{"id":9411,"date":"2017-05-15T07:28:59","date_gmt":"2017-05-15T07:28:59","guid":{"rendered":"https:\/\/www.techdesignforums.com\/practice\/?p=9411"},"modified":"2017-05-17T00:46:32","modified_gmt":"2017-05-17T00:46:32","slug":"the-wally-rhines-interview-part-two-ai-automotive-security","status":"publish","type":"post","link":"https:\/\/www.techdesignforums.com\/practice\/technique\/the-wally-rhines-interview-part-two-ai-automotive-security\/","title":{"rendered":"The Wally Rhines interview – Part Two: AI, automotive and security"},"content":{"rendered":"

Part One of our interview with Mentor chairman and CEO Wally Rhines looked at the mechanics of the company\u2019s acquisition by Siemens. You can read that here<\/a>. In Part Two, Rhines discusses more detail of the deal, particularly regarding R&D and EDA as part of a Siemens flow. He also addresses Mentor\u2019s launch of the DRS360 automotive platform and the emerging challenge of design-for-security.<\/em><\/p>\n

Pick and mix<\/h3>\n

There is always concern about monolithic design flows. Customers shy from putting all their eggs in one basket. Can a single-company flow ever offer the leading solution at every stage? With Mentor\u2019s acquisition by Siemens<\/a>, there is the prospect of one company now delivering essentially\u00a0all the components needed for product delivery\u00a0from initial concept to the factory gate.<\/p>\n

For Wally Rhines though, that simply is not likely to happen.<\/p>\n

\u201cIf you want front-to-back, we can do that. It won\u2019t all be best-in-class though as long as you\u2019re not pushing the technology, it will work pretty well,\u201d says Rhines.<\/p>\n

\u201cBut I have never found anyone who used a single company\u2019s flow front-to-back in EDA. You want best-in-class. To do that everyone tends to have something from here and something from there. Take Calibre<\/a> [Mentor\u2019s market-leading DFM, physical verification and signoff suite]. A customer wants to look at Calibre integration into a Synopsys-led or a Cadence [Design Systems]-led flow. We don\u2019t care. We think we\u2019re better than the equivalents. You want best-in-class. We\u2019ll supply.\u201d<\/p>\n

Rhines says that Siemens shares this philosophy of selling into \u201copen environments\u201d, supplying the connections to other companies\u2019 software where necessary.<\/p>\n

\u201cYou want just Teamcenter<\/a> or NX<\/a> within your PLM flow. Fine. That\u2019s OK. Things will continue to be like that,\u201d says Rhines.<\/p>\n

Wally Rhines talks collaboration on AI<\/h3>\n

That does not rule out software integration and collaboration on R&D across Siemens group companies. As Rhines noted in Part One, some areas of EDA research are specific in their needs so Mentor will chart its own path, but there are areas with more common ground to exploit. One is the combination of machine learning and AI.<\/p>\n

\u201cThis is an area where you can see there\u2019s going to be a lot of leverage early on. While Siemens has already gone deep into the area, we also bring something to contribute,\u201d says Rhines.<\/p>\n

He cites the specific example of Valor<\/a>, the Mentor subsidiary focused on manufacturing automation for PCBs.<\/p>\n

\u201cWe\u2019re among the leading suppliers in that space \u2013 we\u2019re working with Foxconn, we\u2019re working with Flextronics \u2013 and we have a surprisingly deep infrastructure. We\u2019ve 25 people [in Valor] of which 14 are PhDs working on algorithms for components sequencing on tape-and-reel to optimize the arm travel of the pick-and-place machines. We have gateways at the factory floor transmitting data into the cloud and doing deep learning and optimization on the product flow. We have traceability of every IC in every board for failure analysis, for checks against counterfeits, and more,\u201d says Rhines.<\/p>\n

\u201cWhen it came to due diligence, that really impressed the Siemens people. They did not expect that level of technology within Mentor.\u201d<\/p>\n

\"Figure<\/a>

Valor uses AI and machine learning for full part traceability (Mentor – A Siemens Business)<\/p><\/div>\n

Mentor is also flexing its muscles in machine learning in the automotive market. It has launched the DRS360 sensor platform<\/a> for Level 5, fully autonomous vehicles. This uses AI to enable centralized processing of raw data from various sensors to provide real-time responses to safety-critical situations, while reducing power consumption and latency. No more MCUs alongside each sensor for a start.<\/p>\n

The launch caught some observers cold. But as Rhines explains, it also came from expertise already within Mentor, particularly from the 2014 acquisition of XS Embedded (XSe), an automotive embedded software company initially working on infotainment systems.<\/p>\n

\u201cXSe were very successful and very good at what they did \u2013 they came out of Harman [International]. So, we started doing a lot of ADAS [advanced driver-assistance systems] as well. They allowed us to look back at infotainment and see where some problems with ADAS might lie as well,\u201d Rhines says.<\/p>\n

\u201cInfotainment had quickly threatened to become a kludge. You had GPS from here, backscreen TV from there, surround-sound from somewhere else. It was piecemeal, and as we kept adding things, you could see that at a certain point it would become unsustainable to support the software with all these different pieces from different places.<\/p>\n

\u201cWhat the XSe people had realized in infotainment was that you needed to create a single embedded software environment to integrate all this stuff. They developed it and sold it, \u20ac100 per car.\u201d<\/p>\n

That team saw ADAS headed in the same kludgy direction.<\/p>\n

\u201cSo we said, \u2018Stop.\u2019 We halted almost all the ADAS development, took our A-team and said, \u2018Let\u2019s first get the world\u2019s best driverless car architecture and then you can develop ADAS pieces that fit into it.\u2019 That\u2019s how we got to DRS360.\u201d<\/p>\n

\"Figure<\/a>

Wally Rhines notes that DRS360 came out of established Mentor technology (Mentor)<\/p><\/div>\n

A few commentators argued that the launch could bring Mentor into competition with some of its EDA customers. Wally Rhines is phlegmatic. Take, for example, Mentor\u2019s relationship with NXP, the leading automotive silicon supplier.<\/p>\n

\u201cI talk to [NXP president and CEO] Rick Clemmer every other week. He\u2019s my old controller from TI. And [NXP board director] Peter Bonfield is also on our board. So there\u2019s that,\u201d says Rhines.<\/p>\n

\u201cBut this instead encourages all the chip companies to work with us. From our history in automotive, we know that the carmakers pick the embedded software first, then they choose the tier ones and the ICs. In this market, we\u2019re at the front end. For the IC companies, we make their chips look good. So it\u2019s much more about collaboration.\u201d<\/p>\n

The trouble with Trojans<\/h3>\n

When we talked, Wally Rhines was about to deliver a lecture at New York University on design-for-security. As that and his DVCon China keynote show, he sees the topic as one that is reshaping the design cycle, and not just for silicon.<\/p>\n

One element he sees becoming more prominent is the threat posed by embedded Trojans and how companies address their potential presence in silicon. It is an extension of a frequent Rhines observation: Verification is moving from simply verifying that a chip does what it is supposed to do to also verifying that it does not do anything it is not supposed to do.<\/p>\n

For this particular threat \u2013 the best known example being the Stuxnet<\/a> virus used against Iranian nuclear centrifuges \u2013 his immediate concern is that industry is not discussing the problem.<\/p>\n

\"Figure<\/a>

The Stuxnet attack on Iranian centrifuges put embedded Trojans in the spotlight (Islamic Republic of Iran)<\/p><\/div>\n

\u201cI\u2019ve gone to meetings and heard that [embedded Trojans] don\u2019t seem to be a problem. If they are around, our customers who buy design tools don\u2019t seem particularly troubled about it. But then if I tell this to guys at the NSA, they roll their eyes like I came out of La-La-Land,\u201d says Rhines.<\/p>\n

\u201cSo you have to figure that this is happening and that it\u2019s being done by all the players. Why are there restrictions on some imported equipment? Well, you have to guess that it\u2019s partly \u2013 at least \u2013 because you can embed Trojans in them.\u201d<\/p>\n

The problem as it stands takes on two initial forms. The first is cost. The second is indemnification.<\/p>\n

\u201cYou can put things into your chip like PUFs [physically unclonable functions] that address the threat. But they cost money. They add to complexity. So, people don\u2019t want to go through that if they don\u2019t have to,\u201d says Rhines. \u201cBut one of these days, something is going to happen and everybody will\u00a0go, \u2018OK, where\u2019s the toolkit?\u2019\u201d<\/p>\n

But once you offer a toolkit, Rhines notes that something else becomes inevitable: \u201cAt that point, the system builder will go\u00a0to the chip supplier and say, \u2018I\u2019d like to add this sentence to the purchase order. It says that there are no Trojans built into your chips.\u2019<\/p>\n

\u201cYou\u2019ll go to your lawyers and ask if that\u2019s OK, and you know what they\u2019ll say: \u2018Are you kidding<\/em>? Absolutely not. There\u2019s no way in the world we\u2019re doing that.\u2019\u201d<\/p>\n

There are things you can do, Rhines notes. DARPA, the US defense research agency, is looking to promote one scheme \u2013 \u201cBut again, it\u2019s a few cents off your margin.\u201d \u2013 and once you know what a virus looks like you can scan for it (but then there will always be new ones).<\/p>\n

\u201cThe way forward \u2013 and it\u2019s not really happening yet to scale \u2013 is that we’ll have to agree across the industry on a set of best practices. Maybe DARPA takes us in that direction,\u201d says Rhines. \u201cBut what you will have to\u00a0offer ultimately, I think, will be an ability to say, ‘We followed these agreed best practices. We can\u2019t say that there\u2019s nothing in there for sure, but we did use these best practices to prevent it.’ For even that, though, we need the negotiation to start.\u201d<\/p>\n

Wally Rhines has been his own boss for 23 years, notwithstanding that he has always had a boardroom to answer to and the occasional attentions of a Carl Icahn. He is still a big wheel but now one within a huge conglomerate. How is he finding the change?<\/p>\n

\u201cIt\u2019s different. Yes, it\u2019s rigorous. But in many ways, it doesn\u2019t change. On the system design side, we\u2019re all on the same page. We\u2019ve been there for years,\u201d he says. \u201cBut you do have to adapt.<\/p>\n

\u201cOne really good thing is there in those [Siemens] guidelines for integration. The first one is, \u2018The customer comes first.\u2019 So, there you go.\u201d<\/p>\n","protected":false},"excerpt":{"rendered":"

This second part looks at Mentor’s views on flow neutrality, how DRS360 was born, machine learning and the threat from embedded Trojans.<\/p>\n","protected":false},"author":138,"featured_media":8463,"comment_status":"open","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[934,1237,1162],"tags":[2045,707,1479,1180,2047,2046,1827,2016,1856,1826,369,2035,1214],"coauthors":[1441],"class_list":["post-9411","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-eda-topics","category-embedded-topics","category-pcb-topics","tag-ai","tag-automotive","tag-calibre","tag-design-security","tag-drs360","tag-embedded-trojans","tag-ma","tag-machine-learning","tag-pcb","tag-rd","tag-security","tag-system-design","tag-valor","workflow-expert-blog","workflow-interview","workflow-up-to-date","organization-mentor-graphics","organization-nxp-semiconductors-chipmaker","organization-siemens","organization-mentor"],"_links":{"self":[{"href":"https:\/\/www.techdesignforums.com\/practice\/wp-json\/wp\/v2\/posts\/9411","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.techdesignforums.com\/practice\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.techdesignforums.com\/practice\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.techdesignforums.com\/practice\/wp-json\/wp\/v2\/users\/138"}],"replies":[{"embeddable":true,"href":"https:\/\/www.techdesignforums.com\/practice\/wp-json\/wp\/v2\/comments?post=9411"}],"version-history":[{"count":0,"href":"https:\/\/www.techdesignforums.com\/practice\/wp-json\/wp\/v2\/posts\/9411\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/www.techdesignforums.com\/practice\/wp-json\/wp\/v2\/media\/8463"}],"wp:attachment":[{"href":"https:\/\/www.techdesignforums.com\/practice\/wp-json\/wp\/v2\/media?parent=9411"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.techdesignforums.com\/practice\/wp-json\/wp\/v2\/categories?post=9411"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.techdesignforums.com\/practice\/wp-json\/wp\/v2\/tags?post=9411"},{"taxonomy":"author","embeddable":true,"href":"https:\/\/www.techdesignforums.com\/practice\/wp-json\/wp\/v2\/coauthors?post=9411"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}