Source Code: Your daily look at what matters in tech.
To give you the best possible experience, this site uses cookies. If you continue browsing. you accept our use of cookies. You can review our privacy policy to find out more about the cookies we use.
enterprise| enterpriseauthorTom KrazitNoneAre you keeping up with the latest cloud developments? Get Tom Krazit and Joe Williams' newsletter every Monday and Thursday.d3d5b92349
November 6, 2020
Reality sank in this week for software developers who built applications using Docker containers over the past few years: The free ride is over.
Earlier this year, Docker announced that it was going to impose new limits on how free users of its Docker Hub service would be able to access public container images stored in that repository, and those changes started rolling out Monday. The move comes almost exactly a year after Docker sold the enterprise version of its software business and 75% of its employees to Mirantis, leaving a much smaller company behind.
Developers had time to prepare, but many — especially users of the Kubernetes container management service — were still caught off guard this week by the new limits. In response, several vendors offered tips for getting around the issues involved, and AWS announced a plan to offer its own public container registry in the near future.
Containers were a huge step forward for modern software development. They allow developers to package all the building blocks needed to run an application into, well, a container, which can be deployed across a wide variety of cloud or self-managed servers. Docker raised more than $300 million in funding after it created a developer-friendly way to use containers in the mid-2010s, but despite wide use of its container format, the company has struggled to find a business model.
Container images are essentially blueprints for the container, and they are usually stored somewhere readily accessible for developers to grab when they are updating their applications with new code. Developers can store their images in private if they prefer, but over the past few years lots of developers and companies opted to publish public container images to boost adoption and awareness of their software products. The convenience of those building blocks of code being publicly available meant that many people built applications using them.
Docker builds its own certified images for Docker Hub users to employ, along with certified images published by third-party developers and a trove of community-generated images. When it was a fast-growing enterprise tech unicorn, Docker offered those services for free, but the company can't afford such largess at this point in its history.
That entire repository is pretty big — over 15 petabytes worth of container images — and storing that much data is not cheap. Earlier this year Docker said it would delay a plan to delete inactive images after a community uproar, but as of Nov. 2 it imposed new limits on how many times free users of Docker Hub could grab images over a six-hour period, given that the bandwidth costs associated with serving those images are also not cheap.
The rise of automated continuous integration services provided by companies like CircleCI and JFrog exacerbated the problem, said Donnie Berkholz, vice president of products for Docker. Those services automatically check container images for updates when deploying changes to software, which is great for their users but a load on Docker.
"On the order of 30% of our traffic was coming from 1% of our users, and that's not sustainable when those users are free," Berkholz said.
Users of Docker's paid services — which also include features designed for teams and large software organizations — will not face the rate limits, and Docker worked out a deal that will lift the limits for most of CircleCI's customers, too.
Deep-pocketed cloud providers see a different opportunity. Microsoft's GitHub announced plans for its own free public container registry in September, and on Monday AWS announced vague plans for a public container registry that it will likely outline during its upcoming re:Invent virtual conference.
The storage and bandwidth costs associated with hosting container images are rounding errors for companies such as Microsoft and AWS, and developer goodwill is a valuable commodity. AWS will likely encourage users of its public container service to run those containers on AWS, and while GitHub still operates at an arm's length from Microsoft, similar suggestions for Azure users wouldn't be surprising.
In the end, Docker's move is a signal that a relatively permissive and free-wheeling era of cloud computing is winding down as it becomes an enormous business. It also highlights the importance of the software supply chain: Modern software applications pull code from a wide variety of places, and disruptions to those supply chains can have profound effects on application performance or availability.
From Your Site Articles
- How Google Cloud's Kelsey Hightower became one of the most ... ›
- Why SUSE just shelled out millions for Rancher Labs - Protocol ›
- VMware CEO Pat Gelsinger on Kubernetes hype, controlling open ... ›
- Software ate the world. Now it’s eating software companies. - Protocol ›
- How Google Cloud's Kelsey Hightower became one of the most respected people in cloud computing - Protocol ›
- VMware CEO Pat Gelsinger on Kubernetes hype, controlling open-source and freeing your CPUs - Protocol ›
Tom Krazit
Tom Krazit ( @tomkrazit) is a senior reporter at Protocol, covering cloud computing and enterprise technology out of the Pacific Northwest. He has written and edited stories about the technology industry for almost two decades for publications such as IDG, CNET, paidContent, and GeekWire. He served as executive editor of Gigaom and Structure, and most recently produced a leading cloud computing newsletter called Mostly Cloudy.
Big Tech benefits from Biden’s sweeping immigration actions
Tim Cook and Sundar Pichai praised President Biden's immigration actions, which read like a tech industry wishlist.
Newly-inaugurated President Joe Biden signed two immigration-related executive orders on Wednesday.
Photo: Chip Somodevilla/Getty Images
January 20, 2021
Emily Birnbaum ( @birnbaum_e) is a tech policy reporter with Protocol. Her coverage focuses on the U.S. government's attempts to regulate one of the most powerful industries in the world, with a focus on antitrust, privacy and politics. Previously, she worked as a tech policy reporter with The Hill after spending several months as a breaking news reporter. She is a Bethesda, Maryland native and proud Kenyon College alumna.
January 20, 2021
Immediately after being sworn in as president Wednesday, Joe Biden signed two pro-immigration executive orders and delivered an immigration bill to Congress that reads like a tech industry wishlist. The move drew enthusiastic praise from tech leaders, including Apple CEO Tim Cook and Alphabet CEO Sundar Pichai.
President Biden nullified several of former-President Trump's most hawkish immigration policies. His executive orders reversed the so-called "Muslim ban" and instructed the attorney general and the secretary of Homeland Security to preserve the Deferred Action for Childhood Arrivals, or DACA, program, which the Trump administration had sought to end. He also sent an expansive immigration reform bill to Congress that would provide a pathway to citizenship for undocumented individuals and make it easier for foreign U.S. graduates with STEM degrees to stay in the United States, among other provisions.
<p>Biden's Day One moves suggest that immigration reform will be one of the key areas for allyship between Silicon Valley and the Biden administration over the next several years. The industry is sorely in need of common ground as Democrats pledge to use their power over Congress and the White House to crack down on Big Tech. </p><p><div class="ad-tag"><div class="ad-place-holder" data-pos="1">
</div>
</div></p><p>"We welcome President Biden's commitment to pursuing comprehensive immigration reform that reflects the American values of justice, fairness and dignity," said Cook in a statement to Protocol. "This effort will strengthen American communities and the pathways to opportunity this country has long fostered." </p><p>Cook, who is the chair of Business Roundtable's immigration committee, said he looked forward to working with Democrats and Republicans on a "permanent solution for Dreamers that includes a path to citizenship."</p><p>"We applaud @POTUS's quick action on COVID relief, the Paris Climate Accord, and immigration reform," Pichai <a href="https://twitter.com/sundarpichai/status/1352021111479832576?s=21" rel="noopener noreferrer" target="_blank">tweeted</a> Wednesday. "Google has supported action on these important issues & we look forward to working with the new administration to help the US recover from the pandemic + grow our economy."</p><p>An Uber spokesperson also affirmed the company's support for President Biden's actions. "At Uber, we have long supported positive immigration reform and an effective system that upholds core American values," the spokesperson said. "Uber will continue to support Dreamers, and we welcome the new Administration's effort to reform our nation's immigration system."</p><p>The tech industry, which relies heavily on foreign workers, has spent the last four years <a href="https://www.usatoday.com/story/tech/2017/08/31/tech-industry-prepares-new-trump-fight-over-daca-dreamer-protection-program/622897001/" rel="noopener noreferrer" target="_blank">battling with the Trump administration</a> over efforts to shut down vital talent pipelines from countries including India, China and Iran. Over the last four years, tech giants have filed <a href="https://abcnews.go.com/Business/dreamers-embody-apples-innovation-strategy-apple-ceo-tim/story?id=66030999" rel="noopener noreferrer" target="_blank">amicus briefs</a> in high-profile <a href="https://www.vox.com/2020/1/16/21069827/twitter-microsoft-linkedin-public-charge-trump-immigration-tech-companies" rel="noopener noreferrer" target="_blank">court cases</a> and lobbied in favor of opening up U.S. borders to high-skilled workers. Despite protestations from Silicon Valley leaders, the Trump administration continually added uncertainty to the H-1B visa application process, resulting in significant backlogs. </p><p><div class="ad-tag"><div class="ad-place-holder" data-pos="2">
</div>
</div></p><p>Biden's immigration bill would put money towards clearing visa backlogs, eliminate visa caps on certain countries and make it easier for people on temporary work visas to get green cards. Perhaps most significantly, it would immediately give the 11 million undocumented immigrants in the U.S. a pathway to become citizens. </p><p>The executive orders on DACA and the Muslim ban will immediately appease immigration advocates, including those inside the tech industry, but the toughest battle still lies ahead: rallying support in Congress for immigration reform, which has failed for years. Still, there may be some in the Republican party who are willing to work with the Biden administration on key pillars of the proposal. On Wednesday, Sen. Lindsey Graham, who was a leading player in immigration talks during the Obama years and a top Trump ally, told CBS he is open to creating a path to citizenship for undocumented immigrants.</p><p>"What we are really looking for is making sure there is congressional action and there is a legislative solution," said Karolina Filipiak, senior director of government affairs at the Information Technology Industry Council, a tech trade group. "We don't want to be in the same position four years from now or eight years from now [with] a new administration that could rescind the program." </p><p><div class="ad-tag"><div class="ad-place-holder" data-pos="3">
</div>
</div></p><p>Companies including Facebook and Google are hoping to work with lawmakers in Congress to create bipartisan support for reform, particularly as they face regulatory headwinds of their own. </p>
Keep Reading
Show less
Emily Birnbaum ( @birnbaum_e) is a tech policy reporter with Protocol. Her coverage focuses on the U.S. government's attempts to regulate one of the most powerful industries in the world, with a focus on antitrust, privacy and politics. Previously, she worked as a tech policy reporter with The Hill after spending several months as a breaking news reporter. She is a Bethesda, Maryland native and proud Kenyon College alumna.
Protocol | Enterprise
Twilio CEO Jeff Lawson explains how he decided to face off with Parler
Also, why he thinks the $3.2 billion purchase of Segment will help Twilio's customers help their customers and why he's OK with being reliant on AWS.
"I think in a society, words matter, actions matter," Twilio CEO Jeff Lawson said. "That's why companies have things like Terms of Service and acceptable use policies."
Photo: Twilio
January 13, 2021
Tom Krazit ( @tomkrazit) is a senior reporter at Protocol, covering cloud computing and enterprise technology out of the Pacific Northwest. He has written and edited stories about the technology industry for almost two decades for publications such as IDG, CNET, paidContent, and GeekWire. He served as executive editor of Gigaom and Structure, and most recently produced a leading cloud computing newsletter called Mostly Cloudy.
January 13, 2021
Cloud computing companies were one of the few segments of society that enjoyed 2020. But even companies like Twilio, whose stock price tripled over the last 12 months, have had enough of 2021 already.
Last Friday, in the wake of the deadly attack on the Capitol, Twilio sent a letter to the right-wing social media app Parler notifying the company that it was violating Twilio's acceptable use policy for two of its authentication services. Parler decided to turn off Twilio's services rather than moderate calls for violence against elected officials on its app, which became a moot point after AWS cut Parler off from its own computing and storage services Sunday evening.
<p>This was not the way CEO Jeff Lawson wanted to kick off a round of publicity for his new book, "<a href="https://www.amazon.com/Ask-Your-Developer-Software-Developers/dp/0063018292" rel="noopener noreferrer" target="_blank">Ask Your Developer</a>." But little that has happened this month has gone according to plan, and he is quick to defend his decision from last week.</p><p><div class="ad-tag"><div class="ad-place-holder" data-pos="1">
</div>
</div></p><p>"If you walk into a movie theater and yell 'fire,' or you walk into an ice cream parlor and start yelling racist epithets, don't expect to be a customer of that establishment for very long," he said in an interview with Protocol Tuesday.</p><p>Lawson built Twilio into a $60 billion company based around communications APIs that allow app and web developers to build customer service and support capabilities into their own products. Just as Stripe became a central hub for payment technology for a generation of developers, Twilio enjoys a similar position when it comes to communications technology.</p><p>In his conversation with Protocol, Lawson explained the thinking behind Twilio's decision on Parler, the strategy behind <a href="https://www.protocol.com/newsletters/protocol-enterprise/why-did-twilio-pay-3-billion-for-segment-developers" target="_self">last year's $3.2 billion acquisition of Segment</a> and why developers should be free to choose no code, low code or "yo code" tools when building their apps.</p><p><em>This interview has been edited and condensed for clarity.</em></p><p><strong>Given everything that's happened in the last week, I was wondering if you could walk me through your decision to suspend service to Parler?</strong></p><p>I think in a society, words matter, actions matter. And we all have to work together to build a society. That's why companies have things like Terms of Service and acceptable use policies, because they say "What is acceptable?" There's a wide variety of things that are acceptable, but there's also fringe things that aren't.</p><p>This is not new to the online world, but think about the offline world. If you walk into a movie theater and yell "fire," or you walk into an ice cream parlor and start yelling racist epithets, don't expect to be a customer of that establishment for very long. I think what you see is the online world doing the same thing.</p><p><div class="ad-tag"><div class="ad-place-holder" data-pos="2">
</div>
</div></p><p>When people at the fringes of society are doing things like advocating for violence, overthrow[ing] the government, murder and hate, those are things that are largely illegal. And therefore companies like Twilio are saying: That's not how you're allowed to use our products, our platform,</p><p>This is fringe, extremist stuff. And if this were some other religion we'd be having a very different conversation. I just think that because it is happening in America, and it's largely white, it's "why are we shutting it down?" But when it's religious extremism of other varieties? People have a very clear sense: "Oh, yeah, that's wrong. And that shouldn't be allowed."</p><p><strong>What was the motivation behind the Segment deal?</strong></p><p>Our goal is to build the leading customer engagement platform. And customer engagement is essentially the sum of all of the interactions a company has with its customer that makes the customer love the company and want to buy products and want to be a loyal customer and repeat customer. And there's really two parts to building a great customer experience or customer engagement.</p><p>One is knowing your customer, and then, two, engaging with them using communications to say relevant things and to actually be a good communicator. Twilio has long played in the second category of helping companies to communicate with their customers, with text, with voice, [with] video, with chat across many different parts of the customer experience spectrum, whether it's on the ... marketing side, or the sales side, or the customer support side, or while you're using a product.</p><p><div class="ad-tag"><div class="ad-place-holder" data-pos="3">
</div>
</div></p><p>But the first thing you need to do is really understand your customer: understand who they are, what products they're looking for and how they use your product. That's what helps you be really good at engaging with customers. And that's what Segment does.</p><p>If you think about especially B2C companies, the way the story of who you are, what products you're interested in, what your purchase history is, that is all told through data that is on various systems throughout the company. The marketing system knows which emails you clicked on, maybe indicating that you liked those products; the ecommerce system knows which products you bought, historically; the customer support system knows how many times you've interacted with the company and the problems that you've had. The sum of understanding you is the sum of all those data points about you.</p><p>Segment allows companies to take all those bits of data that are siloed, all throughout the company and all those different computer systems, and actually pull them together to build a cohesive picture of, you know, you.</p><p><strong>The way that your customers go about this is they build these customer service products for themselves, using some of your basic plumbing tools. Do you see yourself increasingly competing with companies that sort of just provide that whole experience as a service?</strong></p><p><div class="ad-tag"><div class="ad-place-holder" data-pos="4">
</div>
</div></p><p>Generally speaking, companies need to be able to build parts of their customer experience that their customers value, because that's how you differentiate in the market. If you have the exact same customer experience as all your competitors, you're not going to be very differentiated, especially in the digital era where the perception that customers have so many companies is the digital experience you have.</p><p>Companies who win are companies who are good at building software that differentiates them in the eyes of their customers. Now, you do that by buying some things and building some things. And so a lot of companies will have, you know, SaaS products that they bought, but they need to make those SaaS products do new and interesting things to answer their customers' needs. </p><p>Many of the problems that arise are from companies having bought a whole bunch of SaaS applications that can't talk to each other. They become data silos, and getting the data out of one into the other in order to build a really cohesive experience is tremendously hard. Segment and Twilio allow you to take the investments you've made in a bunch of SaaS products and make them better.</p><p><strong>It appears you're all in on AWS from an infrastructure standpoint, behind your own platform. As you get bigger, have you considered multicloud, or bringing in other providers as you scale?</strong></p><p><div class="ad-tag"><div class="ad-place-holder" data-pos="5">
</div>
</div></p><p>We do have other vendors in the mix at Twilio, although the majority of our infrastructure is with AWS. That doesn't trouble me per se, because if you think about it, what do I want my product teams at Twilio doing? I want them building the things our customers care about.</p><p>Our customers don't really care about where our CPU and storage is. They care about what our products do, and how our products are making our customers' lives better. The more time that we get to spend doing those things that our customers care about and are buying from us, the better.</p><p>That's the great role of infrastructure; infrastructure lets you focus on the things that you do well as your core competency. We are very happy Amazon customers, they provide a great platform. Of course, it helps that I worked there in the very early days of AWS and I know a lot of the people there. But ultimately it's about picking the best tool for the job and then allowing your teams to focus on building the differentiation that your customers care about.</p><p><strong>You mentioned picking the best tool for the job. Increasingly, there are options from Microsoft and Google that are best-in-class tools for a job depending on different needs. But there is obviously some convenience of being all in on AWS, they have a certain way of doing things you understand, so your engineers don't have to learn different tricks. How do you balance that?</strong></p><p><div class="ad-tag"><div class="ad-place-holder" data-pos="6">
</div>
</div></p><p>Developers have the freedom to propose the tool for the job. Now, if it's a new tool we've never used before, there may be an architectural review, there may be a security review, etc. But ultimately, we want developers at Twilio who are building products to be able to essentially pick the right piece of infrastructure that's going to facilitate them building the best product.</p><p>That's one of the points that I make in my book, which is that company policies are useful to protect customers: say, security, and things like that. Of course, you want those things to be taken care of, but you also want to trust your developers to make the right technical decisions for the products they're building.</p><p>In fact, in the areas where we don't use AWS, that's because a developer thought that some other service would be the best tool for the job. And having gone through the appropriate reviews, we let them do that.</p><p><strong>How do low code and no code strategies play into your thinking over the next few years?</strong></p><p>There's three varieties of building tools, if you will: There's low code, no code, and I like to call it "yo code," for the traditional developer approach. There's the right place for each and the right audience for each.</p><p>I think there's builders of many different kinds in the world: There's builders who want to open an IDE and hammer out code, there's builders who want to drag and drop and there's builders who want something in the middle. And we support all those methods of building; we're not religious about saying, in order to build something, you have to write code, you have to love Python, or C or whatever it is.</p><p><div class="ad-tag"><div class="ad-place-holder" data-pos="7">
</div>
</div></p><p>We've got a variety of low code and no code solutions on our platform already. And I think the traction that we're getting with those really speaks to the fact that there's a lot of different ways to build and a lot of different types of builders out there. And ultimately, what we care about is that people are able to innovate at the pace and speed and with the tools that they're most comfortable with.</p><p><strong>Even "yo code" has incorporated a lot of abstractions over time, right? I mean, no one's doing assembly.</strong></p><p>I like to think of it like how we've used Moore's law over the last 40 years. Computers haven't gotten that much faster in terms of how we use them. It's not like your word processor is 10 million times faster than it was in 1990.</p><p>But how we've used Moore's law is to create abstractions that make it easier and easier to build complex, sophisticated software, including the internet itself, right? The fact that you can use Google Docs, which is a word processor, delivered entirely inside the web browser, that's so many layers of abstractions that basically compost the layers below, and make it so that, as a developer or a user, you don't have to care about those details anymore.</p><p><div class="ad-tag"><div class="ad-place-holder" data-pos="8">
</div>
</div></p><p>That frees you to actually build more sophisticated, more agile software that gets updated more frequently. In many ways, as technologists and as [a] society, we've used Moore's law to do that, instead of to make our computers all that much faster for the user.</p>
Keep Reading
Show less
Tom Krazit ( @tomkrazit) is a senior reporter at Protocol, covering cloud computing and enterprise technology out of the Pacific Northwest. He has written and edited stories about the technology industry for almost two decades for publications such as IDG, CNET, paidContent, and GeekWire. He served as executive editor of Gigaom and Structure, and most recently produced a leading cloud computing newsletter called Mostly Cloudy.
Protocol | Enterprise
The GE Mafia: How an old-school firm birthed a generation of tech leaders
The conglomerate hot-housed graduates in the '90s and '00s to create an adaptable army of tech talent. Now those execs are everywhere.
January 11, 2021
Joe Williams is a senior reporter at Protocol covering enterprise software, including industry giants like Salesforce, Microsoft, IBM and Oracle. He previously covered emerging technology for Business Insider. Joe can be reached at JWilliams@Protocol.com. To share information confidentially, he can also be contacted on a non-work device via Signal (+1-309-265-6120) or JPW53189@protonmail.com.
January 11, 2021
Look at the resumes of the top tech executives at the nation's largest companies and you're likely to find at least one theme: a stint at General Electric.
The once-quintessential American conglomerate has served as a launch pad for individuals now spearheading IT operations at companies such as Airbnb, United Airlines, Unilever, Morgan Stanley, AIG and dozens of others, according to analysis by Protocol.
<p>"You look across [top technologists in] the Fortune 100, it's pretty hard to find someone that doesn't have a background in GE," said Nationwide Chief Technology Officer Jim Fowler, who departed the company in 2018 after a 18-year tenure. "It was by design that it became something of a producer of talent," he added. </p><p>GE's ability to cultivate leaders is widely known. The company's famed former CEO Jack Welch, who stepped down in 2001 and <a href="https://www.nytimes.com/2020/03/02/business/jack-welch-died.html" rel="noopener noreferrer" target="_blank">passed away in 2020</a>, was laser-focused on pinpointing the most promising talent across the enterprise. But an early emphasis on IT — and the training of its executives to focus on it — put the company on the precipice of what is now one of the biggest trends in corporate America. </p><p><div class="ad-tag"><div class="ad-place-holder" data-pos="1">
</div>
</div></p><p>In recent years, enterprises have been rushing to digitize their operations, transforming IT leaders from help desk aficionados reporting to chief operating officers into top executives overseeing billion-dollar tech transformation efforts with a direct line to the CEO and board of directors. </p><p>That phenomenon has put new pressure on a position that, until recently, could function largely within its own silo. Chief information officers and similar positions are tasked with figuring out how tech such as cloud computing and artificial intelligence can not only support existing operations, but give the firm a new competitive advantage. Or, more bluntly, turn IT from a cost center into a profit center. </p><p>"That's just the way this role has evolved," Silicon Valley Bank CIO and 17-year GE veteran Mark Rohrwasser said. "If you are just a pure technologist … I don't think you're going to be in the key roles going forward."</p><p>That's perhaps the key reason why so many tech leaders today have been on GE's payroll: The company made it a priority to give its most promising future IT executives a broad range of skills, and pushed them into positions in the conglomerate's many different verticals: everything from healthcare to investment banking. Nationwide's Fowler, for example, worked in seven business units throughout his tenure; Rohrwasser served in eight. </p><p><div class="ad-tag"><div class="ad-place-holder" data-pos="2">
</div>
</div></p><p>"At that period of time, that was pretty unique. You were sitting at the same table as the business leaders, you were expected to engage," Rohrwasser said. "That's commonplace now."</p><h2>Taking tech in-house
</h2><p>GE also pinpointed those individuals early on in their careers and provided ample training and leadership development opportunities. It was a major investment in a relatively small selection of workers, one that former execs say is foreign in today's workforce. </p><p>"They figured out that they needed to pay attention to people throughout the continuum of their careers," said Lynn Boyd, a former talent development executive at GE who helped lead the IT leadership programs. "It's harder to do now, to invest that kind of money [only] to have people leave on." </p><p>While the company's troubles are well-publicized, including the way it <a href="https://www.wsj.com/articles/the-dimming-of-ges-bold-digital-dreams-11595044802" rel="noopener noreferrer" target="_blank">navigated its own digitization efforts</a>, GE's leadership development programs, many of which <a href="https://apnews.com/article/8040b13eb8d980e9ecb4198ff0e314a2" rel="noopener noreferrer" target="_blank">bore the fingerprints of Welch</a>, became a de facto gold standard in corporate America: Outside of CIOs, it's <a href="https://www.investmentnews.com/former-ge-executives-successful-as-ceos-elsewhere-5070" rel="noopener noreferrer" target="_blank">not uncommon to find a GE alumni as a CEO</a>.</p><p>And in a display of just how strong the connection between alumni is as a result of these initiatives, as many as 25 former GE tech execs get together roughly every other month at events sponsored by an outside consulting company, according to Fowler. </p><p>But the foresight to focus on IT, specifically during the late 1990s and beyond, was particularly astute. At a time when a glob of internet startups careened from success to failure amid the dot-com crash, what set GE apart from other corporations, former employees say, was its willingness to spend money to bolster its internal workforce. Instead of relying on the increasingly big providers like SAP and Oracle to manage the tech stacks, it chose to fend for itself. </p><p><div class="ad-tag"><div class="ad-place-holder" data-pos="3">
</div>
</div></p><p>Working with software companies, "if you didn't know what you were doing, you just gave the farm away. There's a lot of companies that don't know what they are doing," said Boyd, who is now an executive coach. "It became evident to GE early on, maybe because they were so big and were spending a pot-load of money on this … that we just can't have somebody come in and mismanage it," she added. </p><h2>CIO boot camp
</h2><p>To cultivate that kind of talent, GE started early, recruiting IT employees straight out of college. Those individuals would immediately be placed into a two-to-three week bootcamp, setting them up for what would hopefully become a decades-long career complete with several more iterations of executive training. Once the introductory sessions were over, each person was rotated to four different positions across the enterprise over the following two years. The program also encouraged these next-generation leaders to specialize in a specific technology field, like application development or coding.</p><p>"We basically took the material that was coming out of colleges and we would 'GE-ize' it and teach it what we do, why we do it and how does technology play into it," Fowler said. </p><p>As they progressed from early to mid-level careers, the coursework would intensify. Each year, roughly 30 first-time IT managers would be tapped to attend the Experience Information Management Program at GE's <a href="https://www.ge.com/news/reports/inside-crotonville-ges-corporate-vault-unlocked" rel="noopener noreferrer" target="_blank">storied Crotonville campus</a>. At the two-week gatherings that occurred four times a year, they could schmooze with other emerging leaders, such as reps from the top business schools in the country and executives from other divisions. (A standout "rock star" at these events was former GE CIO Gary Reiner, who still serves as a mentor to Fowler and others). But it wasn't all social: Each was given a business problem to work on and, after six months, would report in with a division CIO at the company on their progress. </p><p><div class="ad-tag"><div class="ad-place-holder" data-pos="4">
</div>
</div></p><p>"We called it a mini-MBA," Boyd said. "That put you on the map. If you weren't already on the map for other reasons, that put you on the map." </p><p>Ultimately, a segment of that group deemed executive material was then advanced to another gathering at Crotonville that focused more on developing long-term strategic vision. Of that group, as few as three would be ultimately tapped to serve as a group CIO, former staff said, leaving as many as 20 executives who were likely to jump ship in order to ultimately reach the coveted role. </p><p>That is one reason why there is such a large network of GE alumni at the nation's biggest businesses. But it also meant that, when those internal positions did become available, there was little delay in appointing new individuals. </p><p><div class="ad-tag"><div class="ad-place-holder" data-pos="5">
</div>
</div></p><p>"There was never a time when one of those top jobs came open that we didn't have a list of five to six really strong candidates just ready to walk in," Fowler said. </p><p>While GE still runs its postgraduate program for IT staff, it no longer runs the later-stage development programs. Instead, it is planning a new, more accelerated training program that will be rolled out this year. Still, two decades after Welch left GE, the focus on pinpointing leaders early on and exposing them to the numerous facets of the business remains. </p><p>"With the speed of change across the industries in which we operate, developing technology leaders with the skills to take on complex, real-world problems is critical," current CIO Nancy Anderson. "We're proud of the tech leaders we've produced at GE." </p>
Keep Reading
Show less
Joe Williams is a senior reporter at Protocol covering enterprise software, including industry giants like Salesforce, Microsoft, IBM and Oracle. He previously covered emerging technology for Business Insider. Joe can be reached at JWilliams@Protocol.com. To share information confidentially, he can also be contacted on a non-work device via Signal (+1-309-265-6120) or JPW53189@protonmail.com.
Protocol | Enterprise
Why observability is the new monitoring
Understanding software performance is an extremely important — and complex — undertaking for the modern enterprise. Simply watching the meter no longer works.
There's a lot to keep track of in modern software.
Image: Alexander Sinn/Kwamina2
January 11, 2021
Tom Krazit ( @tomkrazit) is a senior reporter at Protocol, covering cloud computing and enterprise technology out of the Pacific Northwest. He has written and edited stories about the technology industry for almost two decades for publications such as IDG, CNET, paidContent, and GeekWire. He served as executive editor of Gigaom and Structure, and most recently produced a leading cloud computing newsletter called Mostly Cloudy.
January 11, 2021
No unhappy complex system is alike: Each is unhappy in its own way. A growing line of business in software development, observability seeks to understand how and why modern software applications and teams become unhappy in order to set them on a path toward happiness, uptime and profit.
An evolution of monitoring software — which became popular during the rise of Web 2.0 applications and spawned companies such as Splunk, Datadog, New Relic and SolarWinds — observability takes the idea of simply watching IT systems a step further. While it's helpful to have dashboards that let administrators determine the health and performance of their applications at a glance, observability advocates believe what modern businesses really need are tools that help them understand the root cause of software issues.
<p>"To help me build better software, you can't just do everything reactively anymore," Bill Staples, president of New Relic, told Protocol. "If you work reactively in today's cloud environment, you're firefighting constantly."</p><p><div class="ad-tag"><div class="ad-place-holder" data-pos="1">
</div>
</div></p><p>The idea is that it's better for software developers to understand exactly which part of their code is causing a problem and why than rely on alerts that flag problems but require painstaking analysis to identify the cause. The time that can save is important: Forward-thinking software organizations move very quickly these days, preferring to deploy small changes to their code on a frequent basis rather than deploying big changes at a slower pace.</p><p>"If it takes you two months to ship your code, you're probably not high enough to ride this ride," said Charity Majors, co-founder and CTO of Honeycomb, one of several startups pushing the boundaries of this emerging field.</p><p>But it's not just coding: Observability tools can also help companies understand how their people are performing, and how the structure of their organization might be causing more problems than it is solving.</p><p>"The tech metrics don't mean anything without understanding the pressures that people are under when they are building the systems," said Nora Jones, co-founder and CEO of startup Jeli.</p><h3>Watching, waiting, commiserating
</h3><p>Administrators have been monitoring the performance of computers since the first was plugged into a wall. But the modern concept of application performance management started to come together alongside the wave of enterprise software innovation that came out of the Great Recession.</p><p>As new SaaS tools started to become some of the most important operational tools inside businesses, performance — always important — took on new meaning. Customers increasingly had expectations for how software delivered over the web should perform, and as lots of these application vendors built their services atop metered cloud computing platforms from AWS and others, they had to be very aware of how much computing resource they were consuming on the back end.</p><p><div class="ad-tag"><div class="ad-place-holder" data-pos="2">
</div>
</div></p><p>The growing consensus around the value of frequent deployment meant that software developers needed tools to quickly measure the impact of those changes so they could pull back a change that introduces a new problem, said John Allspaw, co-founder of Adaptive Capacity Labs, who played key roles at Yahoo's Flickr and later Etsy during the period in which monitoring became table stakes.</p><p>"There was a period of time where some companies totally got [the idea of frequent deployment], and can't imagine working a different way and other companies can't even imagine why you would even try to deploy more than once a day," he said. That latter group gets the benefits of continuous deployment now, he said, which has lifted the <a href="https://www.protocol.com/jfrog-ipo" target="_self">fortunes of companies like JFrog</a>, CircleCI and CloudBees, which have all built businesses around making software pipelines more efficient..</p><p>But early monitoring tools that were used to study software once it was deployed were passive, and didn't provide as complete a picture of how an application was performing.</p><p>"Maybe 10 years ago, the way things would work is developers would write the application and then hand it off to an IT pro, who would probably deploy it onto a server in a data center," Staples said. "Meanwhile, they hope that the system keeps up and running and the IT pro will tell them if something breaks — otherwise, they just go on to the next feature."</p><p><div class="ad-tag"><div class="ad-place-holder" data-pos="3">
</div>
</div></p><p>A decade later, that approach will not fly. Companies no longer separate software development from operations: a shift known as DevOps, which calls for closer cooperation between the teams and forces developers to be more aware of the impact of their changes.</p><p>One of the big risks of making changes to a monolithic application was the chance that you could cause a difficult-to-detect problem in a completely unrelated part of the app. Microservices changed that, allowing developers to break their applications down into lots of smaller pieces that can be operated and tweaked separately. At the same time, cloud adoptees started moving toward deploying their applications in containers, which meant they could be deployed across a wide range of servers.</p><p>Something more sophisticated was required to understand how all of that was working.</p><h3>Disturbance in the system
</h3><p>"Observability is all about looking at [the application]; shipping the code and looking through instrumentation to know if it is doing what I expected it to do," Majors said.</p><p>Honeycomb's approach was modeled after the notion of control theory in mechanical engineering, she said. Its tools give operations engineers a way to build instrumentation into their code to flag problems as they happen, allowing them to discover exactly where something has gone wrong rather than seeing a poor end-user performance and digging through the code to find the problem.</p><p><div class="ad-tag"><div class="ad-place-holder" data-pos="4">
</div>
</div></p><p>Usually software organizations aren't dealing with issues like the massive outage that took down Slack last week: Most software incidents are minor, Staples said.</p><p>"When things fail, they don't fail completely. You see drop-off rates, you see errors fail for 10 percent of users," he said. "What you do as an engineer is you're constantly using those signals: to know where to go invest more, whether it's improving the feature to get the customer through a trouble spot that's slowing them down and keeping them on, whether it's scaling a part of the system to increase the performance of that component or other things."</p><p>Fixing those problems quickly, rather than spending time debugging a poorly-performing application, can give software teams more time to focus on improving their products.</p><p>"Code is like food; it rots," Majors said. Preventing that rot from dragging down an entire system as quickly as possible can prevent bigger outages down the road that cost companies money.</p><p>But it's not just the code that needs observing; it's the people.</p><p>Jones is a veteran of high-performing software teams at Slack, Netflix and Jet.com. Yet even within companies at the forefront of software development practices, organizational structures can make as much of an impact on healthy applications as coding practices, she said.</p><p>At one of those companies (she declined to share which), a disproportionately large number of performance issues happened within a short period of time each year, and the company was having a hard time figuring out what was causing the problem. Turns out, those problematic periods came just after its annual promotions cycle, during which engineers had scrambled to ship as much code as they could in a short period of time to hit their goals for the year.</p><p><div class="ad-tag"><div class="ad-place-holder" data-pos="5">
</div>
</div></p><p>"It wasn't their fault. It was the system that was created at the company," Jones said. "Understanding that these promotion cycles were being correlated to an increase in incidents, because people were trying to get things done really quickly, actually incentivized the company to completely restructure how they did their promotion cycles, which led to this kind of stuff not happening as much."</p><p>Insights like that led Jones to found Jeli, which allows companies to evaluate and monitor how their organizational structures map against their coding practices. The company <a href="https://techcrunch.com/2020/12/07/jeli-io-announces-4m-seed-to-build-incident-analysis-platform/" rel="noopener noreferrer" target="_blank">just raised a $4 million seed round</a> to build out tools for that type of customer.</p><h3>Infinite runway
</h3><p>The promise of observability tools is preventative maintenance: Not only will you be able to see and react to problems faster than current monitoring tools allow, but you'll also be able to glean insights from that data in a way that helps protect against future problems yet to rear their head.</p><p>The surge of interest in this space from upstarts and traditional monitoring companies has lots of ideas flying fast and furious, but it will take some time before that promise will be met, according to Allspaw.</p><p>"We have enough problems with the known unknowns," he said. "The runway to make progress on that is as close to infinite as we can get."</p><p><div class="ad-tag"><div class="ad-place-holder" data-pos="6">
</div>
</div></p><p>Still, 25 years into the internet revolution, we've come to expect certain levels of performance and reliability from our web and mobile applications. Big organizations like AWS, Google, Netflix and others are well down the observability road inside their own companies. And now the tools and companies that will bring those insights to the rest of us are starting to get traction.</p>
From Your Site Articles
Keep Reading
Show less
Tom Krazit ( @tomkrazit) is a senior reporter at Protocol, covering cloud computing and enterprise technology out of the Pacific Northwest. He has written and edited stories about the technology industry for almost two decades for publications such as IDG, CNET, paidContent, and GeekWire. He served as executive editor of Gigaom and Structure, and most recently produced a leading cloud computing newsletter called Mostly Cloudy.
Protocol | Enterprise
Databricks plans to take on Snowflake and Google and score a huge IPO
Even against intensifying competition, Databricks hopes to be a hit when it heads to the public markets this year.
Ali Ghodsi is the CEO of Databricks.
Photo: Databricks
January 11, 2021
Joe Williams is a senior reporter at Protocol covering enterprise software, including industry giants like Salesforce, Microsoft, IBM and Oracle. He previously covered emerging technology for Business Insider. Joe can be reached at JWilliams@Protocol.com. To share information confidentially, he can also be contacted on a non-work device via Signal (+1-309-265-6120) or JPW53189@protonmail.com.
January 6, 2021
Enterprise software had a huge 2020 on Wall Street as companies such as Snowflake and C3.ai went public with blockbuster initial offerings. Databricks CEO Ali Ghodsi is hoping to ride the same wave in 2021.
The public debut of the data analytics startup, valued at $6.2 billion, is among the most-watched IPOs for the year. And for good reason: It competes in a similar space as the much-hyped Snowflake, helping customers find the data to power the algorithms that help with everything from picking which products to order to which candidates to bring in for job interviews. While Databricks has been tight-lipped on its specific plans, including which bankers it is tapping to help navigate the often arduous process, it is taking steps internally to prepare.
<p>"There's a lot of compliance that goes into it," Ghodsi told Protocol. "Much of it is making sure that we have that predictability that you need when you IPO."</p><p><div class="ad-tag"><div class="ad-place-holder" data-pos="1">
</div>
</div></p><p>Unlike other startups that tend to bring on a new chief financial officer to oversee the IPO, Databricks is already set: CFO Dave Conte previously helped take Splunk public in 2012. But it does plan to make other key hires, according to Ghodsi, including a new head of engineering and new board members. </p><p>Databricks is also shoring up its own internal software, including investing "massively" in Workday, according to Ghodsi. And while an IPO is looming, the company is weighing another funding round in the interim amid what Ghodsi described as a rush of calls from venture capital firms looking to invest. </p><p>While Ghodsi says 2020 brought a "significant acceleration" of its business, he declined to provide any up-to-date revenue figures. Databricks said in October that it passed a $350 million revenue run-rate at the end of Q3, up from $200 million a year earlier — an acknowledgement that, despite COVID-19 gutting some sectors, the appetite for data analytics and artificial intelligence is robust. Still, the competition is fierce and, despite <a href="https://databricks.com/company/newsroom/press-releases/databricks-and-snowflake-announce-strategic-partnership" rel="noopener noreferrer" target="_blank">notably partnering with Snowflake in 2018</a> (a partnership that continues today), Databricks faces an increasing rivalry from the current industry darling, as well as Google's BigQuery. </p><p>"For data science and analytics, those are really going to be the three competitors," Sapphire Ventures co-founder Jai Das said. "But the market is still so big, I think all three are going to be successful."</p><h2>First time IPO'er
</h2><p>The investor appetite is, of course, a big vote of confidence in Databricks. But it's also a show of support for Ghodsi, a former researcher at the Swedish Institute of Computer Science who still teaches classes at the University of California, Berkeley. </p><p><div class="ad-tag"><div class="ad-place-holder" data-pos="2">
</div>
</div></p><p>It's not uncommon for companies to bring on CEOs skilled in taking startups public prior to their own offerings. Snowflake, for instance, poached Frank Slootman, who <a href="https://www.cnbc.com/2020/08/30/snowflake-ceo-frank-slootman-pursues-third-ipo.html" rel="noopener noreferrer" target="_blank">led the IPOs for ServiceNow and Data Domain</a>, for its own offering. </p><p>That's not the approach at Databricks. It'll be Ghodsi's first time leading through the process. But while he doesn't come armed with detailed knowledge of the public markets, he believes a deep understanding of Databricks and its offerings is a much more significant advantage. </p><p>"These professional CEOs very rarely are good at that because they don't really understand the products that your company is selling," Ghodsi said. "Oftentimes what happens is that you have great technical founders, they start the company, you launch the product, but you don't really hit your revenue targets. Then the board looks at making changes to fix that. That's basically the death of the company in the long term." </p><p>For Ghodsi, the long-term mindset meant largely ignoring the frothy equity markets in 2020 that led some firms to <a href="https://www.protocol.com/how-a-spac-works" target="_self">quickly launch a public offering</a>. And despite claiming it will be a "really successful" offering in 2021, he views the IPO as just one step in the vision to make Databricks akin to Salesforce — an enterprise software firm that <a href="https://www.protocol.com/salesforce-blockbuster-2020-benioff" target="_self">continues to grow in revenue and influence</a> roughly 16 years after its own IPO. </p><p><div class="ad-tag"><div class="ad-place-holder" data-pos="3">
</div>
</div></p><p>To get to this point, however, took what Ghodsi described as three phases of evolution. </p><p>When he and six other co-founders started the company in 2013, Ghodsi served as vice president of engineering and product management. That gave him a leading voice in figuring out exactly what market needs its offering was solving. But it also meant forgoing lofty revenue targets — it made roughly $1 million that year — to instead focus on "making sure that this is a product that companies really need and love." </p><p>The growth came when Ghodsi took the reins, assuming the role of CEO in 2016. At the time, Databricks poured money into its "go-to-market function," including sales and marketing. That also included developing a "customer success" function, an increasingly common facet to organizations that sell their services via the cloud, designed to keep close contact with customers and ensure they don't pivot to another provider. By the end of 2019, Databricks reported a $200 million revenue run, up from "almost no revenue," according to a statement at the time. </p><p>That also meant hiring beyond 150 total employees, breaching what is commonly referred to as "Dunbar's number" or the maximum number of people that one person can keep in social contact with. In a practical sense, that meant Ghodsi no longer knew the names of all the employees.</p><p>Now, Databricks is in what Ghodsi calls the "operational excellence" phase, or making sure all the company's processes are "aligned and rowing in the same direction." It's produced one of the company's biggest new product releases to date: SQL Analytics, which helps data scientists or analysts more easily run queries on a vast amount of unstructured data through commonly used tools such as Tableau. </p><p><div class="ad-tag"><div class="ad-place-holder" data-pos="4">
</div>
</div></p><p>While similar offerings are available from Snowflake and BigQuery, the release is Databricks' way of championing the "data lakehouse," or a combination of curated information alongside unstructured data. Typically, they reside in two different stacks, often referred to as a data lake (a vast pool of information) and a data warehouse (a more structured storage center). </p><p>And more product announcements are in store for 2021 as Databricks looks to capitalize on its fastest growing industry sectors, including health care. </p><p>"Databricks is behind the scenes of pretty much all the vaccines that are being talked about," Ghodsi said. He added: "Health care providers are using it to track inventory and how crowded their hospitals are." </p><p>But despite Ghodsi's best attempt at downplaying the current financial environment's influence on the company's next steps, it's clear that competitors' successes are top of mind — and that the biggest event of Databricks' year will be held on Wall Street. </p><p>"It's a really, really good IPO climate and we will be IPO-ready," he said.</p>
Keep Reading
Show less
Joe Williams is a senior reporter at Protocol covering enterprise software, including industry giants like Salesforce, Microsoft, IBM and Oracle. He previously covered emerging technology for Business Insider. Joe can be reached at JWilliams@Protocol.com. To share information confidentially, he can also be contacted on a non-work device via Signal (+1-309-265-6120) or JPW53189@protonmail.com.
Latest Stories
See more
Most Popular
Bulletins
Get Source Code in your inbox
David Pierce's daily analysis of the tech news that matters.