Red Hat Summit 2024 | Michael Kim | Samsung Electronics & Taehoon Lee | Red Hat

Clips
News
More from Red Hat Summit 2024

Michael Kim

Head of SMRC

Samsung Electronics

Taehoon Lee

Strategic Account Manager

Red Hat

Samsung and Red Hat join forces for advanced enterprise memory and computing

With steady advances in technologies such as artificial intelligence, the Internet of Things and quantum computing, the innovation boundaries in memory protocol and computing are continually pushed forward.Red Hat Inc.’s alliance with Samsung Electronics Co. Ltd exemplifies this progress. By merging expertise in open-source solutions and hardware excellence, the partnership has introduced a groundbreaking paradigm for data-intensive applications, particularly in the realm of AI.“Samsung is a hardware maker, a special memory maker,” said Michael Kim (pictured, left), senior director of Samsung Memory Research Center and head of the OCP experience center at Samsung Electronics

play_circle_outline Red Hat and Samsung partnership

play_circle_outline CXL as a solution for lowering TCO and improving performance

play_circle_outline Utilizing CXL for Enhanced Performance: Extending Main Memory and Maximizing Deployments with Certifications and Architectures

Info

Michael Kim | Samsung Electronics & Taehoon Lee | Red Hat

Michael Kim

Head of SMRC Samsung Electronics

Taehoon Lee

Strategic Account Manager Red Hat

search

>> Hello and welcome back to day three of Red Hat Summit and AnsibleFest from sunny and breezy Denver. Today we're going to continue to unpack what's been going on. Again, really digging into the ecosystem around Red Hat and what they're bringing to bear in the market. And for that I'm really excited because we actually get to go a little bit deeper into the chips and to the memory and get a little bit better understanding of how you actually stand these things up. I'm going to welcome to the stage here, Michael Kim, who's the head of SMRC for Samsung Electronics. And Taehoon Lee, who's the strategic account manager for Red Hat. Welcome on board both of you.

Michael Kim

>> Hi.

Taehoon Lee

>> Hello.

>> I think one of the things maybe people don't understand how Red Hat and Samsung are working together. Why don't you give us kind of an idea of how this partnership has evolved and really got started?

Michael Kim

>> Yeah, let me start first. Samsung is a hardware maker, the special memory makers. So we need a partner. So actually we product-based company, but we believe we need to move forward the solution company. But to make a solution, we need to do a lot of things, applications, network, hardware, software. But we cannot cover everything by ourselves. So that's why we are looking for partners. So I found that Red Hat is really good partner because they own the operating system and sold their platform. Very good ecosystem. That's why I approached him and he give a lot of good ideas for how we do a partnership. That's how we started.

>> And from the Red Hat side.

Taehoon Lee

>> This is keyword share the market share. So a lot of... More than 70% of enterprise customer use Red Hat platforms. And globally, everybody uses Samsung memory. So there's a new memory generation will come, the name of CXL. So we are collaborating using the market shares to expand our memory market. So yeah, this is a concept of our collaborations. And there's a Samsung CXL memory and there's an enterprise Linux of Red Hat, OpenStack and OpenShift at SMRC. So this block is from the... Just one block cannot have a meaning, but blocks are meaning like our collaboration, we can make a service and expand memory markets.

Michael Kim

>> For more information, you can see the QR code, here's our collaboration article.

>> Great. Yeah, and I think by the way you designed that as well, the Legos, which is I think extremely creative. I think because we go to a lot of shows and we see a lot of things. We actually do a segment at CubeCon called the Swag Segment where we collect from everybody. And I got to say, this is one of the best things I've seen in a long time.

Michael Kim

>> Thank you. I appreciate it.

>> Very good job.

Michael Kim

>> . All right, one for you.

>> Yeah. Oh, well, thank you. Thank you. I love that. So tell us a little bit more about CXL and why it's important and why it's exciting.

Michael Kim

>> Okay. The CXL is a new kind of device, a new protocol. So people looking for a new technology to cover the old technology or some of the limitations. So CXL is one of the new technology to overcome the problem or the previous current IT systems, so especially memory and some more computing process.

>> And everybody here is talking about AI. And AI is very data-intensive, very memory-intensive. How do you see that fitting in with this relationship with things like OpenShift AI and all the things that are being announced?

Michael Kim

>> So actually AI can be anything because we still don't know computing will be... Requires memory, it requires network and the data and the people, many things related. So still memory is very important part of the AI because all about the computing of data but where they're going to store, where they're going to lead the data? Is it from the storage and the memory. So that's why the memory company, we also manufacturing SSD as a storage and the memory to loading a data. So that's why we really close for the AI and the CXL is one of the kind of it.

>> So you're helping accelerate it and helping to actually-

Taehoon Lee

>> Optimize it.

Michael Kim

>> Optimize those deployments.

Taehoon Lee

>> Yes. When data center... The enterprise company who has a data centers, they need to optimize considering ESG's for future. So CXL will be one of the answer for that.

>> Yes. So sustainability is also a big piece of this then?

Taehoon Lee

>> Yes. Yes.

>> And so is this helping people be more efficient with their servers if they're having the CXL?

Michael Kim

>> Yeah., yeah, True. Because the hyperscaler like AWS and Google and Azure, they're looking for the lowering TCO, the price, the efficiencies. So CXL can be one of the solutions to lower their TCO, not only the improving the performance. So they're looking for those kinds of solutions.

>> And it stands for Compute Express link, and you're saying it's a new way for them to back and to communicate with that memory as well?

Michael Kim

>> Yeah. Okay. So the easier to understand is this one, everyone know the SSD, SSD is just a plug-in and plug-out. So they're have pluggable device but easy to put in, but the memory you need to open the door and the plug-in and turn on that later. But the memory, the CXL, you use Edge like SSD in front, just plug-in and recognize your memory. So that's easy to extend your main memory.

>> Oh okay. So you're also... It's, I guess, like circular economy with how you can actually make your servers older servers better now or... And to be able to support workloads like OpenShift AI and some of the Red Hat... Or RHEL AI and things of that nature.

Taehoon Lee

>> Yes, yes.

Michael Kim

>> There will be a lot of limitation though, but later on we are going to have a more server hardware storage, whatever things come out, enabling a CXL. Then you'll see how CXL is enabling the technologies and make it better. So especially the hyperscaler and the enterprise or the IT company looking for efficient way so OpenShift is one of the very important portion of it, but as I said, the memory is the very important part and when the OpenShift require the more memory, I believe CXL can provide more memory for it.

>> Yeah.

Taehoon Lee

>> From RHEL 9.3, every product is enabled the CXL and core. So which means market adoption is very easy to do that. So kind of like portable batteries to extend our memories like CXL.

>> So what you're doing is... Because it's open source, you're being able to bring it in, CXL in, to be able to take actually advantage of that and actually lower latencies or what... How does that work? Does it help?

Michael Kim

>> Okay, so there's a main memory, DRAM. There's a SSD for the storage, but there's some gap in between those kind of two components. But the CXL is in between the lowering of the gap. So from the memory, high speed and the SSD be low, but in the gap you can extend the more memory in same time reduce... Improve performance because we reducing a gap of... Between the SSD and DRAM.

>> Okay. So it can help almost like caching, almost like extending the cache or the SSD for the most part.

Taehoon Lee

>> It is. That is new technology. So we need to certify our product and Samsung Electronics products. So we are continuously processing our certification for next market.

Michael Kim

>> We focus on is this one. So for the end user, for the user of the IT, we make them comfortable, which means we certifying our device to the open system and the platform and we make a best practice and the reference architecture which they can believe on the documentation, then... For them, they're easy to adopt and needs to use it. That's one of the purpose of collaboration.

>> Yeah, no it seems like this would be, like you said, the hyperscalers and enterprises that want to take advantage of going faster basically and be able to get at their data because all these activities are very data intensive, be it analytics, be it ML, be it RHEL AI, what have you. Are you seeing these different use cases or what you're designing this memory for is?

Michael Kim

>> Okay. So there will be a lot the use cases. We still need to look for more and more, but one of the cases . So this memory can be interleaving together because DRM is small and they need more memory than we put the CXM memory as a interleaving as one memory. There is one of the use case to enlarge your memory or as a cache we have a tiering, the main memory and be lower tier memory as a CXL and tiering between those kind of two memory, that's one of the use case. And just be more, the Broadcom... We also working with the Broadcom, they using this memory as a memory virtualization. I believe and OpenStack, OpenShift also can use for the memory virtualization. That's another use case you can think.

>> Yeah so something like OpenShift or OpenShift Vert could take advantage of this and be able to carve it up for the different containers or VMs that are on there.

Taehoon Lee

>> Yeah, true. So certifications of reference architecture with the white papers that we are conducting. Yeah.

>> And are you seeing that this is... Is this going through your same channels and distribution? Are you tied tightly from an ecosystem perspective as well?

Michael Kim

>> It's quite open actually. So it's not only between us, I just want to open to everyone. I'm just welcoming all the ecosystem to come and play together and make a solution together then I believe they will stronger our solution together.

>> Right, right.

Taehoon Lee

>> For Red Hat side is world expanded platforms. That's why we are doing the first time because platform between memories you need to be certified.

>> Yeah that makes sense. And I think when you look at the entire... I used to run data centers a long, long time ago, and when you looked at it you always wanted to figure out how long you could keep your servers running and how you could... Because once you bought it, like you said, you can't really... You could upgrade the memory but you can only get to certain extent. This would seem like a really good way for organizations, especially enterprises, and obviously the hyperscalers at much different scale to go and be able to now bring new life to those servers and to put different workloads on there. Are you seeing a lot of pickup in that where people are going to the Red Hat QR code and looking at how they actually increase that?

Michael Kim

>> Okay, that's a very good point. So as I mentioned before, they want to lowering a TCO, which means one of that is they have all the servers still usable but they need more memory, they need more computing power. So most of time they buy a new server which is a bit waste of the money. So to keep to the old server, they just buy some CXM machine out of the server and connect it to provide a more memory, which means they can use old server longer and better. So that's one of the way to lowering a TCO. So all the... Quite number of the hyperscaler customer looking for those kind of solutions.

>> Yeah. And obviously, OpenShift runs on all the hyperscalers for the most part and Rosas and others. I would assume that this being certified allows those hyperscalers that you can work with them, that this could be in those servers as well.

Taehoon Lee

>> Yes, we forecasting our CXL's future is the memory donor. So AI is very focused now in GPUs now, but there's an option... CXL will be the option for tomorrow. Yeah, maybe. Yeah.

>> No that's great. So what is final thoughts from you guys on what do you want people to know? What's the key for them to know that you want them to take away from this discussion?

Michael Kim

>> Let me start first for that. So most important thing is I believe ecosystem, so all the player come together and make a solution together, they'll make a ecosystem stronger, means that we can go to the market stronger, then more people will use it, more people can use for each for adoption. Then... So the ecosystem, I want to say welcoming all the ecosystem, all the player come and play together to make a solution together.

>> Excellent.

Taehoon Lee

>> As a Red Hat side, Red Hat recently did a lot of compatibility and certification process for existing applications and infrastructures, but now we are expanding to the future for next memory or next markets. So we are doing for next as a Red Hat. Yeah/

>> No, I think that's a great place to leave it. I want thank you both for coming on board. You're now Cube alumni and you're welcome back anytime. So thank you very much and thank you for bringing this. It's awesome. I'm going to build this tonight, so I really-

Michael Kim

>> .

>> Yes, thank you.

Taehoon Lee

>> Okay, thank you.

Michael Kim

>> Thank you.

>> And thank you for watching the Cube's coverage, wall to wall, day three of Red Hat Summit and AnsibleFest. Stick around. We got some more coming. We'll be right back in a few seconds. The Cube, the leader in tech analysis and news. Stay tuned.

Michael Kim | Samsung Electronics & Taehoon Lee | Red Hat

search

>> Hello and welcome back to day three of Red Hat Summit and AnsibleFest from sunny and breezy Denver. Today we're going to continue to unpack what's been going on. Again, really digging into the ecosystem around Red Hat and what they're bringing to bear in the market. And for that I'm really excited because we actually get to go a little bit deeper into the chips and to the memory and get a little bit better understanding of how you actually stand these things up. I'm going to welcome to the stage here, Michael Kim, who's the head of SMRC for Samsung Electronics. And Taehoon Lee, who's the strategic account manager for Red Hat. Welcome on board both of you.

Michael Kim

>> Hi.

Taehoon Lee

>> Hello.

>> I think one of the things maybe people don't understand how Red Hat and Samsung are working together. Why don't you give us kind of an idea of how this partnership has evolved and really got started?

Michael Kim

>> Yeah, let me start first. Samsung is a hardware maker, the special memory makers. So we need a partner. So actually we product-based company, but we believe we need to move forward the solution company. But to make a solution, we need to do a lot of things, applications, network, hardware, software. But we cannot cover everything by ourselves. So that's why we are looking for partners. So I found that Red Hat is really good partner because they own the operating system and sold their platform. Very good ecosystem. That's why I approached him and he give a lot of good ideas for how we do a partnership. That's how we started.

>> And from the Red Hat side.

Taehoon Lee

>> This is keyword share the market share. So a lot of... More than 70% of enterprise customer use Red Hat platforms. And globally, everybody uses Samsung memory. So there's a new memory generation will come, the name of CXL. So we are collaborating using the market shares to expand our memory market. So yeah, this is a concept of our collaborations. And there's a Samsung CXL memory and there's an enterprise Linux of Red Hat, OpenStack and OpenShift at SMRC. So this block is from the... Just one block cannot have a meaning, but blocks are meaning like our collaboration, we can make a service and expand memory markets.

Michael Kim

>> For more information, you can see the QR code, here's our collaboration article.

>> Great. Yeah, and I think by the way you designed that as well, the Legos, which is I think extremely creative. I think because we go to a lot of shows and we see a lot of things. We actually do a segment at CubeCon called the Swag Segment where we collect from everybody. And I got to say, this is one of the best things I've seen in a long time.

Michael Kim

>> Thank you. I appreciate it.

>> Very good job.

Michael Kim

>> . All right, one for you.

>> Yeah. Oh, well, thank you. Thank you. I love that. So tell us a little bit more about CXL and why it's important and why it's exciting.

Michael Kim

>> Okay. The CXL is a new kind of device, a new protocol. So people looking for a new technology to cover the old technology or some of the limitations. So CXL is one of the new technology to overcome the problem or the previous current IT systems, so especially memory and some more computing process.

>> And everybody here is talking about AI. And AI is very data-intensive, very memory-intensive. How do you see that fitting in with this relationship with things like OpenShift AI and all the things that are being announced?

Michael Kim

>> So actually AI can be anything because we still don't know computing will be... Requires memory, it requires network and the data and the people, many things related. So still memory is very important part of the AI because all about the computing of data but where they're going to store, where they're going to lead the data? Is it from the storage and the memory. So that's why the memory company, we also manufacturing SSD as a storage and the memory to loading a data. So that's why we really close for the AI and the CXL is one of the kind of it.

>> So you're helping accelerate it and helping to actually-

Taehoon Lee

>> Optimize it.

Michael Kim

>> Optimize those deployments.

Taehoon Lee

>> Yes. When data center... The enterprise company who has a data centers, they need to optimize considering ESG's for future. So CXL will be one of the answer for that.

>> Yes. So sustainability is also a big piece of this then?

Taehoon Lee

>> Yes. Yes.

>> And so is this helping people be more efficient with their servers if they're having the CXL?

Michael Kim

>> Yeah., yeah, True. Because the hyperscaler like AWS and Google and Azure, they're looking for the lowering TCO, the price, the efficiencies. So CXL can be one of the solutions to lower their TCO, not only the improving the performance. So they're looking for those kinds of solutions.

>> And it stands for Compute Express link, and you're saying it's a new way for them to back and to communicate with that memory as well?

Michael Kim

>> Yeah. Okay. So the easier to understand is this one, everyone know the SSD, SSD is just a plug-in and plug-out. So they're have pluggable device but easy to put in, but the memory you need to open the door and the plug-in and turn on that later. But the memory, the CXL, you use Edge like SSD in front, just plug-in and recognize your memory. So that's easy to extend your main memory.

>> Oh okay. So you're also... It's, I guess, like circular economy with how you can actually make your servers older servers better now or... And to be able to support workloads like OpenShift AI and some of the Red Hat... Or RHEL AI and things of that nature.

Taehoon Lee

>> Yes, yes.

Michael Kim

>> There will be a lot of limitation though, but later on we are going to have a more server hardware storage, whatever things come out, enabling a CXL. Then you'll see how CXL is enabling the technologies and make it better. So especially the hyperscaler and the enterprise or the IT company looking for efficient way so OpenShift is one of the very important portion of it, but as I said, the memory is the very important part and when the OpenShift require the more memory, I believe CXL can provide more memory for it.

>> Yeah.

Taehoon Lee

>> From RHEL 9.3, every product is enabled the CXL and core. So which means market adoption is very easy to do that. So kind of like portable batteries to extend our memories like CXL.

>> So what you're doing is... Because it's open source, you're being able to bring it in, CXL in, to be able to take actually advantage of that and actually lower latencies or what... How does that work? Does it help?

Michael Kim

>> Okay, so there's a main memory, DRAM. There's a SSD for the storage, but there's some gap in between those kind of two components. But the CXL is in between the lowering of the gap. So from the memory, high speed and the SSD be low, but in the gap you can extend the more memory in same time reduce... Improve performance because we reducing a gap of... Between the SSD and DRAM.

>> Okay. So it can help almost like caching, almost like extending the cache or the SSD for the most part.

Taehoon Lee

>> It is. That is new technology. So we need to certify our product and Samsung Electronics products. So we are continuously processing our certification for next market.

Michael Kim

>> We focus on is this one. So for the end user, for the user of the IT, we make them comfortable, which means we certifying our device to the open system and the platform and we make a best practice and the reference architecture which they can believe on the documentation, then... For them, they're easy to adopt and needs to use it. That's one of the purpose of collaboration.

>> Yeah, no it seems like this would be, like you said, the hyperscalers and enterprises that want to take advantage of going faster basically and be able to get at their data because all these activities are very data intensive, be it analytics, be it ML, be it RHEL AI, what have you. Are you seeing these different use cases or what you're designing this memory for is?

Michael Kim

>> Okay. So there will be a lot the use cases. We still need to look for more and more, but one of the cases . So this memory can be interleaving together because DRM is small and they need more memory than we put the CXM memory as a interleaving as one memory. There is one of the use case to enlarge your memory or as a cache we have a tiering, the main memory and be lower tier memory as a CXL and tiering between those kind of two memory, that's one of the use case. And just be more, the Broadcom... We also working with the Broadcom, they using this memory as a memory virtualization. I believe and OpenStack, OpenShift also can use for the memory virtualization. That's another use case you can think.

>> Yeah so something like OpenShift or OpenShift Vert could take advantage of this and be able to carve it up for the different containers or VMs that are on there.

Taehoon Lee

>> Yeah, true. So certifications of reference architecture with the white papers that we are conducting. Yeah.

>> And are you seeing that this is... Is this going through your same channels and distribution? Are you tied tightly from an ecosystem perspective as well?

Michael Kim

>> It's quite open actually. So it's not only between us, I just want to open to everyone. I'm just welcoming all the ecosystem to come and play together and make a solution together then I believe they will stronger our solution together.

>> Right, right.

Taehoon Lee

>> For Red Hat side is world expanded platforms. That's why we are doing the first time because platform between memories you need to be certified.

>> Yeah that makes sense. And I think when you look at the entire... I used to run data centers a long, long time ago, and when you looked at it you always wanted to figure out how long you could keep your servers running and how you could... Because once you bought it, like you said, you can't really... You could upgrade the memory but you can only get to certain extent. This would seem like a really good way for organizations, especially enterprises, and obviously the hyperscalers at much different scale to go and be able to now bring new life to those servers and to put different workloads on there. Are you seeing a lot of pickup in that where people are going to the Red Hat QR code and looking at how they actually increase that?

Michael Kim

>> Okay, that's a very good point. So as I mentioned before, they want to lowering a TCO, which means one of that is they have all the servers still usable but they need more memory, they need more computing power. So most of time they buy a new server which is a bit waste of the money. So to keep to the old server, they just buy some CXM machine out of the server and connect it to provide a more memory, which means they can use old server longer and better. So that's one of the way to lowering a TCO. So all the... Quite number of the hyperscaler customer looking for those kind of solutions.

>> Yeah. And obviously, OpenShift runs on all the hyperscalers for the most part and Rosas and others. I would assume that this being certified allows those hyperscalers that you can work with them, that this could be in those servers as well.

Taehoon Lee

>> Yes, we forecasting our CXL's future is the memory donor. So AI is very focused now in GPUs now, but there's an option... CXL will be the option for tomorrow. Yeah, maybe. Yeah.

>> No that's great. So what is final thoughts from you guys on what do you want people to know? What's the key for them to know that you want them to take away from this discussion?

Michael Kim

>> Let me start first for that. So most important thing is I believe ecosystem, so all the player come together and make a solution together, they'll make a ecosystem stronger, means that we can go to the market stronger, then more people will use it, more people can use for each for adoption. Then... So the ecosystem, I want to say welcoming all the ecosystem, all the player come and play together to make a solution together.

>> Excellent.

Taehoon Lee

>> As a Red Hat side, Red Hat recently did a lot of compatibility and certification process for existing applications and infrastructures, but now we are expanding to the future for next memory or next markets. So we are doing for next as a Red Hat. Yeah/

>> No, I think that's a great place to leave it. I want thank you both for coming on board. You're now Cube alumni and you're welcome back anytime. So thank you very much and thank you for bringing this. It's awesome. I'm going to build this tonight, so I really-

Michael Kim

>> .

>> Yes, thank you.

Taehoon Lee

>> Okay, thank you.

Michael Kim

>> Thank you.

>> And thank you for watching the Cube's coverage, wall to wall, day three of Red Hat Summit and AnsibleFest. Stick around. We got some more coming. We'll be right back in a few seconds. The Cube, the leader in tech analysis and news. Stay tuned.