Earlier this month we hosted the second annual Data-Centric Architecture Forum (#DCAF2020) in Fort Collins, CO. These are the highlights.
Last year, (2019) we hosted the first Data-Centric Architecture conference. (Trip report).
In 2019, the focus was on getting a sketch of a reference architecture (click here to see).
We convened shortly after the event and began planning the next year’s conference. We decided that what we wanted to see were implementations either of a complete stack or at least some of the key layers. We were not disappointed.
Triple Store-Based Offerings
There were two products I expected not to like. Flur.ee was one and Timbr was the other. I must say I was very pleasantly surprised in both cases. Flur.ee is an RDF triplestore implemented on top of blockchain. I was prepared not to like this since it just sounded like fad surfing.
Flur.ee disposed all the proof of work /stake /authority in favor of a simple encrypted ledger. Doing so improved efficiency immensely. Leaning into the immutability aspect, it gives them a number of pretty unique features. It was the first week of February, so use case #1 was staring us in the face— the Iowa caucuses and their inability to count their votes credibly. Imagine an encrypted and immutable block chain with an rdf engine on top. Another use case that Brian Platz demonstrated was their version of a time machine. They whipped up a simple app, with a simple graphic, and then introduced a slider that allowed them to see the contents of the app and the graphic over time. They are well-funded and well-connected, with lots of potential.
The other platform I wasn’t prepared to like was Timbr, a startup from Israel. It is a virtual, federated semantic database implemented in relational technology. At one level this seems like a big step backward, and/or a massive cop out, but I was impressed with what we saw.
The first surprise was that they can import owl directly and cover most of the inferences available in the owl spec (not all, and I’ve forgotten which ones aren’t supported, but enough to be surprising and useful). This was despite the fact that there was no triplestore. Tzvi and Amit Weitzner showed how it seamlessly connects (via jdbc connectors) to analytics tools (such as Tableaux) as well at data stores (not only relational databases, but anything that can act like a relational database, such as Spark).
My take is this may be a good interim technology for companies that are overly attached either to their analytics tools, or their data stores. Injecting semantics in the middle of that stack may be an eye opener for many people and getting a semantic level gearing down of complexity could be a big win. I tend to think of this as interim, but there may be many places where this is the end game.
TerminusDB is an open source microservice triplestore. Very impressive capability and speed. What at first seemed odd, (implementing their own graph query language) becomes less odd with reflection. Kevin Feeney showed us their query language is based on a JSON structure rather than string syntax, which gives it two advantages: one is reducing (perhaps eliminating) the possibility of the equivalent of a SQL injection, and the other is providing some compossibility features, a bit like GraphQL.
There were many more platform reports, but in the interest of space, wanted to limit it to those readers may be least likely to have heard of.
We also had a number of experience reports of people trying out the approach and reporting back what they found.
SchemaApp focuses on helping clients take maximum advantage of schema.org markup in their web sites and content, and thereby improving their SEO. They have decided to eat their own dog food and build both their client-facing as well as much of their internal systems’ functionality using the DCA approach.
Mark vanBerkel walked us though most of the layers in the DCA onion/grapefruit diagram, and described what they had done to implement each, and reported on benefits of each.
Robert Dyas showed us a complete and mature DCA platform that his company has been marketing for many years. It is based on relational technology (many people ask me if DCA can be implemented on relational, and this is one of the reference implementations I go to).
It was very illuminating how much can be done with model-driven and what you learn after years of implementing this. He recounted a recent competitive win where one of their competitors had bid 20,000 hours against the AppSynergy bid of 800 hours. They won the bid. Sometimes an existence proof is more compelling than trying to explain how this is possible.
We all agree that one of the toughest challenges in the DCA approach is how to handle a wide-ranging number of security (authorization) problems in the absence of application code. We saw some aspects of different approaches in many of the platform projects, but Jamie Gulden’s presentation seemed to lay out all the potential approaches and their tradeoffs.
We’re putting forward the notion that DCA needs the equivalent of DevOps, which we are currently calling SemOps. Mark Wallace led a presentation that focused on what is similar and what is different between DevOps and SemOps, along with some of the tooling that is already available.
The DCAF was an outgrowth of the Estes Park Group, which is a monthly teleconference where we discuss issues and technologies related to implementing Data-Centric Architectures. If you are interested in participating contact me or email@example.com.
We are just beginning the planning for next year’s DCAF. There are a few things we know for sure:
- It will be the first week of February
- It will focus on case studies (now that we have existence proof with these technologies we’d like to start reviewing the actual experience of implementing them)
- There will be vendor sponsors (booths and presentations)
- It will be highly interactive, engaging and mind altering
- You will meet a lot of very committed passionate people
- There will be very steep early, early bird discounts as soon as we publish the venue
We have not yet finalized the venue (city or hotel), so either send Melissa a note to be notified, or keep your eye on https://www.semanticarts.com/events/
To get an idea of the rest of the material I didn’t cover in this write up check out: https://www.semanticarts.com/dcc/
P.S. this just reminded me, we sold out this year and had to stop taking registrations about a month before the event, so book early.