We attended the Open19 Foundation Summit 2019 in San Jose, CA. The key message of the leadership team of the Foundation was that it expects 2019 to be the year where Open19 systems begin shipping. Yuval Bachar, President of the Foundation explained that there are 8-10 companies that are deploying or seriously considering, including two mega data center operators in process of evaluation and at least 6 adopters in advanced evaluation/deployment. Open19 does not specify what technology will be inside the systems; it only defines the form factor. Interestingly, GPU and ARM designs are included and we heard about AMD CPU more often than we did Intel.
VaporIO made a presentation about its edge data center systems. Matt Trifiro, CMO of VaporIO explained that its systems nominally consume up to 165 kw in a 9 foot round datacenter. It claims that Vapor IO is deploying in 6 cities, 3 to 4 data centers in each city.
ASUS and a new company, German Edge Cloud, both made interesting presentations. Salim Fedel, Associate VP Enterprise Solutions Business Group, ASUS, presented a few of the company's Open19 oriented products, Alps19, Brick and its Network Switch. Jason Rylands, VP of Data Center Strategy & International Sales, German Edge Cloud, made a very interesting presentation about its company's activities in Germany on how it is participating in the manufacturing industry in Germany. There is reticence on putting manufacturing data and control on the cloud, but there is a desire to build a German-only computing system, driven in part by GDPR and part by the nation's focus on manufacturing - that's the opportunity. German Edge Cloud has decided to use Open19 systems - accessing only the front of the racks is a big deal, it believes.
German Edge Cloud shared many examples of how it is engaging with manufacturers for real-time error tracking during the manufacturing process, using analytics, and sharing information between various companies in the German supply chain.
We attended media briefings at the Keysight World conference held near San Francisco, California, yesterday. Keysight World San Francisco showed a more integrated Ixia as well as a further push towards 5G, automotive and energy, and data center and telecom. Much of the companies conversations during the event related to synergies between these technologies. Such as how to push the automotive industry forward using 5G and the need for edge computing in 5G. As Keysight is involved at the beginning of the launches of many new cellular technologies, it has good insights into the timeline and progress of the 5G market.
One of our key takeaways from the show is the increased pervasiveness of Ethernet in the telecom and automotive industry as well as the significant increase in constant data creation. All this data will need robust, and more intelligent networks to transport it. We were also impressed that Keysight World will actually take place in 10 cities around the world, with San Francisco being the first.
Another interesting observation was just how global Keysight was in looking at products, for example in one session that we attended, much of the automotive effort in the company was located in Germany.
According to news reports and press and social media announcements by high-ranking members of US government, the US government has put Huawei on its so-called "Entity List" of the Bureau of Industry and Security (BIS). Our read on this is it similar to what happened with ZTE during C2Q18 last year, a move that severely curtailed ZTE's shipments and revenue until ZTE made concessions and was removed from the list. Many, but not all, Huawei products use technology only available from US suppliers. US-made semiconductors are the most significant Entity List target that Huawei needs to ship its products. Significant US semiconductor suppliers to Huawei include Intel, Xilinx, and Broadcom.
Huawei is such a significant vendor in many of our coverage areas, including Mobile Radio Access Networks (RAN), Ethernet Switching, and Servers, for instance, that we feel it is a good time to point out that 2019 market-level estimates may be at risk. Additionally, since Chinese cloud services players, like Baidu, Alibaba and Tencent cannot delay their capital infrastructure build-outs, alternate suppliers may benefit.
We think it comes as no surprise to Huawei that the US is putting it under pressure. Just over a year ago, we attended the Huawei analyst summit (April 16, 2018) and its then-chairman said in response to the question "Will Huawei find alternate suppliers for data center products, "Today, Intel is the dominant player. Our point of view, we look forward to a more diversified landscape; but we work with Intel mainly now." Additionally, at Huawei's most recent analyst summit (mid-April 2019), the three main keynote speakers, all high-ranking executives of the company spoke about how much progress Huawei has made in developing in-house semiconductors and what its plans are to continue developing more. We do, however, think that despite Huawei's diversification efforts that it still has significant reliance upon key US chip companies.
The main message of Ribbon keynotes was that Ribbon has moved from a voice and SBC focus to a broader scope including data and analytics, such as UCaaS, Security-as-a-Service, Service Assurance-as-a-Service, SIP Trunking-as-a-Service, SD-WAN, and analytics. The company identified two major catalysts to the change in company scope: the August 2018 acquisition of enterprise SBC company Edgewater and the February 2019 acquisition of analytics company, anova. Additionally, the company confirmed that it is not pursuing service provider mobile market opportunities such as EPC or “full IMS” that are components of larget mobile equipment larger vendors. Instead, Ribbon is pursuing the mobile market by offering overlay analytics services that allow SP customers to derive value from mobile operators using monetization, network, marketing, and customer care.
The company’s core business is evolving. For instance, the company has doubled down on the Enterprise market with its Edgewater acquisition, and a result, it emphasized that it is unique because it has both service provider and enterprise SBC product offerings. The company has offered SBC on public cloud and reiterated that this capability has been available for one year. The company highlighted that it has a Tier-1 Mobile Network Operator VoLTE Interconnect contract win.
Analytics. The company sees its opportunity in the analytics market is to deliver technology to the customer for the following cases:
• Monetization (Targeting, Advertising, Sponsored Data, Campaign Efficacy)
• Network (Reduce Network Cost, Improved Quality of Experience, Proactive Alarms, Network Service Assurance)
• Marketing (Service bundles, churn reduction, product insights, inferred demographics)
• Customer Care (customer experience, bill shock, most probably cause)
We are impressed with Ribbon’s technical capabilities when it comes to using GPUs. The company has previously discussed the performance of its GPU based media interworking function as being 3.5x more powerful than a DSP-based system, or 9x more powerful than a CPU-based system and over 2x more power efficient. The company says its GPU-based systems are generally available for its D-SBC and i-SBC functions. It has evaluations at three operators: Tier-1 US MNO, Tier-1 US CSP, and Tier-1 Japan CSP. We would not be surprised if some of these customers begin deploying soon. One of its Tier-1 operators found that the GPU-based system costs half as much in capital spending and is 800% more power efficient.
Kandy, the white-label voice/messaging services brand of Ribbon discussed its success with customers such as BT, NUWAVE, AT&T, Hong Kong Broadband, Optus, and ecosystem partner, Five9. It enables UCaaS, CPaaS and WebRTC services. The company reiterated that it plans to go to market with partners, mainly service providers, instead of opening its own store-front that would compete with service providers. The company says its customer pipeline is growing.
Huawei hosted 700 analysts and media participants in Shenzhen China last week to attend its annual analyst summit, nick-named HAS2019. The company's high-level message was simple - the company is an innovator and is moving down the stack into semiconductors and is partnering with and funding university projects to develop basic research. This year’s message was different from than the prior-year meeting, but several transformative events have occurred between this meeting and the prior year's, most notably the 2Q18 shipment ban on ZTE, the US / China trade dispute and US efforts to thwart Huawei’s participation in the 5G infrastructure of its allies. Interestingly, during HAS2019, the Apple and Qualcomm announced their chip-supply and patent settlement, Samsung announced its foldable phone (which has been met with criticism), and Ericsson & Swisscom announced that the operator went live with its 5G network. All three of non-Huawei events highlighted the importance of Huawei’s chips and innovation announcements.
The company made announcements in its main keynote presentations on day one about seven different chip projects delivered recently or planned shortly. Chip-level is unusual for what are typically high-level presentations from a keynote-level presentation. These chips (seen in accompanying pictures) are:
The company shared more details about other chips in breakout sessions on the second and third days of the conference, as well. The point we are making, though, is that upper-level management provided significant detail about semiconductor developments at Huawei. Another relevant semiconductor-related point to make is that the company is de-emphasizing its reliance on Intel-based architecture and instead is focusing on devices such as ARM-based processors, as well as GPU, FPGA and NPU semiconductors.
We would be remiss if we did not mention some of the system-level announcements and observations related to 5G that were made at the HAS2019 conference, which include:
even more follow on
We attended the Aruba Atmosphere 2019 user conference in Las Vegas. What we learned was that Aruba had made solid progress since last year’s Atmosphere conference. It has delivered on 802.11ax, SD-Branch (and SD-WAN), AI/ML, and Zigbee/Bluetooth 5.0, and elements of the IoT market. The company also introduced a new access point that was not hinted at last year, an 802.11ad outdoor access point. If we were to sum up the company’s main message for the show, it’s all about SD-Branch. The company took great efforts to emphasize that its portfolio has greater breadth than ever and is among the few vendors that can deliver all the networking a company may desire.
802.11ax. At last year’s event, the company told customers to expect 802.11ax products by Nov/Dec of 2018. Our market share tables show the company shipped 802.11ax for revenue in 4Q18. At the show, the company also announced some new, full-featured 802.11ax Access Points, the 530 and 550 series. These new Access Points support Bluetooth 5 and Zigbee, to allow support of IoT devices. These new 802.11ax Access Points will be available this month, April 2019.
802.11ad. The company also introduced a point to point outdoor access point. The new AP387 allows 1 Gbps at 400 meters using 802.11ad and has a backup of 5 GHz 802.11ac in case of inclement weather. This device has been shipping for a “couple months” according to the stage presentation (personnel at the show booth said since January 2019).
Machine Learning. Using Machine Learning for Client Steering and for managing Transient clients. At last year’s Atmosphere event, the company was just rolling out AI/ML to customers to improve networking capabilities for wireless users.
SD-Branch. The company disclosed that it has 25,000 SD-Branch “wins,” which means that it has many contracts to sell “at some point in the future” SD-WAN and other branch equipment systems such as WLAN and switching. At last year’s event, the company had not sold any SD-WAN, so this is a big accomplishment and signifies Aruba’s progress in delivering what it calls a Single Pane of Glass approach that includes four parts: SD-WAN orchestration, Dynamic Path Steering, Secure Connectivity, and Dynamic Segmentation.
Clearpass Device Insight. The company introduced its device recognition system, intended to simplify the discovery of IoT devices on the network. Clearpass Device Insight is available in April 2019. This cloud service uses a fingerprint database, as well as AI/ML, to find devices on the network, and then presents them by category on a single screen.
During the day-2 presentations, the company had some fun and CTO, Partha Narasimhan, showed a picture of him pretending to be an IT executive of a fictitious university.
Today Intel had a major Data Center event in San Francisco. It was a multi-hour announcement showcasing all the different products Intel is/will be launching.
Some interesting background that Intel talked about was that only 2% of worlds data has been analyzed and a 5G will be a major driver move compute to the edge. They also touted that over 50% of AI workloads are inference and runs best on Intel (X86). Noticeably absent at the beginning of the presentation was Intel’s work on training.
We found the most interesting parts of the announcement being that AWS talked about custom versions of Intel’s CPU and a Up to 14X inference improvement from just July 2017 in its XEON processors. Overall there is a 30% gen/gen improvement in XEON, the biggest jump in 5 years. While staying at 10NM, Intel is able to continue to squeeze performance gains out of the server.
It is important as we hit the limits of process geometry that everything be accelerated, especially with Intel having a similar view of AI workloads and that’s what we saw with Intel. New Optane memory, new persistent memory, faster Adapter cards which will lead to more Smart NIC announcements, and a 10nm FPGA. It was clear at the event that large cloud provers like AWS, Azure, and Tencent are looking at all avenues to increase performance and reduce power consumption via software and hardware advancements.
Some interesting highlights included AWS touting over 100 unique instances that leverage Intel processers with more SAP instances running on AWS than anywhere else and Formula 1 using of 65 years of historic race data to train its models in order to make real-time race predictions.
Cisco is in the 802.11ax market now, with two sets product products under each of its main brands. These are: Cisco Meraki MR45/55 and Cisco Catalyst 9100 Access Points. You cannot find these kinds of products on Amazon, but some are listed on CDW for instance.
The Meraki products don't have the full set of 802.11ax features, and the Catalyst 9100 APs are categorized as "certifiable" and "pre-standard" depending on what models you chose. For our market share purposes, we'll be calling all of these products as "802.11ax" (aka Wi-Fi 6) in our market share tables, so it looks like Cisco will have shipped some product in the C1Q19 quarter. This is interesting because it is common for Cisco to introduce new enterprise products at its Cisco Live show, and that's on Jun 9-13, 2019. So, we were thinking we'd find out about the products in San Diego at Cisco Live. It's possible that Cisco felt some competitive pressure to get the products out there; many of its competitors have already been shipping products (HPE Aruba, Ruckus, H3C are some examples).
We'll be reporting on C1Q19 shares in a couple months and will have more to say about how many Cisco 802.11ax products actually shipped, but we are fairly confident the 802.11ax enterprise-class product class is hitting the knee of the curve, so to speak, very soon.
We witnessed two significant OCP Summit announcements:
a) open sourcing of semiconductor design
b) improved server system security.
First, Microsoft announced a new compression standard called Project Zipline. The intellectual property for this compression is being made available as open source Verilog (RTL), which will allow others to program semiconductors like FGGAs. The claim is that Project Zipline outperforms existing compression by more than 20%. This is the first semiconductor design member contribution.
Second, and continuing a theme started at last year’s OCP show, Microsoft announced improvements to Project Cerebus, which is a systems level design and specification to improve server security. Last year, the project addressed boot time security for the CPU. This year’s announcement addresses other subsystems associated with servers, such as accelerators, storage and NICs. We see this project as a reaction to recent concerns of data center hardware compromises that came to light in the public eye over the last couple years.
Juniper made a significant announcement at OFC on the photonics front. It announced the launch of silicon photonics optics designed and built by Juniper. This is a big deal for the industry and for data center networking.
To give some market background, as the market moves from 100 Gbps to 400 Gbps, and then beyond, the cost of the optic is becoming a larger part of the spend. As hyperscalers tend to buy optics separately from the switch and from different vendors, this is lost Total Addressable Market (TAM) for the networking vendors if they don’t adjust product offerings. Vendors risk losing wallet share and thought leadership due to the transition.
Juniper is building their own photonics optics, and growing its own silicon to create a product that is compelling to cloud providers. Engineering like this is not only hard, but cool and pushing the upper limits of science. As we look at data center fabrics beyond 2020, this type of technology is needed to get speeds beyond 25.6 Tbps in a single 1RU switch.
We note that by targeting just large Cloud customers, Juniper is actually targeting nearly 90% of the market, so while the announcement may seem limited in customer count, it comprises almost the entire revenue and shipment opportunity for 400 Gbps in the next two years. By engineering the solution in house, Juniper will be able to offer better power and better cost. Power should not be underestimated as data center locations are struggling to get additional power.
At a higher level, over 25% of the Ethernet Switch market in the data center is actually optics. That percent is growing with time. Juniper’s addressable market has gone up significantly by offering optics which is a good thing for the data center market and its evolution to higher speeds.