease in power, a 76% decrease in rack units, and a 76% decrease in physical weight compared to using the traditional setup of separate routing and OTN systems.
Private Line Emulation (PLE) is now available for customers as a function on several of Cisco’s router product lines. We asked whether PLE can be made available to older routers, and the answer was yes, depending on customer demand to incorporate this feature. Cisco expects that it will have a significant number of PLE customers by this time next year who will take advantage of the PLE function to drive traffic that historically ran over OTN networks. The company shared its view that PLE improves latency when compared to an OTN system because its routers introduce a 5-10 microsecond latency (roughly equivalent to that of a 1 km span of fiber optic cable) compared to an OTN switch, which takes 40 microseconds.
We also learned that Cisco’s integration of recently acquired Acacia is proceeding smoothly; its 400G ZR modules are shipping. It plans for other modules to become available in the future, including 0 dBm ZR modules (autumn 2022 availability), a QSFP-DD Open Line System (OLS) (First Customer Ship in C1Q23). Google presented on Day 1 of the conference and explained that it sees a 50% CapEx savings by using 400G ZR instead of traditional transponder systems. It shared that it experienced significant OpEx savings related to power, space, and management cost reductions. And, Google said in the past three years, its bandwidth has growth between four to five times. Several of Cisco’s customers shared their experiences working with Cisco ZR and related products, including Arelion, CENIC, Colt, WebSprix, Windstream, and Zain. Windstream said that it got ZR+ working to over 1,000 km in its tests and deployments. Cisco’s team summarized its ZR success by sharing that its 400G Coherent journey from 2011 to current days has led to significant size/power/cost reductions of 40% Y/Y in the past decade.
AI/ML Workloads and Next-Generation Server Designs to Drive Growth
New network architectures and increased server speeds are causing the connectivity market inside the data center to move away from Direct Attach Copper (DAC) to active solutions. The current DAC technology can’t scale with the speed coming from next-generation server designs in the Cloud. Similar to other technologies, the Cloud will lead the way, but the broader markets of enterprise and service providers will enjoy the benefits of the technology.
Server speed is quickly moving to 112 Gbps SERDES and 100 Gbps ports for server access. There are several reasons behind this, such as the rapid move to PCIe Gen 4 and 5, faster processors, and the introduction of AI/ML workloads. But the more significant reason is the increase in accelerators via Smart NICs and DPUs.
2022-2024 server architectures take advantage of these factors and can drive multiple 100 Gbps lanes per server. DAC can’t scale as the distance becomes too small and a cable gauge becomes too large, and there is no appetite or need to move towards fiber or middle-of-row architectures. Cloud customers will be the first to proceed with the overall enterprise market following when they transition to higher speeds. These short distant server interconnects (up to 7m) account for over 50% of all data center network interconnects.
Beyond server access, customer preference continues to shift away from large Modular chassis and toward Fixed 1RU systems. At the same time, the aggregation and core layers are being moved closer together in a move away from the physical blast radius to the virtual blast radius. These trends allow aggregation switches to connect at short distances, not necessarily across a whole data center. As a result, these Distributed Disaggregated Chassis (DDC) architectures grow in volume each year and are prevalent in the three major verticals of Enterprise, Cloud, and Service Provider. DDC doesn’t need fiber for short distances, and one can think about active copper cables replacing the modular switch fabric in a chassis.
Last year, 90% of servers used DAC cables for connectivity, with the remaining 10% being a mix of 10G-Base-T, fiber, blade server enclosures, and other technologies. By 2026, we expect over two-thirds of the Cloud server market to be active copper (Figure 1). Furthermore, the percentage of 100 Gbps and above will be even higher, with almost the entire market at that speed using active copper. From a revenue perspective, the market will quickly exceed $1B and can exceed $3B by 2026, far higher than the DAC market it will replace.
Active copper cables fall into 2 main categories, ACC and AEC. ACC cables use an analog redriver approach to recover the signal that is attenuated by the losses in the copper conductors. AEC cables use a DSP retimer approach to accomplish the same. While AEC technology is able to achieve longer lengths, it also uses significantly more power. Spectra7 is a leader in the ACC segment.
The industry has moved technologies in the past, with the 1 Gbps to 10 Gbps as an excellent example of a drastic and rapid shift for server connectivity. The market for 1 Gbps was dominated by 1G-Base-T, but the need for 10G shifted rapidly to DAC, with 10G-Base-T never gaining traction beyond some enterprises. 100 Gbps should have the same trend, where there is a rapid inflection over the next 2-3 years, first starting with the Cloud providers and then shifting to the broader enterprise market.
Cloud providers currently are examining when to transition away from DAC, and we expect to see many new designs between now and the end of 2023 transition away from DAC. This times well with servers moving to 100 Gbps links and new 25.6T and 51.2T switch ASICs. While Cloud providers tend to be secretive, we expect to see demonstrations throughout the year as both cable and switch vendors demonstrate products and interoperability.
- Alan Weckel, Founder and Technology Analyst at 650 Group.
In 1Q'22, revenues for the Optical Transport market were 6% greater than the year-ago quarter, according to 650 Group's recently published report. The market overall has been inconsistent in performance over the past four quarters, oscillating between Y/Y growth and Y/Y declines.
On a regional basis, North America had the strongest Y/Y growth and on a Q/Q basis. We have maintained our revenue forecast for 2022 and 2023. For the outer years of our forecast, we are maintaining our long-term forecast since the previous report.
Cisco experienced the strongest Y/Y revenue performance, while ZTE experienced the strongest sequential (Q/Q) revenue performance.
The top 5 vendors in the quarter (ranked by revenue size in the report) were the following:
1) Huawei - grew Y/Y
2) Ciena - grew Y/Y
3) Nokia - declined Y/Y
4) ZTE - declined Y/Y
5) Infinera - grew Y/Y
The Optical Transport report breaks out over five vendors and includes detailed segmentation, forecasting, and market shares. Please contact us for press inquiries and further report information. In 650 Group's top vendor performance, we only include vendors broken out directly in the report. Vendor's in our rest-of-market category are not included in the top performing list.
650 Group is a leading Market Intelligence Research firm for communications, data center, cloud, enterprise, and SP markets. Our team has decades of research experience, has worked in the technology industry, and is active throughout the supply chain throughout the world.
In 1Q'22, revenues of the Routing market were 2% greater than the year-ago quarter, according to our recently published report. Over the last four quarters, the market has consistently grown on a Y/Y basis. Our report also includes an updated long-term forecast. 650 Group maintains our revenue forecast for 2022 and 2023. For the outer years of our forecast, we are maintaining our long-term forecast since the previous report.
The top 5 vendors in the quarter (ranked by revenue size in the report) were the following:
1. Cisco - grew Y/Y
2. Huawei - grew Y/Y
3. Nokia - declined Y/Y
4.Juniper - declined Y/Y
5. ZTE - declined Y/Y
The Routing report breaks out over five vendors and includes detailed segmentation, forecasting, and market shares. Please contact us for press inquiries and further report information. In 650 Group's top vendor performance, we only include vendors broken out directly in the report. Vendor's in our rest-of-market category are not included in the top performing list.
Different Scenarios can Play Out for the Networking Industry
While the potential Broadcom / VMWare combination could affect many industries, this blog shares our view on its impact on the networking industry. As an opening thought, this big deal may experience regulatory challenges, considering regulators prohibited Broadcom from acquiring Qualcomm and NVidia from acquiring ARM. This deal is not a lock.
If the deal closes, we expect Broadcom will follow its familiar playbook and cut headcount significantly. If past is prologue, Broadcom will slash sales staff and administrative positions immediately after the deal closes. Broadcom runs lean, and we would be surprised if it didn’t relocate the remaining VMware employees from lovely locations like Palo Alto to more spartan facilities. Once the initial synergistic changes are complete, we view three scenarios depending on how Broadcom executes following the acquisition.
Companies will seriously look at M&A. We find this interesting as the enterprise market for DC equipment is peaking as we speak (ASPs are driving revenue growth on mostly flat units/ports). The Hyperscaler market is now 2+ generations ahead of enterprise-focused systems, so there is already a complete bifurcation between what enterprise and cloud customers buy and use. Maybe we get two unique supply chains and very different market shares as a result of this corporate combination.
In 1Q22, revenues of the Telecom Core market were 4% greater than the year-ago quarter, according to our recently updated report. The market has been inconsistent in performance over the past four quarters, oscillating between Y/Y growth and Y/Y declines. The report also includes an updated long-term forecast. For the outer years of our forecast, we are maintaining our long-term forecast since the previous report.
Amdocs experienced the strongest Y/Y revenue performance, while Ingate experienced the strongest sequential (Q/Q) revenue performance.
The top 5 vendors in the quarter (ranked by revenue size in the report) were the following:
1. Huawei - grew Y/Y
2. Nokia - grew Y/Y
3. Amdocs - grew Y/Y
4. ZTE - declined Y/Y
5. Ericsson - declined Y/Y
Regional revenue performance in the quarter for the largest four regions was as follows:
North America - grew Y/Y
Europe - grew Y/Y
Greater China - declined Y/Y
Asia Pacific (excluding China) - grew Y/Y
650 Group's Telecom Core report breaks out over fifteen vendors and includes detailed segmentation, forecasting, and market shares. Please contact us for press inquiries and further report information. In 650 Group's top vendor performance, we only include vendors broken out directly in the report. Other vendor's in our rest-of-market category are not included in the top performing list.
650 Group is a leading Market Intelligence Research firm for communications, data center, cloud, enterprise, and SP markets. Our team has decades of research experience, has worked in the technology industry, and is active throughout the supply chain throughout the world. Learn more at http://www.650group.com
Fortinet had its Accelerate 2022 conference today. Its executive team presented its current strategy and expectations for the future and highlighted what it thinks are its key differentiators in the marketplace. The company explained: (a) the importance of its ASIC and its software code-base called FortiOS, (b) its services bundling strategy called "Fabric," and (c) its sales and marketing differentiation (the presentation was intended mainly for its selling partners as well as for financial analysts). We were reminded of the relative simplicity of Fortinet's messaging (Fabric, ASICs) and the solid financial results that illustrated its sales team and partners' success (it grew product revenue 54% Y/Y in the most recent quarter).
ASICs and FortiOS. The company explained that it takes around four years to develop an ASIC. According to its ASIC roadmap, it has three types of ASICs: Network Processing, Content Inspection, and Entry-level Systems-on-Chip. CEO Ken Xie explained that the company uses 7nm process geometry semiconductors for its ASICs, which we assume means for its CP10 (content inspection) based on the timeframe shown in its ASIC Roadmap. The company's marketing strategy around its ASICs is that these systems allow its firewall platforms to perform better than competitors.
The company's operating system software that runs on its firewall platforms has evolved to include many functions besides standard firewalling, including SD-WAN, Enterprise WLAN Controller, and SASE. What's interesting is that Fortinet is up till now has been taking an appliance-focused strategy and integrating many non-firewall services into its FortiOS. However, the company indicated that the OS will find itself in many environments in the future. Indeed, management highlighted that it sells FortiOS to four different environments today, including on appliances, as virtual software (that can operate in a data center, on a cloud hyperscaler infrastructure, or as cloud-native SaaS), as a container, or as cloud-delivered services. The company said that VM revenues are growing very fast, but expects its VM revenue stream to "go cloud-native." The company says it is "working on" a SaaS business and that its customers will consume these SaaS services through Fortinet's own cloud service. To summarize, management said that its R&D team is heavily investing in cloud designs and expects to make new cloud-related announcements soon.
Fabric. The company highlighted its sales strategy called "Fabric." CEO Ken Xie said that Fortinet charges less for its services, only around 20% of product prices and that when it uses bundling (which we interpret as synonymous with its "Fabric" strategy), it "discounts a lot." The company shared a slide highlighting "Broad Service with Half the Cost." The slide highlighted its FortiCare, FortiGuard, and FortiTrust services on this slide. On the company's earnings call earlier in May, it said it is "giving away" services in some cases. The company's Fabric offering includes services/functions/software such as "Access & Endpoint Security," FortiGuard Threat Intelligence," "Secure Networking," "Cloud Security," and "Network and Security Operations." If we were a competitor to Fortinet, we would find it frustrating to compete with Fortinet's discounting of these services – yet customers seem to embrace Fortinet's integrated product offerings.
Sales and Marketing. The company rapidly embraced SD-WAN a few years back, and now many of its firewall customers are using the same appliances to run SD-WAN networks. Likewise, Fortinet has embraced the SASE marketing moniker and has found large selling partners for this offering. Likewise, the company has adopted a selling strategy called "OT" that takes advantage of its ruggedized products (first introduced in 4Q20); it has experienced substantial growth in new vertical markets such as manufacturing and construction. The company's management has likewise embraced the industry trend of converging networking and security functions. It expects to sell using its positioning as a leader in "Secure Networking." Fortinet has experienced strong growth LTM product growth of 44% growth, which supports Fortinet's decisions to embrace SD-WAN, SASE, OT, and Secure Networking.
We participated in the Huawei Analyst Summit 2022 (#HAS22) for three days this week. The conference started with a keynote featuring Rotating Chairman, Ken Hu and then there were many executives and managers in charge of various operating groups at Huawei. To summarize, Hu said there will be challenges ahead and emphasized software, services and partnerships instead of asserting leadership in hardware as it did last year. Additionally, other speakers discussed the progress Huawei is making in F5G, introduced its latest Massive MIMO 64T64R radio system, discussed its progress with both indoor cellular and Fixed Wireless Access markets.
At the keynote, Ken Hu, and his other colleagues on stage made no mention of hardware systems, semiconductor, nor optical component plans, in sharp contrast to the year-ago HAS keynote made by last year's Rotating Chairman, Eric Xu. Instead Ken and his colleagues focused on the importance of software development (re-engineering software compliers to boost performance, for instance), the opportunity to work with the Huawei cloud service, entry to new markets (automobiles, the metaverse), and of course, the company's green initiatives. Mr. Hu also addressed supply chain challenges, saying that "it is true that we face a chip shortage," and that "we have no plans to build chip factories ourselves." Hu also said that the shortage will be "fixed in a few years." Hu also said that the company is organizing itself to allow for the shutdown of divisions or product areas more easily than could have been done in the past, which we think could mean it is planning for what might happen if Huawei cannot obtain chips, for instance, for its smartphone business. The last of Hu's prepared remarks was to remark that "Huawei has quite a few challenges ahead."
In the Telecom Core meetings that were held, there were several interesting points raised by the presenters. The team says that container based systems are not as mature as that of systems based on Open Stack and that Containers are not standards based. Regarding Mobile Edge Computing (MEC) market, Huawei said that its experience is that "almost all" of its customers are using MEC for network connectivity and though it had offered a marketplace for third party applications, they haven't been adopted much as of yet. Huawei sees Slicing as an emerging opportunity that will allow its operator customers to monetize their networks and to offer differentiated experiences.
Huawei spent a lot of time conveying that participation in the F5G (Fixed 5G) market is shared with other participants. Among those who shared the stage with Huawei were standards body ETSI, China Telecom, China Mobile, China Unicom, Globe Telecom (Phillipines), MTN Group (South Africa). Besides its product managers discussing F5G, Ken Hu also highlighted its importance during his aforementioned keynote.
Huawei's Radio Access Network (RAN) team highlighted a new 64T64R Massive MIMO system that has 384 antennae elements. Through multiple presentations, various managers highlighted how the new M-MIMO system enhances throughput and range, reduces power and other advantages of using its new product. The company also highlighted its Digital Indoor System (DIS) as a means of enabling enterprises across various vertical markets including manufacturing, coal mines, the steel industry, the petrochemical industry, shipping ports, healthcare, power-grid companies and others. Huawei said that in 2H22, it will bring to market a low-power chip-based system, as well. The company also highlighted a customer, Schneider Electric, who expects to enter its deployment phase of Huawei DIS systems later in 2022.
We found that Huawei's participating in the Fixed Wireless Access (FWA) market is interesting. Similar to Ericsson, Huawei isn't participating meaningfully in the Customer Premises Equipment (CPE) part of the FWA market and instead is selling just the RAN equipment. It has a certification program where it endorses other manufacturers, including greenpacket, who made a presentation at the Huawei conference. greenpacket explained that it is seeing a lot of demand lately for "hybrid" FWA CPE systems, which we took to mean that these were battery-based devices that users can travel around with. It is the experience of the presenters (both Huawei and greenpacket) that their customers in places like Kuwait and South Africa consume about 2-3 times more data using FWA than do average mobile (eMBB) customers. The presenters contrasted this with the T-Mobile experience in the US market where 10% of its FWA customers consume 1 TB per month. We found this contrast very interesting and feel that the screen sizes of customers in the US market (probably lots of high-end TVs) dramatically exceed that of the South African market (predominantly smart phones). In our own FWA report, we have seen significant ramp-ups in volume in the 3GPP FWA market.
51.2 Tbps Switch ASICs to Drive the Migration to 800 Gbps and 1.6 Tbps Starting in 2H22
NVIDIA continues to drive products and solutions at a rapid pace. They continued to push the boundaries and accelerate the pace of innovation in the data center through the recent GTC announcements. With the launch of Spectrum-4, we got to see NVIDIA’s first Ethernet Switch announcement and what they were busy developing since acquiring Mellanox. Keeping with the accelerating theme, NVIDIA clearly accelerated the pace of Innovation and proudly showcased Spectrum-4 at GTC.
Over the next five years, 51.2 Tbps ASICs will be responsible for over $20B in Data Center Switching revenue. About 2X the size of the 100 Gbps upgrade cycle and 4X the size of the 400 Gbps upgrade cycle. This class of ASIC is critical to allowing data centers to scale as more workloads become hardware accelerated. Hardware acceleration can come via the server, NiC, or DPU and causes networking bandwidth to double every year compared to the more traditional growth of 30-40%.
NVIDIA’s 51.2 Tbps ASIC
NVIDIA’s Spectrum-4 announcement included many new and first-to-market capabilities. It was the first 51.2 Tbps ASIC announced in the market and the first switch that can do 64 ports of 800 Gbps from a single ASIC. It also supports 1.6 Tbps speeds, another first and the preferred speed for the two most prominent Hyperscalers.
In addition, NVIDIA announced Spectrum-4 as a fully integrated Switch that will sample later in 2022 and begin shipping in 2023. The chip will be produced at TSMC in the 4N process. NVIDIA did over $1B in Ethernet Switch and NIC revenue in 2021.
It's All About The SERDES
NVIDIA will be using its own, homegrown, 112 Gbps SERDES. SERDES has always been a challenge in switch development, with many ASIC companies having to integrate 3rd party SERDES. However, we view vertically integrated as a source of differentiation and will become more common in the future as ASICs increase in speed. SERDES will also be a key to adding additional IP blocks and photonics to the networking ASIC, as those markets evolve.
The Speeds And Feeds Race
NVIDIA’s Spectrum-4 introduced several features that help it distinguish and route traffic. These traffic classes are often described as elephant and mouse flows. The key at a market level is that AI and accelerated computing put a considerable burden on the network. So AI networking switches need to keep pace with this new class of ‘elephant flow’ traffic and not slow or drop packets. At the same time, accelerated traffic is growing as a percent of traffic. Multiple 100G ports are already standard at the RFP level for AI workloads. 12.8 Tbps and 25.6 Tbps ASICs can not keep pace with a fully loaded AI cluster.
While the announcement focused on NVIDIA examples, such as connecting Spectrum-4 to NVIDIA Certified OVX OEM servers, we view the market as benefiting from NVIDIA taking networking seriously. Cloud customers have always requested multi-vendor at the ASIC and system level, and the ability to collapse networking tiers with 51.2 Tbps will be a critical driver to adoption. We, therefore, expect Spectrum-4 to apply to the whole data center, not just focused on the high-end or NVIDIA-connected solutions. We are excited to see this switch show up in 2H22 in the customer’s hands.
We attended Aruba, a Hewlett Packard Enterprise’s Las Vegas-based Atmosphere conference. It was very well attended, and this was Aruba’s first in-person user conference in over two years as we come out of the pandemic. Aruba made several important announcements at the event, including a formal Network as a Service (NaaS) offering, a series of broad cloud-services announcements, a formal unveiling of the Aruba/Pensando partnership, and the availability of industry first location-based access points.
The biggest splash came with Aruba’s NaaS announcement. While the company has been offering NaaS for two years, most of the deals were custom deals. In one of the smaller group meetings, the company shared a few of the logo wins, which included Texas A&M, Brookfield Properties, and Trevecca. What Aruba has done is standardized its offerings to a number of “service packs” that resellers can approach customers with. The company basically took its two years of experience and and packaged up what it felt were the best practices and the most common packages, which include wireless, switching and more. On a related topic, 650 Group announced that it is publishing a NaaS report, which of course includes historical and forecast data for this up and coming market. Additionally, in the small group breakout meeting, some resellers were asking questions about what happens at the end of a three or five year subscription and whether the offering can be split up or must it be sold as one. We feel that Aruba will be working closely with their partners over the next year or so to support them as NaaS gains stickiness, but that there is some genuine interest as markets move away from pure CAPEX models towards subscription and as-a-Service.
The company also announced significant enhancements to its Aruba Central Cloud-Managed Services offerings, with Aruba Central NetConductor. NetConductor allows for cloud-based central management of wired, wireless and WAN systems as well as to SASE systems. It also integrates the capabilities of two other functions well known to Aruba customers, Network Access Control (NAC) and Dynamic Segmentation. Aruba has based NetConductor on some widely used protocols like EVPN, VXLAN and BGP which would allow for integration with both Aruba gear and non-Aruba gear.
During the first day’s keynote sessions, HPE CEO, Antonio Neri, spent quite a bit of time on stage. Former Cisco CEO John Chambers joined him on stage so that they could discuss the HPE and Pensando’s partnership. It was interesting to see Chambers literally hugging Neri, especially considering that Chambers ran Aruba’s arch-nemesis, Cisco Systems, for two decades. It made the keynote very interesting. We covered the Aruba-Pensando announcement last fall and you can find more about it here. Another thing that is interesting is that Neri’s presence was felt beyond the Chambers discussion, as Aruba Central is now part of HPE GreenLake. HPE claims that this gives IT admins a single operating model for network, compute and storage services across edges, data centers, and public clouds.
Aruba announced that its newer access points have all been shipping with a location-based feature that leverages GPS. This is a differentiator for Aruba because it said it is not charging extra hardware or subscriptions charges to use GPS. We’ve been thinking about what this function can do for Aruba customers. Relative location capabilities have been around on WLAN Access Points and on Bluetooth beacons for several years now and many vendors offer it. However, the way we see it is that having a GPS capability makes location capabilities far easier to use because what Aruba’s location capabilities do is position all the APs on an absolute basis, and thereafter, a relative service like wayfinding, dwell time analytics and the like can be anchored to specific places on a floor plan without having to get the ruler out, so to speak. Also, maintaining control over inventory can be simplified because these Aruba APs know where they are, like, what ceiling tile they can be found in, or what building they were moved to.
In summary, at the Atmosphere show, Network as a Service has become a major talking point in the industry, Aruba’s cloud service now shares the stage with other top players in the industry, and Aruba announced a really cool function on its access points that simplifies location services.