The final day of the Summit, which for me had the fewest formal sessions and the most one-on-ones with Red Hat folks.
Women in Open Source Award
Inaugurated in 2015, so the second year Red Hat have run this.
Jessica McKellar: Community award - works on Python @ Dropbox, Python development, and Python community work.
Preeti Murthy: Academic award - currently working in academia, Preeti started kernel hacking with the power management systems (your laptop says “thank you”) and moved into kernel development generally.
Tom Soderstrom - CTO, NASA JPL
Over 5,000 curious people work at JPL: capture data from cool spacecraft and do interesting things with it. JPL have a history of firsts and innovation, including Voyager and “cars on mars”.
Sustaining innovation: periodically the JPL will look at what’s coming; the current trend is to look at the consumer space (where most of the rapid innovation is happening) and bring it in the enterprise space.
Consider Internet of things: is it a toy? Yes. Is it a security problem? Oh yes. Will it be useful? Probably.
How does JPL work: adopting open development and startup approaches. If people care about a thing, keep working on it, if they don’t, drop it.
“Today’s toy is tomorrow’s tool.”
Earth 2.0 found at Kepler-186, in the habitable zone.
Financial services breakout: Banks’ “Uber Moment”
Moderated by John Marx of Red Hat. This was a two hour session; one hour of the speakers talking, and one hour of general Q&A.
Kfir Godrich - BlackRock CTO Brad Harrison - TD Bank - Senior Managing Engineer Jason Valentino - Capital One
This talk is in part motivated by the thinking that the financial sector is next on the list of industries to be upturned by VC-backed companies. A report by Anthony Jenkins claims: “Half of branches and retail staff will dissappear in the next few years.” A huge amount of seed and Series A funding is already moving into Fintech.
Red Hat see their large customers partnering with, studying, and aquiring fintech companies.
“We are not a bank, even though regulators treat us like one, based on the structure of some of our products.”
“We are in the top 20 places to work. We are the only non-technology company in the top 20. Many people don’t know who we are, and sometimes we are happy about that.”
- $4.6T of funds under management by 115 investment teams.
- “Technology without people is nothing. Technology with people is everything.”
- “We are still building datacentres. We are still building our own software. We like to own our own destiny. We embrace open source.” <<<<<<< HEAD
- One software stack is I.C., an operating system for asset managers. Where do we go from here? Can we democratise this system for financial services?
- Aladdin was their Uber moment: sharing something that worked with other people. Cloud 2.5 = AladdinStack is what’s next.
- “If you put this new tech stack on an old stack, you have achieved nothing. People who build new stuff on old stuff, you are going to get hurt.”
- Develop new stuff as a speedboat alongside the barge, but develop the speedboat so it can run slowly when you need it to.
- “Simplexity” - you need to have processes that are simple to understand. If things aren’t simple, it will bite you.
Advice: * Know your business. * Don’t be scared. Embrace the opportunities.
- TD Bank is the largest bank in Canada; 1300 branches, 80,000+ employees.
- One of the largest OpenStack deployments around.
- Everything Brad’s unit does is about getting engineers off the critical path. “If you take nothing else away from this talk, it’s that getting your engineers off the critical path to create a legendary experience.”
- “Large financial institutions need to learn to innovate before the fintechs learn to scale.”
- TD Bank are focused on delivering infrastructure via self-service, of allowing developers to pick and choose tools more easily. Containers, immutable infrastructure, etc.
- All applications are not equal.
- Do not assume all applications will be able to take advatange of new infrastructure.
- Not all applications will make it - some will run in legacy space, some will just be eliminated by new applications.
- Microservices, APIs, and mobile dev will outpace the back end.
- Build and release processes need to accomodate that reality.
- A dream: one platform, any cloud.
- Get out of your developers’ way.
- “If your title is DevOps anything you’re probably not.”
- Capital One is a young bank - IPO in 1994. Hit the Fortune 500 in 2000.
- Invented a credit modelling system for credit cards. “We were big data before big data was cool.”
- Developed into a fat, inefficient company during early growth, over-focused on credit cards.
- Became a more orthodox bank, branching into other forms of banking like retail. This was driven partly by a desire to access alternative sources of funds, like customer deposits, and partly to de-risk a dependency on one product line.
- Worst Decision Ever: starting to operate like competing banks. Outsourced everything technology-relaed in the mid-2000s because that’s what everyone else was doing.
- It was great until people had new product ideas.
- The outsourcers ground all innovation to a halt.
- “Outsourcing was the shittiest decision we ever made as a company.”
- Rebuilding your tech capability was hard. Really, really hard.
- With a history of outsourcing, no-one you want to hire wants to come and work for you.
- Set up new offices, cool offices, nice environments.
- Had to bring in talent from startup land because they had no-one qualified to build things any more.
- Moving to agile was hard - too much of the ceremony of agile and not enough of the intent of agile.
- Perserverence and showing value was the foot in the door to convince the broader business that the Labs were worthwhile.
- Recent aquisitions have put the aquired teams in charge of global capability - they aren’t product aquisitions any more, but team and talent aquisitions. It was noted by all the speakers at this point that there’s a common US pattern of aquiring startups for a product, API, or team, and then closing them off to competitors who might be using them. That’s a great deal more ruthless that we typically see in NZ business culture, and also a warning to people who might be thinking they can just rely on an API.
- Mobile is great, but IoT will be the next big thing - talking to your Alexa about finances is weird, but it’ll happen more.
- What are you doing around blockchain?
- Jason: Watching. Staying informed.
- Kfir: Experimenting over the last year. A chaotic landscape, but he wants to have more fun. One third of the Blacrock sponsored hackathon (100 projects) working on blockchain. There’s about 5 years of regulatory work to do there. Mostly interested in private blockchains. Not yet strategic.
- Brad: Watching brief.
- Kfir: Most interesting deployment is for enterprise authentication.
- Red Hat have a blockchain group because all their finsec customers are working with it.
- How do Capital One organise their lab?
- Lots of tension betweem product development and technology.
- Needed to be a drastic change. Admitted we don’t know what we don’t know.
- Brought in Skip Potter to set up a lab.
- Don’t just use your best executive. They won’t change enough.
- It’s not just a tech change, it’s a product change as well.
- Teams shouldn’t be bigger than can be fed by two pizzas.
- When building new things, embed your ops people in the teams. Agile will fail when one team has to reach out to a chain of other agile teams.
- Brad, where are you on your journey with containers?
- Incubation, narrow use cases around JBoss, node.js. Needs to be well-understood.
- Working with the newer apps, not retrofitting.
- What are Capital One doing with containers?
- No formal strategy, but people are using them. Mostly Docker being used ad-hoc for development environments.
- Trying to avoid being too dependent on AWS services.
- When you talk about speed of experiments, what does fast look like?
- It depends what you’re doing, where you’ve come from, what good looks like for your organissation is doing.
- Innocation dies if you run RFPs. Just build things and chose what works better. Getting to market with the wrong decision is better than spending time in RFP.
- What’s the average developer desktop look like?
- Brad: Very diverse. Depends on the developer and what stack they’re using. Ranges from locked-down traditional corporate Windows to local admin, Windows, Macs, etc. Lack of rigid mandates is a strength.
- Kfir: It’s not just developers, it’s everyone - if the developers have something cool, everyone wants it. So it’s an uplift for the whole organisation.
- Jason: We have a shopping mall approach. We have 50,000 people, so we give people a choice from an actual inhouse store.
- How do you decide when to build vs when to buy software/consume APIs:
- Jason: Anything that’s a competitive difference, and not just for the customer touch points, that’s right down to e.g. the APIs, the credit decisioning - the whole chain.
- Red Hat: A lot of aquisitions are being driven by both aquiring those capabilities and denying them to competitors who have come to depend on e.g. a product or service.
- What would you do differently with the lab?
- Jason: I’d burn a lot fewer bridges. I wouldn’t pick fewer unnecessary fights, to be better partners.
- Don’t have little fights, save your political capital for big fights.
- App resilience challenges:
- It’s transforming the mindset of the developers.
- The challenges of ROI, regulation: how do you justify what you want to do in that context? How do you measure it?
- Kfir: I have a flat budget. It’s up to me to spend it as I see best, so I do what makes sense. There is no “technology/business” divide.
- Jason: Mobile’s easy. There are NPS scores, A/B testing, and so on.
- Getting cloud approval:
- Jason: AWS are interesting because they won’t jump through your hoops like other vendors. They are helpful, but they won’t change for you. They reserve the right to cut you off with 6 months notice, so don’t be too dependent on their stack.
- Kfir: if you educate, regulators can be very very helpful, and many of them are themselves adopting new technologies.
- Red Hat: The cloud providers will often be helpful.
- APIs hosted in cloud:
- Jason: staying well clear of anything that’s not encrypted, and they need to control the keys. Decisioning stays in the datacentre.
- Kfir: No APIs in the cloud.
- TD Bank: Ditto.
- How do you balance legacy with net new development?
- Jason: MUMPS looks great 5 years ago, but mobile is driving more load than it can cope with, to the point where they have to grow the mainframe. Don’t do that; cache for a few years to tide you over the load problems while build a new capability to replace the legacy system.
- TD Bank were looking to see core systems on OpenStack. How is that going 18 months into that?
- It’s working. A lot of the problem is developer education. They have grown faster than they expected - well beyond their original projection.
- How do you solve talent aquisition problems?
- Going right back to school to get better diversity.