Algora | Algora TV

Open source Interviews #29 - Jove Zhong, co-founder of Timeplus.com

Algora

Watching now

00:00 Hey everyone, welcome to the Open Source Founder Podcast.

00:04 Joining me today is Jov, co-founder and head of product at TimePlus,

00:09 a company that recently open sourced Proton.

00:12 I'm very excited to talk about this.

00:15 Jov, thanks so much for joining.

00:16 And let's kick it off with some information about your personal background

00:20 and then how TimePlus gets started.

00:26 My name is Jov John. I'm based in Vancouver. I'm one of the co-founders for TimePlus.

00:31 And we have been working on the next generation streaming processing engine to make the data

00:38 engineer and the platform engineer much easier to process their data. And at a high level,

00:43 this is really a new way to

00:46 process data without worry too much regarding tuning JVM and set up a cluster

00:51 this is really a new way to process data in a much efficient way worry about less

00:57 infrastructure stuff yeah I'm happy to discuss more about this space and do some demo later

01:04 yeah perfect yeah we'd love to hear a little more about this space and do some demo later.

01:11 Perfect. Yeah, I would love to hear a little more about how you experienced this problem throughout your career so far,

01:16 and then all the way to actually starting your open source motion with Proton.

01:20 I think it's been a few months of the first release.

01:22 So how do you experience this problem before?

01:28 Yeah, of course. So actually, I'm in this space for over 20 years.

01:32 I work for a big company like IBM for four and a half years.

01:40 I work for EMC Documenting, which is a content management enterprise search product for the other four and a half years.

01:46 I work for a company called Spunk, which is the very first kind of big data company that gets IPO and the focus a lot on the log analysis, availability, security.

01:53 In the company I worked for them seven year and a half. So even I already, this is my 21, 22

02:07 This is my 21, 22 year of working, but I only work for not a lot of companies.

02:14 But I do see a lot of opportunities slash frustration and opportunity.

02:15 There's more and more data.

02:18 In the past, we talked a lot about big data.

02:23 So that's how you get familiar with Hadoop, HDFS, MapReduce.

02:26 But not many people talk about big data right now because there's a lot of live data, stuff like Kafka, like Repanda, Pulsar, and

02:32 in the recent one or two years, it is not a large data model.

02:40 Even this week in CES, people're talking about a large action model.

02:46 So there's a lot of things going on.

02:47 But essentially, people needed to absorb much bigger and immediate data in a very efficient way.

02:57 But underlying technology doesn't really give you a good foundation.

03:03 You have to learn a lot of open source knowledge.

03:06 You have to learn the programming language.

03:08 You have set up big cluster.

03:10 There's a lot of things to work on.

03:12 For example, my last two years in Spark,

03:16 we did leverage stuff like Quintus,

03:19 we leveraged Apache Flink, leveraged Kafka,

03:22 leveraged Kinesis.

03:24 It's true, there's many components in that project to set up a large-scale, multi-talent

03:29 data pipeline.

03:32 And it is not cheap.

03:34 And we have almost 100 engineers.

03:36 Again, we do a very interesting product.

03:39 We process sometimes over multiple data a day.

03:44 But in that kind of a scale,

03:46 this entire stack is very complex.

03:49 We make the work for sure,

03:50 but we feel like there can be a different way,

03:53 an easier way.

03:54 So that's why some of us coming from background,

03:59 thinking that is, hey, if this processing data

04:03 in the emulsion

04:06 to reduce sending too many garbage

04:09 or some low-level data to the destination,

04:11 can we have a system in the middle

04:13 to press data in more efficient way?

04:16 And is the Apache Flink the only option?

04:20 So we ask those questions

04:21 and we take a very aggressive approach,

04:24 that is that we build a new engine.

04:27 And this is based on CPI-Path, not based on Java.

04:30 And this gives us a very independent tool.

04:36 That is, you don't have to worry too much about other components.

04:38 You just set up either a single node or cluster,

04:41 very efficiently leverage the latest Linux or OS by system

04:47 level features such as OURAIN that you're aiming for the lowest possible infrastructure

04:55 overhead, but also you gain the best performance.

05:00 And yeah, we work on that.

05:03 And also we leverage Clearhouse because there are, I guess,

05:06 maybe one or two thousand different SQL functions

05:09 in Clearhouse.

05:11 That database is very solid.

05:13 So we leverage Clearhouse

05:15 as kind of a SQL engine

05:18 mainly focusing on the historical side,

05:21 but we implemented something new

05:22 for the streaming processing

05:24 and also try to unify them together so that you don't have to set up historical site, but we implemented something new for the streaming processing and

05:28 also try to unify them together so that you don't have to set up

05:33 a site. 10 Plus and K-Plus actually have a single

05:37 binary. So the journey that is, we start from

05:44 this unified approach and we have our first cloud release, I think almost one year and a half ago.

05:49 And we also have some on-prem customers,

05:54 but in the same time,

05:56 we feel like this is such an amazing product.

05:58 We do want to open to more users,

06:01 especially for the whole source community.

06:03 So three months ago, we made a decision to have our core engine open source under Apache

06:10 License v2, which is a very friendly license.

06:13 And we started to really get more feedback from users, even if they don't really want

06:18 to make a decision to purchase our cloud right now.

06:22 Thank you so much for this background.

06:25 Certainly makes a lot of sense.

06:27 I understand that it's difficult infrastructure to work with

06:30 and very expensive for organizations

06:32 and these real-time streaming data needs

06:37 as well as analyzing historical data,

06:39 putting everything together in real time.

06:41 The applications are in every industry.

06:43 My understanding is in fintech, for example,

06:46 or with Internet of Things.

06:49 DevOps, there's a real pain point there.

06:53 And now coming to today,

06:56 if you could talk a little more about the decision to open source,

07:00 how it influences distribution,

07:03 as well as the ease of new people to play with

07:06 the technology, all the way to actually enterprises adopting it.

07:10 I'm curious how you're mapping out that process for your product and for Proton and what it's

07:16 looked like today in terms of the adoption, the conversations you're having and monetization

07:21 as well, how it can help the company be sustainable.

07:25 Yeah, yeah, yeah.

07:26 That's actually one of the kind of fun part

07:30 and also I'm not saying that dark part,

07:33 but it's also some challenging part

07:34 because in the past,

07:35 I was always in the role such as engineer,

07:38 engineering lead,

07:39 engineering manager, director,

07:40 and I don't have to worry too much

07:42 regarding the go-to-matic strategy,

07:44 how to do the customer

07:45 call, all those things, and also pricing.

07:49 But again, many of us in this 10-plus company, we are coming from a technical background,

07:55 but we really learn things fast.

07:57 Maybe we make some mistakes, but we try to improve ourselves and adapt to what we need

08:03 to do. So I guess the reason why we stopped our company

08:10 for the first one or two years really has

08:12 to be the product and the don't,

08:14 we discussed open source strategy a little bit,

08:17 but we don't really make the open source

08:18 at the very beginning.

08:19 Part of the reason is that we know we are working

08:23 on a very challenging problem.

08:34 And it is not easy to suddenly do something amazing to some other project or some commercial product which has been there for 10 or 20 years. For example, in the fintech space, you may know that there's a company called KX.

08:39 The product is called KDP, I think.

08:42 It is very expensive software.

08:44 I guess they don't really support cluster.

08:48 But yeah, it is already widely adopted by mini bank or like a hash fund.

08:55 And they also use a very special language called Q.

08:58 So they do give you a very low latency, high performance.

09:04 But people who spent maybe two years learning it don't

09:08 want to turn to other projects because that kind of learning experience becomes their

09:14 assets.

09:15 Just like, I'm not sure how many people right now in the world are still using COBAR.

09:20 So those programmers, maybe they can be very rich because they are having a very unique skill.

09:26 So it's very interesting. But my point is, there's a lot of mature players in this space, like KDB in the fintech area. and for example, Apache Flink, Apache Spark,

09:45 or even KsqlDB, they are already been there

09:49 space for a long time, but I do see a lot of campaign,

09:52 for example, for the KsqlDB, and the future

09:55 of that project is not very clear,

09:57 especially one confounds put so much effort,

10:00 money to build the Flink call service and the solutions.

10:07 So we know this is not an easy project.

10:13 I know that some projects today people just wrap the open API to have a new product to

10:20 like release the product within two or three weeks.

10:22 It's not the case for us.

10:23 So we have a lot of database engineer, we do our talent or cloud services, we even have a UI which I'll demonstrate

10:31 later that is really wanting to give a new experience to the developer that is very fast,

10:38 but it's very easy to use. So there's a lot of stuff to do, so we don't worry too much about

10:43 open source at the beginning.

10:46 Actually also because we're coming from the background of Sprung.

10:50 You know, Sprung is not really an open source product. It's very successful.

10:54 And also one of the key reasons that is Sprung, they have very nice infrastructure written into the class,

11:03 but under the cover,

11:05 they have very nice UI and they have dashboard,

11:09 they have out-of-box solution for security,

11:12 for monitoring.

11:15 So you don't really have to be experts for

11:18 SQL or the programming just to use the product out-of-box.

11:21 So we are in some degree we're really

11:24 influenced by Spark.

11:25 And also other products like Snowflake is not open source.

11:28 So there's a lot of debate.

11:30 So somehow we made the decision that is,

11:32 we really needed to spend the first one or two years

11:34 really have the product very ready.

11:37 And also we start on the cloud first.

11:39 So I think that was much an alternative to,

11:44 I think we have a very first private beta,

11:49 then we go to public beta, then we go to GA,

11:51 got stuck to.

11:54 But I joined, in the same time,

11:56 I joined many technical conference on ETA,

11:59 and also I joined as many Discord or Slack

12:03 as possible, trying to reach out to those developers

12:06 to see what kind of problem they are facing,

12:09 what's the challenge, what's on their wishlist.

12:12 And then I try to present, hey, this is a new tool.

12:15 It is very easy to use.

12:16 We do have the free tier on the cloud,

12:18 but it's very common that people say,

12:22 sorry, either by company policy or I'm still at the early stage, I don't want to use a cloud service or I don't want to commit to using a commercial product.

12:33 Can I have this on my laptop? Can I deploy this in our internal communities faster? I get so many answers like this, and we think that, hey, since developer is our main user,

12:49 maybe manager or CIO, CTO can be our buyer, but developer is the user.

12:54 And a port like this, it is very hard not to open source.

12:59 At least they have the core engine for them to get experience and get some bit more confidence.

13:04 And if they want to let us to manage the infra, yeah, that'd be great.

13:08 But open source seem to be I mean, we are not snowflake.

13:13 We are not Splunk.

13:14 So that's only a few companies out there can say no to open source.

13:18 But I think

13:20 we do think having open source model ready

13:23 will open the door for more people to try our product.

13:27 And also, we contribute back to the community.

13:29 For example, we leverage Clearhouse, and now we are contributing some streaming processing code back to Clearhouse.

13:39 But again, as a company, we move very fast.

13:43 Clearhouse has a lot of other priorities.

13:45 They are more focused on the cloud.

13:47 So we cannot share every code to them.

13:49 They don't really have time to merge them.

13:51 But the idea is that on one hand,

13:54 we open source ourself, CoreEngine as part of.

13:57 On the other hand, we contribute to those codes

14:00 to say ConnectHouse or we leverage a project called Bensos,

14:06 a source of things, and we're also coming back to it.

14:09 So we want to be a good player in the space,

14:12 in the same time, more importantly,

14:13 we want to get more user feedback from the developers,

14:18 even if they don't have no intention right now

14:20 to use our commercial products.

14:23 Absolutely, excellent.

14:24 And I think we just spent to use our commercial products. Absolutely. Excellent.

14:31 I think we just spent about the last two, three months on this.

14:35 And to be honest, I'm still new to this open source,

14:39 not saying marketing, but at least community building.

14:43 But yeah, it gives me more opportunity to talk to more developers because we are not really trying to sell you

14:46 anything. We are building an interesting product and just try to let you know and you can try it

14:51 for free and give us some feedback no matter having a feature request or using our PR,

14:55 that would be great. So more conversation started because we do have an open source project.

15:03 Phenomenal. And this you know, this voucher for the

15:05 open source motion, great to hear

15:08 that already you're getting all this new feedback

15:09 and traction and requests from people.

15:12 And, of course, it goes hand in hand with making

15:14 it easy to use. And I think you've been doing

15:16 a great job so far showing how easy

15:17 it is to use with all these demo videos

15:19 you make on Twitter, for example,

15:22 on X, like you publish them very often

15:24 at Joe, your handle. So I was very impressed. And, like, for example, on X, like you publish them very often, at Joe, your handle.

15:25 So I was very impressed.

15:27 And like, you know, you overlay yourself all the time,

15:29 try different formats.

15:30 So I really liked that.

15:32 I don't know if there's anything you could share with people

15:34 in terms of, you know, building in public

15:36 and doing those kinds of demos,

15:38 putting yourself in front of the camera.

15:40 If there's anything you learned so far

15:42 or a tool you found that you really enjoyed.

15:45 And then all the way to actually podcasting, you have your own podcast, Streaming Caffeine.

15:51 And I'm curious how it has helped with gathering this sort of feedback, hearing the problems

15:57 people experience in the industry and, you know, positioning the company as well to solve

16:03 those problems and, you problems and foster these relationships.

16:07 Just curious to hear how you've been experiencing all this.

16:10 Yeah, I do try a different approach in the past one or two years,

16:16 especially the last year.

16:18 I started a podcast.

16:20 I joined more conferences.

16:23 I'm so lucky to be a presenter in conference.

16:26 So I invite offline and I send more blog.

16:31 Essentially I think me as the co-founder

16:34 and no matter what kind of title I have,

16:37 I really want to do whatever I can

16:39 to help the company, help the product to be more successful.

16:43 If now, for example, we already have awesome database

16:48 engineers to write the code,

16:49 I don't have the good synchronous engineers to write the code,

16:53 but we do need people to further to let more people know

16:59 the awesomeness of 10 Plus and Proton.

17:02 And also we need to build more partnership.

17:04 We need to educate the market a little bit

17:09 about what is streaming processing.

17:11 So I feel like this is really something I can contribute

17:15 no matter to the company, to the team

17:17 or to the wider peer in this market.

17:21 So that's the reason why a couple of things

17:23 you mentioned that is the streaming company

17:25 is a very casual podcast.

17:29 It's more like, hey, I talked to some of my friends

17:31 or some new friends in this space.

17:34 It's not really about 10 plus or pro term.

17:37 It's really about, hey, what kind of cool project

17:39 you are working on in this streaming space,

17:41 no matter you are working on say message bus layer,

17:44 or you are working on, say, message bus layer, or you are working on

17:46 streaming processing engine, or you work on some database.

17:50 I just want to add this some fun element,

17:53 that's why I'm using the streaming cafe,

17:55 so people sometimes usually have a cup of espresso or latte.

18:00 At the very beginning, actually,

18:02 I was using this as a streaming cafe,

18:04 but my first guest

18:06 actually is Xander coming from Bivox.

18:08 Bivox, by the way, is an awesome open source project, especially for Python developers.

18:13 So we are a strong believer for CQ, but I know that a lot of people prefer using Python

18:18 to have more flexibility, and that Bivox is a cool way to use Samsung as the interface

18:26 but another means they have

18:27 a Rust engine which is very

18:29 fast. So my

18:31 first guest is Xander from

18:33 Bawax and he said, yeah, maybe

18:35 we should be more inclusive.

18:37 People might drink tea, for example,

18:40 but all those things have caffeine

18:42 so let's call this

18:43 streaming caffeine.

18:47 And we have a lot of cool conversation. And sometimes I do ask a similar question as this one that is, hey, I do have the interest, but you may want to share more to the audience.

18:55 That is, what's your lesson learned when you bring a product or have a new project?

18:59 So this streaming caffeine, I think right now it's 10 or 11 episodes.

19:05 I interview people in different ways, virtual or in person.

19:13 Maybe I mentioned one or two sentences about 10 plus or 14, but the whole point is this

19:17 is not a commercial purpose.

19:20 It is really a very casual way for people who want to join this space or

19:26 already in this space want to know what's happening. In the same time, I also have other,

19:30 not really podcasts, but more like a playlist that is, for example, how to learn streaming

19:35 music in three minutes. Every video I try to edit or record this within about three minutes. So

19:42 every three minutes you learn a new concept. course i also do some other demos during some questions yeah for example we are now using uh

19:54 riverside i think which is awesome too i think about like 10 months ago i think i have

20:00 the very first time someone has me, also using Riverside.

20:05 I think he also published some video on the blog.

20:09 I said, oh, this is such a nice tool,

20:10 and I started using it.

20:12 Also some of my streaming podcast,

20:15 typing podcast using it.

20:16 It is a very nice tool, and at the same time,

20:18 you know there's some AI magic in this tool,

20:20 but a lot of person needed to,

20:23 nothing just be perfect, but sometimes lot of people need to... I'm not saying it has to be perfect, but

20:25 sometimes we do have different requirements.

20:28 For example, it's okay if you have an initial

20:29 video for, say, 40 minutes.

20:32 But on Twitter, for example, I think

20:33 the limit is 1

20:35 40 second, just like 1 40

20:37 carat. So about 2 minutes.

20:40 So

20:40 maybe

20:42 you can put a long video, but people don't watch for that one minute video on LinkedIn.

20:48 So you don't need to edit in different versions, add some subtitle or official effects.

20:54 So I do use some other tools such as Final Cut, I'm using Riverside, I'm using some other tool.

21:03 That's awesome. I think you're doing a great job so far. Keep it up.

21:07 It gives

21:08 ourselves as the host the opportunity to

21:10 form these connections, to learn

21:12 from the guests as well as for the

21:14 guests to share everything they

21:16 know and everyone get through these reps

21:18 of putting ourselves in front of

21:20 the camera, discussing.

21:22 One thing I want to add,

21:25 that is, this is also

21:27 kind of a mindset shift,

21:29 because when I was

21:31 a senior manager or director

21:33 in a company like Spunk,

21:35 sometimes manage like 50 people,

21:38 I do

21:39 spend more time on how we can

21:41 define OKR, how we can have the process,

21:44 how we call retention, of course, how we can have the process, how we can call retention,

21:46 of course, informants, things like that.

21:48 When I was in that position, I was not really so passionate regarding social media or LinkedIn.

21:54 But now I'm in this founder position, I do need to make more connections.

22:00 I'm looking for opportunities.

22:01 But also, in the same time, I do see a need to build our own brand

22:06 and position better for ourselves, that is, we are the domain experts. Because we are so quiet

22:14 for me in the past, and now we are building a new next-generation product, why is open source?

22:22 People don't really have a lot of context about the product.

22:27 So if they start to believe that, yes, you are a person with the right knowledge,

22:32 experience, skill, and you already experienced so many things,

22:36 and if they start to trust you as an individual,

22:39 then it will be easier if you say, oh, this is the product.

22:42 I've spent so much effort, and it gets a lot

22:45 of good conditions for all my people around.

22:48 So that will be a higher chance for your audience to be the conference and to be the product.

22:55 So I'm not saying I have to be an influencer.

22:59 I will never be.

23:00 But the point that is, the reason why I spend so much time writing everyday or publish videos

23:08 from time to time is to build this cadence and awareness.

23:13 So no matter you are new to data streaming, you check my blog, actually sometimes not

23:20 really related to my products.

23:22 It's more about my opinions.

23:24 Things like that. So I don't want to my products. It's more about my opinions. Things like that.

23:25 So I don't want to get confused.

23:27 But anyway, the key point that is,

23:29 also, I sometimes help people using AirBuy,

23:32 helping people using NetTunnel.

23:35 They are still in this data space.

23:37 So I don't really have to sell them to using campus,

23:40 but if they really see my article, my blog,

23:44 give them help, and I'm not asking for return,

23:48 but that kind of trust, that kind of

23:53 inference can help me to reach out to more people

23:58 and make their life easier.

24:00 I mean, that would be.

24:04 Also, more importantly, talk to the users.

24:08 Exactly.

24:09 You think they'll get the feedback, validate your assumptions.

24:13 No matter if it's fault or it doesn't really matter.

24:16 No matter whether it's infrastructure features or it's a solution for a certain industrial,

24:24 it doesn't really matter either.

24:25 As long as it's not for a particular user, because the key thing we want to avoid is

24:31 we don't want to spend too much effort just to make one user happy. If that's very unique for

24:36 them, then you cannot leverage that kind of effort to sell to the other user or at least make the other user more efficient. So the key point that is, be very careful to spend effort on your gadgets.

24:50 Focus more on the product.

24:52 No matter the product, as long as it's a repeatable pattern,

24:56 as long as it really solves a real problem, you are on the good course.

25:03 Exactly. That was very insightful.

25:05 Thank you so much for sharing this. And. That was very insightful. Thank you.

25:05 Thank you so much for sharing this.

25:07 And it is a tension for a lot of infrastructure companies,

25:09 you know, navigating,

25:10 building more for the organizations that need to self-host

25:13 and maybe have different compliance requirements

25:15 or, you know, for the cloud version.

25:18 And it can be a challenge to balance everything

25:20 and where to put engineering bandwidth.

25:22 But, you know, as you said, at the end of the day,

25:24 you need to do something that is repeatable and is going to benefit a lot of

25:27 people, a lot of organizations and developers.

25:31 So that was really insightful.

25:32 Thanks so much for sharing this.

25:34 And without further ado, I think we can transition into the demo section.

25:39 And for that, I have my co-founder, Zaf,

25:42 a joint to help go through the demo.

25:47 I can share it.

25:48 And then on your end, sharing your screen, I think, should be good to go.

25:54 Yeah, I can share the screen.

25:55 Awesome. Thanks.

25:57 Hey, Jov. It's great to have you here.

25:59 Yeah, of course.

26:00 So let me demonstrate a little bit about our products,

26:04 both on the template side and on the platform side, to give you an idea about what I'm talking about.

26:09 So, Proton is the open source engine we published to make the data engineer more

26:15 efficient to process data in Kafka, in the IoT devices, in fintech data sources.

26:22 And you can think of this as a SQL engine

26:26 for you to process data in an efficient way.

26:29 In the past, or even today, people using tools

26:33 like Apache Flink or ksqlDB all write their own code

26:38 to process the data by themselves.

26:40 And it requires a lot of learning and tuning and the infrastructure costs because those are not easy enough to learn and they are complex.

26:52 Because of the staff is in Nigeria, there are a lot of components.

26:58 Sometimes you have to spend two or three months to really set up a very healthy system. But at this point, we build a

27:08 for the clear path, so it is very efficient.

27:12 And we add a lot of stream processing,

27:15 processing capability so that you can use this

27:19 using CQS interface to process data easily.

27:21 So for example, I have this demo stack using Repanda as a message bus.

27:29 So for people who don't really know too much about Repanda, it is similar to Kafka.

27:34 Actually, they're using the same API or protocol as Kafka, but Kafka is written in Java or

27:43 maybe Star a little bit.

27:45 It is a very mature product and the company behind it is Confluent Cloud, the Confluent

27:52 Cloud platform.

27:53 Apache Kafka is almost a de facto data streaming service number one or top three in this space. And the Rekanda is a

28:06 talk-to-talk compatible data streaming system,

28:09 but they implement using Superfast just like us.

28:12 So we share a lot of similarity.

28:14 We are aiming for easy to use, very efficient,

28:18 and we have very good integration.

28:21 So what I'm showing right now is in the,

28:25 in the ProTerm example folder,

28:28 and you can see there's a lot of different examples,

28:31 for example, for the cost sharing, which is IoT use case,

28:36 Coinbase, which is a FinTech fraud detection,

28:39 even Hacker News.

28:41 So in the e-commerce folder,

28:44 this is using Stocker Comp stock compose it's a very uh this is a quite

28:51 a simple stack we have the bottom latest version here and we have the repanda as the message bus

28:58 to different topics for different data and we have this this console, which is, this ReHANA console is an SUI

29:07 for you to review all the message in the topic.

29:13 The modern data engineer today,

29:14 that is, you have all the data in different systems,

29:17 maybe in PostgreSQL database,

29:19 maybe in Snowflake, in our SMAB HLF system.

29:24 It is very common to consolidate them together

29:27 in Kafka or Rangando or Kinesis as a mesh device.

29:31 Then you need a tool to process that data.

29:35 And a very common choice today is

29:39 a spec.

29:40 The problem with that is it is a relatively heavy

29:44 fun-facts tool for you to learn on the team.

29:48 So, for example, if now we have this list of topics, imagine you want to monitor how

29:56 your website is being visited.

29:59 So if you, which user is purchasing which product and whether your recent marketer campaign

30:09 gets the right impact you want. So Repanda and or Kafka give you this opportunity to

30:17 put any data into the system, it's JSON or maybe an overall protocol, all those messages can be put in in a very flexible

30:26 way and you need to analyze those data. For example, here we have all the IP address,

30:31 we have the IP method, or maybe the size. You may want to understand how many requests is using post, how many is using guest.

30:45 Or maybe you can do more interesting stuff

30:47 for which country or which city

30:51 is visiting which category.

30:54 So the idea that is you have,

30:57 you've consolidated your data in,

31:00 this is not the problem we saw from the start.

31:02 This is assuming this is already set up.

31:06 How important can you tell me that is,

31:08 you already have all the data consolidated

31:10 in a very nice data streaming effort,

31:14 then let's using Fortum to understand this

31:19 without being set up a very complex step.

31:23 So we have Fortem, for example, I think I would have Portem right now, for example,

31:30 this is the Portem client.

31:32 We are using the latest Portem and I show stream which is there's no data available,

31:40 no data defined. So, example that is for, say, this is the docs-time-pass.com,

31:48 bottom, this is our documentation, feel free to read them all. So, we do have a way to

31:55 read or write popcorn. So, in this example, we can, we can use an Excel stream like this.

32:12 we can use an XNodeString like this. So we define a table or a string and point to the Rappanrinda topic and we can get this data easily. So I copy this one.

32:33 And you can see if I say select a star from the front end JSON, you can, I mean, this query never ends.

32:34 This is also an interesting way that is, traditionally, CQ is a bounded query.

32:41 You query a range for your snowflake or your goosegrass

32:45 and you wait for one second or one minute,

32:48 you get the result.

32:50 But in our case, the CPU is always running

32:52 because it's.

32:58 For sure you can have ways, for example,

33:01 I only care about the method,

33:02 I only care about the address. For example, in this case about the method, I only care about the address.

33:05 For example, in this case, you can do the select certain field or you can have some

33:12 where calls for sure, or you can have aggregation of the method. So this is a very simple example that is we can list how many events for certain posts

33:38 or certain methods.

33:39 And this will give you a theoretical results and so this is also based on our default behavior that

33:48 is every two seconds that if you want to get the output say every five five seconds so you can have

33:57 stuff like this maybe so i explain a little bit that is, we have this window called combo window,

34:09 which so every, it's misspelled, so.

34:16 So every five seconds, there's a window.

34:20 In that window, we calculate how many events

34:23 in that window.

34:24 And there's some other window, for example, sliding window that window, we calculate how many events in that window.

34:25 There's another window, for example, sliding window, that is, if not like tumble, it's

34:33 every five seconds we can have a window moving in a period.

34:38 So it's more like give me results every one second for the past five seconds, for example. So those are very interesting ways for you to

34:47 keep monitoring the data without

34:51 waiting for the next interval. So traditionally, people

34:55 know there's some ground job, there's some

34:58 ETL. You either wait for one day or you wait for every

35:03 hour to get the new data.

35:07 But in the streaming world, with the popular tools like Pulsar and Redundant and Pulsar,

35:15 now it is very easy for you to keep getting data.

35:19 Now you need a way to process data in an efficient way, in a continuous way.

35:23 And that's how we can work.

35:32 This is the engine side, but in the same time we have the time pass, which is more like adding extra UI in the Cloud Thread Delay interface. For example, it gives you a very quick data, let's say,

35:40 you can define different source or different processing logic and you can combine them together

35:48 as a data lineage or sometimes we call it a DAG. So you can easily understand what, say for some,

35:58 say this might be too small, but say for a certain data source, this is the view defined,

36:04 how other views is created

36:06 based on that view.

36:07 So, you can better understand if there are some data quality issues, which component

36:14 we should be spending time on.

36:16 Similarly, we have some query interface.

36:20 We have our dashboard alerting systems.

36:31 We have a show a little bit. For example, we do have a lot of customers in the free time area. So for example, you notice the price of the luxury of Bitcoin in last year.

36:39 So if you can really set up a very kind of a new project to understand the Bitcoin price

36:48 of some strategy implemented, you might have a chance to get some actual return.

36:54 So we do provide a very solid tool for them to do the basic data ETL, or they can do some specific trading strategy or at least apply

37:09 some alerts for them to either buy or sell or pay attention.

37:13 We have some customers using us for security use cases, for example, whether you are being

37:18 attacked or some other IoT cases, if you have many cars on the street, which area needs more cars,

37:27 or whether people have any incentives for your logging or something. So there's a lot of things

37:33 you can do, but essentially those are data problems, and we have very good processing.

37:44 Thank you so much for the demo, Joe.

37:46 So with TimePlus, basically, if we have an existing Kafka or Red Panda pipeline,

37:53 we can just throw TimePlus in front of it, and we can get real-time insights via

38:00 these queries and charts and more, right?