Apr 13, 2016 · 23m · mad

A Virtual Distributed Storage System // Haoyuan Li, Alluxio (Hosted by FirstMark)

Haoyuan Li · 18m spoken Matt Turck · 42s 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

In this DataDrivenNYC presentation hosted by FirstMark, Alluxio founder and CEO Haoyuan Li introduces Alluxio, an open-source virtual distributed storage system operating at memory speed. He explains how Alluxio decouples compute and storage to solve big data integration challenges, showcasing real-world enterprise case studies and the project's roots at UC Berkeley AMPLab.

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 3.5% of the talking time here. How this is scored →

Matt as informed peer 0.5 Guest teaching 0.5 Guest disagreement 0.3 Matt pushing back 0.5
05100:0010:0020:001:17–4:04 · Matt as informed peer 0/10 Alluxio Definition and Project Background This segment is an uninterrupted solo presentation by Haoyuan Li introducing Alluxio and its open-source lineage. Since the host does not participate, all host-related scores are zero.4:04–11:05 · Matt as informed peer 0/10 Hardware Trends in Memory Performance and Pricing Haoyuan Li continues his slide deck presentation on DRAM pricing trends and data duplication bottlenecks. Without host involvement, host and interaction scores remain zero.11:05–14:55 · Matt as informed peer 0/10 Enterprise Case Studies and Talk Conclusion The guest details enterprise case studies at Baidu and Barclays to conclude his formal talk. Because this remains a monologue, all host scores are set to zero.14:55–23:03 · Matt as informed peer 2/10 Fireside Q&A Session with Matt Turck Matt Turck asks friendly Q&A questions about AMPLab culture and monetization before opening the floor to audience questions. The dynamic remains casual and collaborative, with the guest declining to discuss revenue details while in stealth mode.1:17–4:04 · Guest teaching 0/10 Alluxio Definition and Project Background This segment is an uninterrupted solo presentation by Haoyuan Li introducing Alluxio and its open-source lineage. Since the host does not participate, all host-related scores are zero.4:04–11:05 · Guest teaching 0/10 Hardware Trends in Memory Performance and Pricing Haoyuan Li continues his slide deck presentation on DRAM pricing trends and data duplication bottlenecks. Without host involvement, host and interaction scores remain zero.11:05–14:55 · Guest teaching 0/10 Enterprise Case Studies and Talk Conclusion The guest details enterprise case studies at Baidu and Barclays to conclude his formal talk. Because this remains a monologue, all host scores are set to zero.14:55–23:03 · Guest teaching 2/10 Fireside Q&A Session with Matt Turck Matt Turck asks friendly Q&A questions about AMPLab culture and monetization before opening the floor to audience questions. The dynamic remains casual and collaborative, with the guest declining to discuss revenue details while in stealth mode.1:17–4:04 · Guest disagreement 0/10 Alluxio Definition and Project Background This segment is an uninterrupted solo presentation by Haoyuan Li introducing Alluxio and its open-source lineage. Since the host does not participate, all host-related scores are zero.4:04–11:05 · Guest disagreement 0/10 Hardware Trends in Memory Performance and Pricing Haoyuan Li continues his slide deck presentation on DRAM pricing trends and data duplication bottlenecks. Without host involvement, host and interaction scores remain zero.11:05–14:55 · Guest disagreement 0/10 Enterprise Case Studies and Talk Conclusion The guest details enterprise case studies at Baidu and Barclays to conclude his formal talk. Because this remains a monologue, all host scores are set to zero.14:55–23:03 · Guest disagreement 1/10 Fireside Q&A Session with Matt Turck Matt Turck asks friendly Q&A questions about AMPLab culture and monetization before opening the floor to audience questions. The dynamic remains casual and collaborative, with the guest declining to discuss revenue details while in stealth mode.1:17–4:04 · Matt pushing back 0/10 Alluxio Definition and Project Background This segment is an uninterrupted solo presentation by Haoyuan Li introducing Alluxio and its open-source lineage. Since the host does not participate, all host-related scores are zero.4:04–11:05 · Matt pushing back 0/10 Hardware Trends in Memory Performance and Pricing Haoyuan Li continues his slide deck presentation on DRAM pricing trends and data duplication bottlenecks. Without host involvement, host and interaction scores remain zero.11:05–14:55 · Matt pushing back 0/10 Enterprise Case Studies and Talk Conclusion The guest details enterprise case studies at Baidu and Barclays to conclude his formal talk. Because this remains a monologue, all host scores are set to zero.14:55–23:03 · Matt pushing back 2/10 Fireside Q&A Session with Matt Turck Matt Turck asks friendly Q&A questions about AMPLab culture and monetization before opening the floor to audience questions. The dynamic remains casual and collaborative, with the guest declining to discuss revenue details while in stealth mode.

speaking balance: gold is Matt, purple is the guest (3 minute bins)

0:00 · Matt 0% · guest 100%0:00 · Matt 0% · guest 100%3:00 · Matt 0% · guest 100%3:00 · Matt 0% · guest 100%6:00 · Matt 0% · guest 100%6:00 · Matt 0% · guest 100%9:00 · Matt 0% · guest 100%9:00 · Matt 0% · guest 100%12:00 · Matt 0.5% · guest 99.5%12:00 · Matt 0.5% · guest 99.5%15:00 · Matt 16% · guest 84%15:00 · Matt 16% · guest 84%18:00 · Matt 7.4% · guest 92.6%18:00 · Matt 7.4% · guest 92.6%21:00 · Matt 4.8% · guest 95.2%21:00 · Matt 4.8% · guest 95.2%
Sharpest disagreement ▶ 18:13 Stealth Mode Deflection on Revenue

Haoyuan Li politely deflects the host's direct question about monetization by stating they are still in stealth mode and do not talk about revenue publicly.

Hardest push from Matt ▶ 18:08 New York Question on Revenue

Matt Turck presses the guest on business strategy by framing a direct question about how the company intends to make money.

Biggest teaching moment ▶ 20:45 Explaining Virtual Storage vs Storage Connectors

Haoyuan Li educates an audience member on why creating a middle virtualization layer provides greater architectural utility than building individual ecosystem connectors.

Matt holds his own ▶ 14:59 Citing Berkeley AMPLab Alumni

Matt Turck demonstrates industry domain familiarity by listing specific successful companies and founders that originated from UC Berkeley's AMPLab.

the scores for every segment, with the reasoning behind each
ChapterTopicMatt as informed peerGuest teachingGuest disagreementMatt pushing backWhy
Alluxio Definition and Project Background 0000 This segment is an uninterrupted solo presentation by Haoyuan Li introducing Alluxio and its open-source lineage. Since the host does not participate, all host-related scores are zero.
Hardware Trends in Memory Performance and Pricing 0000 Haoyuan Li continues his slide deck presentation on DRAM pricing trends and data duplication bottlenecks. Without host involvement, host and interaction scores remain zero.
Enterprise Case Studies and Talk Conclusion 0000 The guest details enterprise case studies at Baidu and Barclays to conclude his formal talk. Because this remains a monologue, all host scores are set to zero.
Fireside Q&A Session with Matt Turck 2212 Matt Turck asks friendly Q&A questions about AMPLab culture and monetization before opening the floor to audience questions. The dynamic remains casual and collaborative, with the guest declining to discuss revenue details while in stealth mode.

Statements from this episode (9)

Assertion Not checkable as stated
Alluxio was deployed at over 100 companies by April 2016
“And it's deployed at more than a hundred companies nowadays.”
Haoyuan Li Apr 13, 2016 ▶ 2:08
Assertion Supported
Alluxio open-source project reaches nearly 250 contributors across 50 organizations
“We have close to 250 contributors to the developers contributing to this project. It's about three times growth over the last year, and we have over 50 organizations are contributing to this project as well.”
Haoyuan Li Apr 13, 2016 ▶ 2:31
Assertion Partly supported
DRAM costs decrease by 50% every 18 months
“Every 18 months, the DRAM cost decreased about 50%.”
Haoyuan Li Apr 13, 2016 ▶ 4:42
Insight
Decoupling storage and compute enables independent resource scaling
“You can separate your storage cluster as well as your compute cluster, which enable you to scale your compute cluster as well as your storage cluster independently.”
Haoyuan Li Apr 13, 2016 ▶ 10:05
Assertion Not checkable as stated
Baidu queries data 30 times faster using Alluxio
“They're coring data 30 times faster with Aluxio.”
Haoyuan Li Apr 13, 2016 ▶ 11:19
Assertion Not checkable as stated
Baidu manages over two petabytes in memory across 200 nodes using Alluxio
“They manage more than two, 200 nodes in this deployment with more than two petabytes of memory managed space inside Alluxio. And this has been in production for more than a year.”
Haoyuan Li Apr 13, 2016 ▶ 11:39
Assertion Not checkable as stated
Barclays accelerated Spark jobs from hours to seconds using Alluxio
“They used Tachyon to accelerate Spark jobs from hours to seconds”
Haoyuan Li Apr 13, 2016 ▶ 12:03
Assertion Not checkable as stated
Impala on Alluxio over Amazon S3 yields a 15x performance gain
“And there's another one, you run Impala on top of Aluxio, Aluxio on top of Amazon S III. They have a 15 times performance improvement in this case.”
Haoyuan Li Apr 13, 2016 ▶ 13:31
Assertion Supported
Databricks founding CEO Ion Stoica was Haoyuan Li's PhD co-advisor
“Yang, who, who's the founding CEO of Databricks, he's my PhD co-advisor.”
Haoyuan Li Apr 13, 2016 ▶ 20:25
Made with StarZero

Turn any episode into a week of clips.

This entire site, over 400 conversations transcribed, diarized, checked and made playable, runs on the StarZero media pipeline. Drop in your own episode and the podcast clipper finds the moments worth sharing, cuts them, captions them, and reframes them for every feed.