YT YouTube2Text

YouTube2Text → Transcripts → Flexnodes Explained: Decentralizing Blockchain Performance with Muriel Médard & Sajida Zouarhi

Flexnodes Explained: Decentralizing Blockchain Performance with Muriel Médard & Sajida Zouarhi — Transcript

by Optimum · 3,849 words · 544 segments · language en · Watch on YouTube

Open in transcriber Summarize with AI Download SRT/VTT

Full transcript

0:05Hello. Hello. Welcome to another one of
0:07our Optimum chats uh where we have
0:10discussions with uh members of the
0:13Optimum team, sometimes uh folks from
0:16the outside, advisers, partners and so
0:18on. Uh my name is Muriel Medard, Muriel
0:22Midar. I'm co-founder CEO and I'm really
0:25happy today uh to be joined by Sashida.
0:28Sashida, would you like to introduce
0:30yourself?
0:31>> Sure. Um, hi everyone. So, I'm Sashid
0:35Desari. I work at Optimum as chief
0:38product officer. Uh, been in the web 3
0:40space for the last 10 years. Uh, from
0:43working on the merge at consensus or
0:45more recently, MEV and block building uh
0:49at block native. Um, very uh very much
0:53looking forward to this conversation.
0:55>> Yes, thank you so much. And today's
0:58topic is flex. Uh you know what are
1:01they? What do they do? Uh what is the
1:04future? How do they fit into the whole
1:07optimum story? Uh and that's uh that's
1:10our discussion today. So
1:13maybe we can start a little bit with the
1:16optimum network uh and you know our
1:18first product basically data
1:20acceleration network. Um so you know
1:24just to give a little bit of background
1:26again particularly for people who might
1:28be new to optimum um what we're doing is
1:31this data acceleration network that
1:33accelerates data propagation to start
1:35with um we use our technology randomly
1:39network coding of which I'm co-inventor
1:41you know developed in my lab at MIT um
1:44where I've been working in this area for
1:47many many years as a professor and now
1:50uh doing the tech transfer with optimum
1:54and uh I wanted to hear from you sidak
1:58you know as somebody who's really
2:00focused on product uh somebody who has
2:03so much background perspective expertise
2:07uh in web 3 how you describe the optimum
2:11network
2:12>> right uh thanks Muriel so I think um
2:17optimum network is a bit uh a hybrid
2:21because it's uh mostly like a
2:24performance layer that connects several
2:26uh chains and actors of those chains in
2:29order to grant them better performance,
2:32faster access uh to data and some more
2:34features that they might need such as
2:36data availability. And so in that way
2:39it's like connective tissue between
2:41several blockchains without being one uh
2:45in itself. And so I think for a web3
2:48project that's uh kind of an interesting
2:50positioning. Uh what is also interesting
2:53is that it does have some blockchain
2:55properties because we do have notions of
2:57consensus uh and some security
3:00principles that we are enforcing at the
3:02level of the optimum network uh and also
3:04are used for some other things that we
3:07are going to do down the line such as
3:09decentralized memory. So from a
3:11technological perspective but also
3:13economical perspective it is a very uh
3:16interesting product that bridges those
3:18two world from the infra networking side
3:21to the blockchain web3 side.
3:24Yeah, absolutely. And and I think you
3:26you bring that to for that sort of
3:29uniqueness of the fact that you know we
3:31we often say scale any chain. Um, and to
3:35some extent it's almost like web 3 is
3:37still in the in the sort of early
3:42time of computers where you would have
3:46that the computer and the operating
3:50system and everything was just one
3:51monolithic product. And of course
3:54eventually you go hey you know there's
3:56best practices that can be used across
3:58different um let's say VMs chains L1
4:03whatever you want to call them
4:05>> uh and you know what we're doing is
4:08really we're doing the whole data flow
4:11the whole data movement uh which
4:14includes the propagation the memory etc.
4:17And as you said on the memory there has
4:19to be of course consensus but
4:20interestingly enough on the propagation
4:22you can actually be very strongly
4:26resistant to
4:29uh Byzantine attacks or to um uh you
4:34know any any sort of censorship attempts
4:38while not having consensus. Um so yeah
4:41so so that's basically what we do right
4:43we do the the the we we do the data uh
4:46the data flow the data management
4:48effectively for everybody um and then we
4:52allow these these chains to do what they
4:54do best which is be chains you know
4:57rather than
4:59uh rather than having to to manage
5:01everything end to end uh themselves. Um,
5:05and you know, I you we mentioned a
5:08little bit uh how we do it. You know, I
5:10mentioned RLNC.
5:12Um,
5:14I'd like to hear how you describe RLNC
5:16in general. Sajida, I know, you know,
5:18I'm I'm always interested in in hearing
5:21different members of the group always
5:23have such really fantastic perspectives
5:27and very insightful ways. So I I want to
5:30hear your way just
5:31>> Mariel I'd love to pitch to you.
5:35>> Please please pitch it.
5:37>> So is a great technique Mariel as you
5:39know. Um so the idea here is that um how
5:43can you um make uh best use of bandwidth
5:48of a node and so RNC actually helps us
5:51do that by breaking the data into pieces
5:54and encoding it. Each node actually on
5:57the propagation pass contribute to that
5:59because it actually regenerates new
6:02equations and innovative pieces. And so
6:06what is uh quite robust and resilient
6:09here is that if you gather a certain
6:11number of those uh pieces, those shards,
6:14you are able to retrieve the original
6:16message. And so that is a interesting
6:18design especially for distributed and
6:20decentralized network so that any node
6:23can participate uh and and gossip the
6:25information uh and we're not relying on
6:28any source or any uh path for the data
6:31propagation for the message arrival and
6:35the sort of successful transmission to
6:37happen.
6:39>> Wow. Okay. You saw me on it. That was a
6:41really good pitch. Yeah. And I think you
6:44brought up a very important aspect you
6:47know which I had mentioned a little bit
6:48with respect to um you know the
6:50censorship resistance uh which is basic
6:53resilience. It is because you know
6:55you're not you're not really depending
6:57on anybody right uh you you you have
7:00this dependence on a on a collective
7:03rather than uh a dedicated a dedicated
7:06path which actually brings us to the
7:08flex nodes right because what do we mean
7:10by not depending on anybody it's going
7:11to the flex nodes
7:13>> um
7:15and um you know we we get asked a lot
7:17what do you call it a flex node you know
7:19I like to say well you know it can be
7:21big it can be small and it's strong
7:23you're flexing Yeah. Um but you know
7:26this is now really when you look at it
7:29as a product um wanted to to hear you
7:32know h how what your thinking is around
7:35flex notes.
7:36>> Yeah for me it's like the the atom of uh
7:40of the network. uh it's the the basic
7:42component that creates the overall
7:45optimum network and uh what is
7:48interesting is that they come in many
7:49form uh whether it's in terms of
7:51resource they can contribute to the
7:53network uh services they could support
7:56uh we're talking about propagation today
7:58but uh we intend to add more um chains
8:02they could be connected to uh one could
8:05be helping propagation for Ethereum as
8:07it could be helping for another chain
8:09and so there is a lot
8:11versatility in a in a flex node and its
8:13role. Uh but also from an audience
8:15perspective because you could be a you
8:17know sophisticated node operator wanting
8:20to run a flex node. Uh or you could be
8:23uh you know someone at home interested
8:25in participating and uh you know
8:28bringing some resources to the service
8:30of uh Optimum and its users uh with like
8:34uh a simple device, a laptop uh maybe
8:36even a smartphone. Uh so yeah this is
8:39how I see the sort of like flex node
8:42nebula worldwide completely
8:44decentralized no single point of failure
8:47everyone contributing and uh providing a
8:50service in the end to users.
8:52>> I love Nebula. I think we should uh we
8:55should uh reuse that uh that term
8:58because it is it's
9:01the atoms in the nebula. I think that
9:03that's that's very poetic and very apt.
9:06Yeah. No, I I I completely agree. And
9:08you know, I I think in particular when
9:10we think of um people just regular
9:13people with just a regular computer,
9:15regular internet connection running a
9:17flex node, you know, the idea is also
9:20they're helping
9:22in many cases the local community,
9:23right? And every little bit helps. Uh so
9:27it's not like, well, you know, I'm
9:28running a flex node, but you know,
9:31everybody's going to win out because
9:32it's a competition. it it's not, you
9:35know, if if I only provide two
9:37equations. Sure, I only get paid to for
9:39two equations, but you actually get paid
9:40for what you did. Uh as opposed to
9:43having to be in a race with the big guys
9:45and you're never going to win.
9:47>> And you're sort of unique as well, like
9:49uh no two flex are alike. If if someone
9:52spins a flex node in a region where
9:54there is a you know low supply uh the
9:58equation that they're going to
9:59contribute uh from the perspective of
10:01the local peers are going to be very
10:04valuable. So everyone has this chance to
10:07meaningfully contribute uh to the
10:09network.
10:10>> That's right. That's right. It's like
10:12your corner store, you know, they they
10:15they really help you out. They're right
10:17where you are and uh and they're giving
10:19you what you need. I I completely agree
10:22and and I think um you know the other
10:24thing about flex nodes versus for
10:26instance running a validator right a
10:27validator is big investment
10:31um you also need to have a lot of uptime
10:34which can be really hard uh here with
10:37the flex node you know you get paid for
10:40what you do but you don't get
10:43you know punished for what you didn't do
10:46uh so as long as you didn't do something
10:49wrong uh you know on purpose in which
10:51case you get punished you. So, so I
10:53think also from a philosophical
10:54perspective that that also goes into the
10:56flexibility, right? It's like it's as
10:59big as you want to be, as small as you
11:01want to be, and it's also on when you
11:02can do it, you know, it's like whenever
11:04whenever it works for you.
11:07>> So, yeah, that's uh but I I'm going to
11:10remember the remember the nebula. Um I
11:13think maybe at this point it might be
11:16good for us to transition and say, well,
11:18what do they do? Right? We say they
11:20exchange equations. Right? Now they do
11:22propagation then the road. Um we're
11:25going to add basically not just data in
11:27flight but data at rest which is memory
11:30uh memory that can be updated like say
11:32random access memory on your computer.
11:35Uh but what does it look like? like you
11:37know the do you get a you know do you
11:40get a performance boost or do you only
11:42just get you know this decentralization
11:46and very often also people think that
11:47decentralization and performance are
11:49sort of at odds with each other you know
11:52actually our first uh uh discussion of
11:56this type was with Nicholas and Kishori
11:58about the trilmma right where we try to
12:00deconstruct the trilmma um so saj do you
12:04want do you want to show us what it
12:06looks like.
12:07>> Sure.
12:08>> Thank you.
12:08>> So performance. So there are many
12:11aspects to to that obviously, but I can
12:15sort of go through one.
12:17>> Yeah.
12:25>> Okay. Can you see my screen?
12:35Yeah, I can see it.
12:36>> Okay, great. So what we can see here is
12:40uh on the top right uh me P2P protocol
12:44which is the protocol that we developed
12:46at optimum based on air valency uh
12:50propagating block payload on ethereum
12:53throughout the network over 40 plus
12:57location all around the world and it's
12:59able to do that um in 300 milliseconds
13:03uh at the moment and this is not uh how
13:07can I Today this is not the most uh
13:10intense uh benchmark data that we can
13:14achieve actually. So this is a pretty
13:15conservative figure.
13:16>> Well, it's actually it's real time,
13:18right? You just went on it and this is
13:19what it gives you right now. We we
13:21didn't cherry pick. We just went right
13:24now and and picked at a random point.
13:26>> Exactly. And we can see there is some
13:27noise here. So So as you can see that
13:29that sort of skew the average. And so uh
13:32here you can see uh the same thing being
13:36measured but on the lip P2P pass which
13:38is the gossip protocol of Ethereum and
13:41you can see that we're more on the order
13:43of uh seconds 2.4 seconds. So yeah
13:47that's a big gap and that is also a sort
13:51of testimony to what uh you know
13:53incorporating good coding techniques
13:56into the P2P layer of a blockchain can
13:59help achieve. So we often talk about
14:01accelerating,
14:03scaling, increasing the performance etc.
14:05of blockchain. Uh one key aspect of that
14:07could just be uh improving on the
14:10peer-to-peer layer. Um so yeah.
14:14>> Yeah. No, thank you. And uh you know, as
14:17I mentioned to the team often, I I love
14:19looking at our our dashboard which you
14:23just shared with us, Sida, and you know,
14:25seeing all the green and I think you
14:26said that's like touching grass, you
14:28know. his endorphins. It's all green.
14:31This is so nice. Um, and just to maybe
14:35highlight that was on hoodie test net.
14:37So, it's um, you know, it's not an
14:38internal test net. It's it's a real um,
14:43you know, bonafide
14:44>> in the wild.
14:45>> It's in the wild. That's right. It's
14:47it's it's it's not uh it's not grass
14:49that I grew in my kitchen. It's it's out
14:53out in the wild. It's in the field. Um
14:57and uh and of course you know we're
14:59going to be going to to mainet in just a
15:02a couple months or so. So um you know we
15:07talked a little bit about okay um
15:12roughly what is a flex node
15:15what it supports you know at this point
15:17it supports our data acceleration which
15:19we call m P2P uh we showed that it gives
15:24you a huge benefit this month P2P this
15:27you know network of flex nodes we talked
15:30a little bit about who can run a flex
15:32node Um but may maybe maybe we can talk
15:36a little bit more about the the
15:39requirements
15:40from just a product point of view. Um
15:44>> yes
15:44>> of you know what it takes to connect
15:48here.
15:49>> Yes. So I mean we we touched on this the
15:53the flex uh side uh sort of trickles
15:56down to even the experience the operator
16:00experience itself where depending on the
16:04you know technical level in terms of
16:08skills or uh resources uh in terms of
16:12bandwidth uh and just ability to run it
16:15with like pretty good uptime or more
16:18opportunistically.
16:19um who can run a flex node may vary uh a
16:23lot. So in terms of resource requirement
16:27uh based on what we know currently it is
16:29pretty uh lightweight. Uh so
16:34we assume people are going to be able to
16:36uh run it on their laptop. Uh we're
16:39going to push it so that it's you know
16:41uh even doable on uh a smartphone. That
16:44would be uh ideal. Uh and then in terms
16:47of skill set, we have I think we can uh
16:49mention that Marielle uh we have our
16:51plan of running an actual program. Uh
16:54something that uh helps people learn the
16:57basic concepts that they need to know in
16:59order to uh run a flex node in the way
17:01that they see fit for their you know uh
17:04day-to-day and also the goals that they
17:06may have uh financially. Uh and so
17:10whatever you put uh in this is what
17:12you're going to get basically and that
17:14uh maps even to how we see the economics
17:16of a flex node. Uh whatever you
17:18contribute to the system is uh the basis
17:21on how you're re rewarded at the end of
17:23the day. Um yeah. Oh, absolutely. And
17:27and thank you for mentioning uh our
17:29plans to have these basically flex node
17:32training, you know, which will be open
17:35to anybody who wants to to run a flex
17:38node. Uh but the idea is it should be
17:40very very light. It's mostly so that you
17:42know walking people through just how do
17:44you install it, you know, and right and
17:46then just watching it run. Um so no,
17:49that's uh that that's really great. And
17:50I I I wanted to also um talk about the
17:54the the aspect about the financial
17:56goals. You know, one of the things is
17:58the more flex nodes there are, the more
18:02chains we can put, the more networks uh
18:05and not just chains, but also different
18:09participants in chains. Like right now,
18:12um we have really concentrated
18:15to start with with validators on
18:17Ethereum.
18:18Well, you were showing our our hoodie
18:20test net. It's basically that's who it's
18:22geared towards. Uh but we've also
18:25started working with block builders um
18:29who also can really take advantage of
18:34acceleration to benefit their
18:36operations. Uh and then we'll go into
18:38other chains right as as we mentioned.
18:40So of course you know more flex nodes
18:42means that more chains and also more
18:46participants in different chains can
18:49join and get advantage and then that
18:51leads to better value which then
18:53enhances the economics for folks and so
18:56people may choose to you know maybe make
18:58their flex node bigger dedicate more
19:01resources or maybe run more flex nodes
19:03or new people may start running flex
19:05nodes. So what we see is a is a virtuous
19:08cycle where there's demand that's coming
19:11because of value that's accured to
19:13different L1s by different players in
19:16these L1s and then as that value is
19:18accured it sort of creates a natural
19:20pull um for more people to to
19:24participate whether it's participants
19:25participating more actively or new
19:28participants entering.
19:30>> Yeah. Yeah. I really like this uh this
19:33way of mapping it because it shows that
19:35there are many strategies also people
19:37can adopt. Uh so it can be um for
19:40example deciding to opt in on
19:43propagation topics that are related to
19:46an ecosystem or another or multiple and
19:49you could even adapt it depending on the
19:51demand uh at a certain point in time uh
19:54on those propagation path and then uh
19:57it's sort of multiplied by all of the
19:59places in the world that you could be
20:01running a node whether it's at your
20:03location or in a distant place and so
20:05all of that creates sort of a multi-
20:08factor uh you know setup uh by which you
20:12can try to optimize uh you know
20:15basically the contribution impact that
20:16you can have on the network which uh you
20:19know relates to the rewards as well. So
20:21I think that that's going to be a fun
20:23thing for many people that are oriented
20:26uh for strategy optimization to to play
20:28with.
20:29>> That's right. And you know it could also
20:32be that you want to prioritize your
20:33neighborhood, your community or you know
20:36as you said uh a particular ecosystem
20:38that you you believe in and you want to
20:41really help out and accelerate. Um no
20:44absolutely. So um just you know to to
20:48try to kind of pull it together how how
20:52do we feel we're going to measure the
20:54success of flex notes you know just to
20:56to cap tap cap conversation.
20:59Um
21:01you know to to to me the success is if
21:03people are using it uh and people are
21:08benefiting from it from the point of
21:10view of you know
21:14the validators the the block builders
21:17the me searchers the users in the end um
21:21through lower gas fees you know faster
21:25uh service for whatever they're trying
21:26to do and on the other side if our flex
21:30node operators are seeing the rewards
21:32and they go this is great you know I I'm
21:35make you know I'm getting rewarded and I
21:37feel good about what I'm doing I'm
21:38actually providing a service that that
21:40people find uh you know intrinsically
21:43valuable so that's how I say it but you
21:46know love to hear your your takes jeta
21:50>> yeah so I think for me it's um it's
21:53going to be how we impact the the user
21:56experience in the end because most of
21:58the direct user of optimum and and flex
22:02nodes are going to be you know protocols
22:05or maybe DAPs and in the end uh the
22:09value is you know for the end user and
22:12right now the big focus in blockchain is
22:14how do we get to this level of maturity
22:16where everything is like you know sub
22:19200 millisecond and we're able to have
22:21the greatest uh user experience in web 3
22:25such as what we experience uh usually in
22:27web 2 from a um a latency perspective
22:31and so if optimum is able to impact that
22:34uh move it in the right direction that
22:35would be a great success. Uh from a flex
22:39node operator perspective I would say um
22:42how much uh how much people are getting
22:45back from their contribution is going to
22:47be a big aspect. Uh it would be easy to
22:51create this type of service and just do
22:53it in a centralized way. uh it would you
22:55know be be uh be fast uh and be
22:59performant but that's not the idea. The
23:01idea is that we support and enable uh
23:05blockchains and decentralized systems
23:07and protocol in general grow and there
23:10is a big component of that which is the
23:12community. So there is a model here that
23:14we're also trying to improve which is uh
23:16can we create a sustainable marketplace
23:18here uh where people contribute and have
23:22an incentive to keep contributing that
23:24is matching also uh the demand on the
23:28other side who's actually seeing uh
23:30benefits from that latency uh or
23:32otherwise improvements uh could be also
23:34on on the memory aspect. So that's going
23:37to be a pretty exciting uh sort of
23:39experiment uh to see unfold in the next
23:41coming months.
23:43>> Yes. And we'll be doing that together.
23:45So uh looking forward to it. Thank you
23:47so much Sida. This was a a fun
23:50conversation. Uh I hope that uh people
23:53who look at this online also find it
23:56interesting and engaging. And uh yeah,
24:00we'll be having more of these. Thank you
24:01so much.
24:02>> Thank you.

About this transcript

This page contains the full transcript of Flexnodes Explained: Decentralizing Blockchain Performance with Muriel Médard & Sajida Zouarhi by Optimum, generated from the public captions YouTube serves with the video. The transcript has 3,849 words across 544 segments, with the original timestamps preserved so you can click any line to jump to that moment in the embedded player.

What you can do with it

Use the transcript to take notes, quote the speaker, build a study guide, generate a summary with ChatGPT or Claude via the YouTube Summary tool, or export it as a timed subtitle file with YouTube to SRT. You can also re-open it in the transcriber to translate the transcript into 100+ languages.

Free YouTube transcript tool

YouTube2Text is a free YouTube transcript generator — no signup, no daily limit. Paste any YouTube link and get the full transcript instantly, with timestamps, click-to-jump, translation to 100+ languages, AI prompts for ChatGPT, Claude, and Gemini, and exports to TXT, SRT, VTT, or Markdown.