Dec 5, 2013 · 44m · mad
Panel discussion // Data Driven NYC #13 // Dec 2012
Oren (Sqrrl) · 9m spoken Eric (MemSQL) · 8m spoken Drew (InfoChimps) · 8m spoken Matt Turck · 3m spoken Alexander Younghoff · 1m spoken Joseph Adler · 1m spoken Charles Amor · 1m spoken Carter Schoenwald · 50s spoken Benjamin Swamy · 30s spoken Matthew Rutley · 26s spoken Mandeep Singh · 21s spoken
0:00 / 0:00
▶ Watch on YouTube →
gold bands on the timeline = statements, start to end. Hover to read, click to jump. CC turns on captions
At Data Driven NYC #13, host Matt Turck moderates a panel featuring founders Oren Falkowitz (Sqrrl), Drew Bansal (InfoChimps), and Eric Frenkiel (MemSQL) discussing startup origins, database architectures, enterprise security, and the future of big data.
How this conversation actually went
Every chapter scored 0–10 on four independent dynamics. Hover any point for the reasoning behind the score. Matt holds 8.3% of the talking time here. How this is scored →
Matt as informed peer 1.7 Guest teaching 2.8 Guest disagreement 1.7 Matt pushing back 1.0
speaking balance: gold is Matt, purple is the guest (3 minute bins)
Sharpest disagreement ▶ 33:55 Audience member challenges SQL relevance
An audience member forcefully asserts that no data scientists use SQL anymore and directly questions why anyone would choose MemSQL over MongoDB.
Hardest push from Matt ▶ 40:36 Matt cuts off off-topic questionMatt politely but firmly shuts down an unproductive and confusing audience query about satellite-free GPS, insisting it be taken offline.
Biggest teaching moment ▶ 35:14 Eric details technical flaws of MongoDBEric educates the audience on MongoDB's performance limitations, specifically contrasting its global write lock and schema repetition with MemSQL's concurrent relational engine.
Matt holds his own ▶ 10:35 Matt challenges premise on open-source softwareWhen Drew suggests that difficult technical problems are open-sourced, Matt counters with domain knowledge by pointing out that databases are notoriously hard yet predominantly open-sourced.
the scores for every segment, with the reasoning behind each
| Chapter | Topic | Matt as informed peer | Guest teaching | Guest disagreement | Matt pushing back | Why |
|---|---|---|---|---|---|---|
| Panelist Origins and Company Evolution | 1 | 2 | 1 | 1 | Matt opens the panel with lightweight origin questions about the founders' transitions into entrepreneurship. Oren playfully corrects Matt's narrative about VCs forcing Sqrrl to relocate to Boston. | |
| MemSQL Stealth Development and In-Memory Architecture | 1 | 2 | 1 | 0 | An audience member asks about MemSQL's stealth period, prompting Eric to explain the technical necessity of building an in-memory database from scratch. The host acts purely as a facilitator passing the microphone. | |
| Open Source vs. Proprietary Database Models | 3 | 3 | 2 | 2 | Matt prompts a debate on open-source versus proprietary models and interjects to note that complex database software is frequently open-sourced. Eric argues open source is a great marketing tool but a poor business model. | |
| Database Paradigms: NoSQL vs. NewSQL | 4 | 4 | 2 | 2 | Matt questions whether NoSQL and NewSQL vendors compete directly and asks about enterprise cloud adoption across industry verticals. The guests explain how relational models are evolving past NoSQL key-value limitations. | |
| Audience Q&A: Security Policies and Unstructured Data | 1 | 2 | 1 | 0 | Audience members inquire about enterprise security standards and unstructured data capabilities. Oren explains how fine-grained security policies can be embedded directly into database keys. | |
| Audience Q&A: Data Locality and Enterprise Compliance | 1 | 3 | 2 | 1 | Audience questions focus on memory locality and enterprise compliance. Oren uses a colorful analogy comparing specialized extreme performance tools to a Formula One car versus ten thousand versatile Kias. | |
| Audience Q&A: Shared-Nothing Architecture and SQL vs. MongoDB | 1 | 4 | 3 | 0 | Audience members challenge panelists on shared-nothing architectures, data marketplace value, and SQL versus MongoDB relevance. Eric details MongoDB's performance constraints, including its global write lock. | |
| Audience Q&A: Data Obsolescence and Multi-Database Architectures | 1 | 3 | 1 | 0 | An audience member asks about handling data obsolescence over time. Drew outlines how enterprise platforms tier data storage dynamically based on time horizons. | |
| Audience Q&A: Network Inventions and Ethics in Big Data | 2 | 2 | 2 | 3 | An audience member proposes an unfocused question about a nation-scale GPS invention, prompting Matt to firmly step in and redirect the discussion offline. Drew then shares thoughts on big data ethics and privacy. |
Statements from this episode (18)
InfoChimps pivoted from data content marketplace to infrastructure platform
“We couldn't get enough revenue by just selling the content. But as soon as people found out, we pivoted to, hey, this platform we built to support our marketplace, you can have one of these.”
MemSQL raised millions post-YC and stayed stealth for 15 months
“We were able to raise a couple million following the program. We shut up shop in San Francisco and then basically went stealth for 1415 months. Launched the product just this year in June”
Eric Frenkiel claims building a database in 14 months is a record
“Doing it in 14 months is pretty much, I think, you know, a record.”
MemSQL founders entered Y Combinator with zero lines of code
“The funny thing is, is that we entered the program with zero lines of code.”
InfoChimps receives more external code commits than internal ones
“And we're getting more commits from folks outside the company now than we are from inside”
MemSQL's Eric Frenkiel: Open source is a marketing channel, not business model
“I think open source is, is fantastic. It's a great marketing channel. I don't think it's a great business model.”
Eric Frenkiel: MemSQL would not exist if launched as an open-source project
“Fundamentally I don't think MemSQL would exist if we had to go into it the open source way.”
Sqrrl's Oren Falkowitz predicts large enterprises will adopt polyglot database architectures
“So, I think for large enterprises, they'll probably want one of everything.”
MemSQL's Eric Frenkiel argues analytics cannot be performed in key-value stores
“I mean I think, I mean, it's, you can't do analytics in key value.”
Drew Bansal predicts enterprises will put sensitive data in cloud by 2022
“I can see a world in which 10 years from now those are also ok.”
MemSQL's Eric Frenkiel argues on-premise infrastructure costs a fraction of AWS EC2
“I have a lot of you know, folks that spend tens of thousands of dollars a month on EC-II and they could do it for a fraction of the price if they did it on premise.”
Frenkiel: MemSQL on EC2 achieves 1ms latency vs 60ms for MySQL
“Whenever you go with MM-SQL on EC-II, for example, you will see, like, one millisecond response times, and a Postgres or a MySQL is on the order of 60.”
Eric Frenkiel: Writing a lightweight parser is easier than unstructured analysis
“What we typically see with unstructured data is it's far easier to write a lightweight parser around that log file, for example, and get it structured very readily.”
Frenkiel: Moving data from disk to memory yields a 50x speedup
“Just putting data from disk into memory, you'll see, you know, on the order of a 50 x improvement.”
InfoChimps co-founder: Companies pay for internal data integration, not public data
“So I feel like data mashing together and shareability is hugely important, and it's something that folks are willing to pay for within an organization for internal data. Maybe not necessarily, at least in our company's experience, for data that is coming exter…”
Eric Frenkiel criticizes MongoDB for relying on a global write lock
“Mongo also has a global write lock. So one writer will block everyone else until it completes.”