New CIQ Offering: Upstream Kernels in Rocky Linux

lot as they are successful I like to call you Mr Mainline stable Colonel around ciq because you run it I think you probably run it more than everybody else you stay current every week anybody else does that that's pretty amazing uh actually you're right I mean I've been running it for almost two years and it's actually running on this laptop right now so um so you're right it does update weekly on Sundays um and it is something that uh is very very automated um and we're applying more and more rigor to it every week um so that when we do deliver it to customers um

they're not uh they'll have high confidence in putting it in production um and you know this is where we'll we'll find all sorts of other use cases to really add value and you know one you know you know I think col might be joining us shortly but you know Colin and I are are heading over to Google today for a conference um to talk to a lot of the Google Cloud customers for example um it would be really nice to say that hey can we save you five to 10% on your Cloud bill just because we're running five to 10 percent faster because we're married

to all the beautiful work that the Upstream folks do with the colonels um that then gets pass down to all the diff different distributions um but this is the power of that Upstream kernel and it's the power of having the latest capabilities the latest access to the latest Hardware uh and be able to run that with confidence with the rocky 9 user space um that uh everybody is now have a lot of confidence in to run their apps that are isv certified etc etc so this is you know one of the things we got asked for um we want to go we heard that a

lot right so we work with a lot of HPC folks a lot of AI folks a lot of performance intensive Computing folks in general um Zane you hear this all the time too Rose you hear it all the time right so this is something we wanted to go to lier for them and uh and they're seeing a lot of benefits from it um and you know we're really excited about where we can take it and do more with it um and then I think the little known fact is there there's all the security starts there all the drivers start there right so it's it's it's

one of the it's one of the best places to get out ahead and then there's also not just the stable Colonel um or the main line which is Linus but the stable colel which is Greg Crow Hartman and then there's the 6.6 long-term kernel there's the 61 kernel um I actually run all those on different servers at home and those update as well and they all run pretty flawlessly as well so you do have a choice between stable 6661 um and depending on the feature sets that you want um you can choose the different kernels and we can help you choose the different ones that

might work for you um I hope that makes sense it does I think one of the questions I get asked quite often is why doesn't Enterprise Linux just run the mainline kernel to begin with why why is it so far behind or why is 18 months to two years behind can you give us a little clue um you know I'm sure there's all sorts of reasons for that um just to be clear uh uh but I I uh what I do know is that freezing the kernel in a certain spot and then creating a lot of have to do back porting um is is a

lot of work um it's not easy and red Hat's done a beautiful job at that for customers for many years and they have a army of people that do it Susa Etc right so um that being said it's it's also not necessarily what customers are asking us for um now customers are definitively asking us for you know traditional Enterprise Linux kernel so I want to be really clear ciq and you know as far as I know the Rocky the enterprise software Foundation has no interest in not doing that and they will always do that to some degree um so that all stays the same this

does not change that in any way this is in addition to that you now have Choice um similar to where maybe you'd have choice to go run Fedora or you'd go run something uh from auntu that's a little bit based on the Upstream kernels now with Rocky Linux uh with full support from ciq you have choice of three different Upstream kernels and it just gives you that choice to be able to either test Dev uh with your Dev teams or actually put those into production and try to get the benefits of additional enhancements of speed maybe some security enhancements that you're getting the first of

the security bug fixes uh etc etc so and security and Bug fixes you know at the stable kernel and the 6661 I mean that's where they all go first right um so it is very important in my mind to make sure our customers are secure they're compliant but we want to get as much performance out of that as possible for them and and they do too right they want to maximize their Investments whether it's on the cloud or you know that $100,000 AI server they just bought from Super Micro Dell or HP um you know or the superod from Nvidia Etc right so we we

want to make uh Rocky Linux optimized for NVIDIA we want to make it optimized for whatever that customer wants to run on and have that experience and get the most value for their money out of that experience does that sense it does I don't think most people realize that the Linux kernel that it comes with or just the version you get straight off the iso is compiled to run the broadest set of infrastructure it is not there for performance in any way it's just to make sure that it works with the most amount of stuff and I think that's something that it's interesting when you

say performance Improvement and just making some changes in the colonel could do that yeah and you know what's funny is uh the the Upstream kernel maintainers are extremely good at this right I mean they you know it's I've had conversations with many of them about well what what would you strip out you know of even the Upstream kernels and they were like like actually I've stripped everything out that I thought I could and it really wasn't that much improvement like and then you know the stability of you know the ABI stability of the Upstream Colonel is is looked after extremely closely by that you know

all those maintainers and you know what I'm talking about here isn't like a five people right this is you know you know Greg Crow Hartman is the leader but you know you got folks like Google you got Oracle you got red hat you got some of our team right you got sus you got like all sorts of major organizations AWS you got all sorts of Nvidia developers AMD developers Intel developers and and the beauty of trying to move more and more towards the Upstream kernels is all those other folks you know AMD they don't have to focus on backporting they don't have to focus on

getting their customers up and running on a different version from somebody else right they can focus on innovating they can focus on pushing all that Innovation upstream and working closely with all their Upstream Partners to take the kernel further faster and that's what I hear even from all those different organizations is that would be wonderful if you you all started working with those Upstream kernels and you could handle whatever features set the customer wants and get that to them and support it to them so that they can execute exactly how they want to execute with as much compliance which as much security and as with

as much performance but in an easy fashion they don't have to go rebuild it themselves they can just trust that they got something from someone like us and go execute on it and you know others have done this and others do this well so I'm not sitting here saying only ciq I'm not sitting here saying you know uh you know that you know this is the right way or the only way this is a way and we got asked by a lot of customers a lot of very big customers um clouds asked us right governments ask this right like this isn't you like small people

you don't listen to right um and you know these are people I want to succeed right like I want all the people in our government succeeding you know what I mean I kind of live here right you know you know I want all the cloud succeeding I use them on a regular basis right like so like and they're partners of ours right so like we want these people to do great work you know so it goes back to what I originally said doing cool stuff with cool people and you know kind of like hanging out with the two of you all right well we're gonna

actually bring a fourth person on so let's make Colin live here hello Mr Vander Smith how's it going guys going good how's everybody doing feel like I heard cloud and AI like 10 times and I'm like someone my ears were burning I guess and here you are it's like a genie you say it three times and Colin shows up he magically here yeah trying to get us you know sponsored by topa Chico so these every time I come on you know so Colin's working on something specific to Performance and optimizing in the cloud as an example it's a presentation he's going to do later at

Google for all their customers um that we'll be at so Colin I don't know if maybe you want to give everybody a little preview see I stole Zay and Rose's stick I love it I mean that was a good intro that's solid Robert um no I mean honestly so you've heard me say a lot of this stuff it's it's present in a lot of the materials that we've published both in our blogs as well as in previous uh webinars um but to isolate to kind of what are the kind of the five key focal points coming out of Google next a few weeks ago as

well as uh reinvent a few months ago and then looking at kind of the stage uh moving forward for this year there really are kind of like five key things that our customer custers like overwhelming are asking us for and the first part really falls into kind of cost and operational management then bleeds into other areas that we're helping with um you know they look both at their on premise uh cost of operations both in people process technology which also includes licensing and support and the rising uh cost you know in some cases from our competitors but just in a holistic view if you're thinking

at the top with a CIO that it's getting very very expensive and I need to find ways to short my balance sheet in order to invest in the geni and llm projects that I want to make possible I mean it's become a mandate where you know in a recent Gartner study uh roughly 90% of all Enterprises have a AI uh first mentality or things that they're trying to do but only 9% actually have a plan down on paper for what they're going to use it for and so when you get into you know the the cloud structures and and cost of operations that's where things

start getting very tricky so then the second part becomes well what does that even mean and so what we end up uh walking into a lot of customers is their their Readiness so again where's there going to be the fastest place to deploy and try uh a lot of these uh projects out is going to be in the cloud but then they find out that their data isn't ready so again a similar report both uh reinforced from Forester flexer and Gartner is that less than 5% of customer data is actually ready to be deployed and accessible in the cloud and is secure enough to provide

qualifying answers and so that just gets to availability the next layer to that is semantics so how uh how is the data structured to give you qualified answers to the questions and solutions that you're proposing so you have all these things kind of all working in conjunction and where we fit in that picture is really in Cloud optimization so getting you there so that you are ready getting you there so that you can start investing in those and freeing up the spend uh for your existing environments as they grow or move to the cloud and making all those realities even remotely possible and so you

know Zan you heard a lot of this uh from speeds and feeds um you know it's it's really surprising to me how much performance means just raw speed from you know let's say a silicon dynamic but then customers completely miss the efficiency side of how their workloads are running especially in the cloud and so it's very unique for us is that we can Bridge both sides of that making it operations more effective and efficient for on premise and making those same teams as they deploy or manage their large fleets across any cloud provider making those even more performant and efficient and that can mean workload

performance isolated to the compute structure or that can mean optimizing for the the fine-tuning requirements of that workload as they scale up and down so we've got a lot of things to help customers with now and I I feel super ready for this presentation at Google I mean it's GNA be a lot of fun there's gonna be a couple hundred customers we'll be at my old campus in South Lake Union uh so be back on familiar ground um we're gonna have a lot of fun today is it tonight or is it tomorrow it's in three hours three hours oh wow that's coming up really quick

Robert you better run that's why I'm in the office [Laughter] today so I have a question you guys as I'm thinking about this and we're using the word optimization and you have an upstream kernel so then and this is maybe a weird question but like then the operating system on top of that do they have a choice of that or it comes with Rocky yeah so would in this particular case be the rocky nine user bace um so yes um something that they use every day um and that their applications sit in so it would definitely come with an an enhancement um that you have

an option to run something different uh with rocky9 and then the other beauty of this is we make it extremely easy with our uh with our repositories inside of mountain which is our artifact delivery yes your logo is right there and uh and this also gives us the ability to also secure that delivery so we can secure that delivery with r-based access and we can secure that delivery with uh Project based access with our Edge deployment and then even artifa artifact based access so folks have a lot of control over securing these different artifacts and the updates and access to them so that uh an

administrator can set it up in addition this will fit right into our Enterprise Linux platform so not just Mountain but Ascender which is our awx anible solution we've automated the deployment of this um to where it will actually pull from that repository for you put that set that server up for you and now set up the updates weekly for you um as well so the complete solution is completely automated through the Enterprise Linux platform as well so again you'll hit your compliance wants you'll hit your security wants you'll hit your performance wants and we make it easy for you um that's that was our goal

when we were doing this was to just make it easy for people to get the benefit of what Colin just described you know specifically for gcp there um but this would fall over to anything on premise HBC clusters AI clusters anything that people want to run you know we're there to run with them well I think in that regard too I think there's even a a very blanket statement that we can make that there is not a workload on the planet on premise or in the cloud that we cannot make at least 5 to 10% better that can be performance that can be efficiency that

can be uh even in some cases security this hits Robert's three key points three bullet points he likes to secure performance easy I like it Miss Sylvie thanks for watching what versions of the mainline Kel does ciq work with that's a great question uh we do have this working for rocky9 with the stable kernel the long-term 66 kernel and the long-term 61 kernel today um doesn't mean we couldn't add others uh but those are the three that we have uh working today so you said there's updates every week correct so what does that mean to ciqs offering is that also updated weekly it is today

um now you don't have to necessarily put uh those updates weekly into production um unless you want to um that being said that those are available to you immediately um and you know Greg Crow Hartman said this to me once and it kind of sticks with me he goes every update every bug fix is a security fix um so we want to make sure security is available to everybody immediately um and uh and obviously any bug fix as well um so those are those are important to our customers to have that security and that compliance um and then the beauty of having bug fixes as

well is in a lot of cases that's also enhan in performance um or fixing something to make it run more uh more efficiently Etc or you know making something over time run a little bit uh uh faster Etc so it it again hits all those key points right it's compliance security performance and just making it easy for everybody to to use so we have customers that are wanting not only to use the newer kernel for performance but they also like that we can actually customize that kernel and get even more performance so we have anywhere from what 5 to 20% Improvement if you have the

new kernel and you compil it for your infrastructure so it depends right so early on say a new version of uh a new version of Enterprise Linux comes out um it's usually based on you know one of the Upstream kernels that is a little bit newer right as that kernel ages and it doesn't change for say a year or two um that's when the performance keeps accelerating with the Upstream so if it's uh six months older maybe it's closer to three to five if it's uh if it's you know more like a year plus now you start looking at maybe five to 10 um and

then don't forget there is performance enhancements with different versions of the user space too so even going from 8 to nine in the user space you get performance and then there's Flags we can put in there there's other areas um Intel's done some really amazing work with Fedora with the ISA um that I've actually run ran that with uh with the 66 Kel it worked fairly flawlessly from what I've seen um so there's investigation that we're doing there as well but that's you know an extensive flagging um that Intel built into there that makes and gets some of that performance um so we want to

make sure it's compatible Etc so there's a lot of other ways to get performance um that we are investigating as well with customers and and by ourselves um but this again it goes back to the central core of what customers just keep asking us for and you know a lot of folks especially the the the clouds especially the um the different uh uh capabilities for like Intel Nvidia AMD these people really want to focus on the Upstream kernels and and not worry about all the backport so we want we want that because we want all those people doing all the great work they're doing to

bring out better faster capabilities so we can just pass that on to our customers and then support them with it absolutely so Colin we talk about performance at a workload that's running in a hyperscaler I mean everybody obviously wants things to get done faster but from a financial perspective why is that important I think the biggest thing is just understanding cost management across uh all the lines of business it's not just an IT issue any longer um and so the best practices that uh I try and Foster for a lot of our customers is to adopt a critical finops uh culture inside of the company

uh and so in other words you would have uh both regular uh reviews you would look at how much is being spent across different uh lines of business uh the applications and things that they're using and and effectively do that across the board so that you know exactly you know what both cogs as well as you know uh Roi looks like across everything you're doing in the cloud um and those that have adopted that have seen not really cost reduction but they've been able to get to time to value or time to Market uh short of tremendously and so it's not just the Silicon it's

uh the entire workloads uh themselves and then also what that you know helps the business achieve you know from from that standpoint um and so in our world um yes it starts at the the infrastructure up um but it's something that's so often overlooked that it allows us to make that statement that there's not a a workload out there we can't make more performant or more efficient um and you know hopefully this uh helps a lot of customers really kind of reassess starting with inventory you know what they're doing right now both on premise and cloud and then take a healthy look at uh how

much they're spending and and what the return is I think that's really kind of the most critical place to start thank you Robert you mentioned earlier that people like AMD and Intel contribute code back Upstream so they're writing drivers for themselves I don't think people think about that as much as who's actually contributing back I think most people think of it as guys like us just sitting at home writing code but it is these larger Enterprises that are putting that code back out there for their own stuff yeah 100% And you know it's uh they do way more than you think too so know Greg

actually had a a diagram on this once I forget who the top was but I believe uh Google and Intel uh for example were two of the top Oracle is a huge huge uh Colonel maintainer right so um they do incredible work over there and we you know we know some of those team members and we're you know very proud to have a an alliance with them with the open Ela right Susa has a ton of folks that put incredible work Red Hat has great folks right so you know it's not just you know the amds and the intels of the world and nvidias but

they do a ton of that work because they got to get their Hardware to work with all this um so all the latest Hardware all those drivers get built for that Upstream kernels um and all the latest features feature ads get built in the mainline kernel and then that gets passed down on a basically a quarterly basis to become a stable Kel and then the stable kernels you know they'll be one chosen a year to be a long-term kernel and then those stay around for different amounts of time depending on which Kel it is but in general at least three years so this is where

uh you know the lot of the value is added uh as a entire Community um to the kernel so it's not just you know a distribution at this point it's every everyone it's clouds it's all the oems Dell HP you know so all those folks add a ton to this as well so and they're always fixing and making it better um you know I I was just talking to uh one of our folks that maintain Samba Upstream Jeremy Allison um him and Jonathan Maple have been working on an nvme shutdown issue um and you know somebody from Red Hat joined them somebody from joined them

um somebody from the Upstream maintainers joined them from Google right so Google originally started it like you know Jeremy picked it up and ran with it like now it's now they're they're getting it to where an nvme you know large nvme box with 28 different drives in it which used to be a linear little shutdown used to take literally 1020 minutes to shut down and and reboot back up they're getting it down to under three minutes so instead of like doing a reboot and walking away to fix something you know and coming back a half hour later now folks are going to be able to

you know reboot and you know immediately be able to go know if their fix worked and be able to itinerate on making things better and faster and you know one of our sponsors on this is you know run some of the biggest luster clusters in the world and do amazing things so you know it's very impressive what they do and you know we don't want folks like that having the weight half an hour to make an update right we want folks like that who are running massive luster clusters 24 by7 uh be able to do that in a two minutes three minutes fix things much

faster I think what goes with that too Robert is like what makes the headlines right now is you know a lot of the hardware capabilities that power gen and other Cloud workloads both on premise and and in the cloud and um I think one of the most recent ones that actually gets really interesting is like the power from a chip on up for confidential Computing which is really a hardware enabled uh piece that then the operating system because everything passes through the OS uh gets overlooked and that's really where both the security and stability and reliability posture for Rocky comes into play and then the

fine-tuning and kind of certification if you will that we provide you know as a safety net for a lot of our customers to enable those capabilities that are present on the workloads that they want to run there and so um I think you know between a lot of the hardware vendors that we work with and the cloud providers that we work with I think that'll be a big area that'll get exposed very quickly as a lot more of these uh environments and data sets get moved to the cloud is that air gapping you know zero trust Etc will not just be a network structure it's

going to be end to end right and I think that's part of the value that we provide working with a lot of those vend one to one Beyond just the the stability of the colonel absolutely yeah the only thing I I'd add to is I mean we talked to a lot of new newer AI companies and older ones could you imagine them being able to save five 10% on all their training you think any of them would be interested in that right like yeah you know we talked through some big transaction processors do you think if they could go 5 to 10% faster you think

their customers would be happier you think that you know they'd save five to 10 percent because it's just how long did it take to go so we just want to make them go faster um and keep them secure while they're doing it and you know this is one of the ways we found to serve them um and we want to continue to figure out better ways there's going to be more and better and you know as ciq grows we add more resources to add more of this capability and as more customers come to us with these similar demands we and we can deliver it we

can add more resources to do more and more capabilities like this for people and you know that's why we're here right to really serve our customers and the communities um that we try to you know give as much as we can to absolutely I a house I mean I don't know what else you can say after that you kind of summed it all up perfectly let you have anything else you want to talk about guys we're up on time here no it was real good to see uh Rose and Zane thank you for inviting absolutely colen welcome back hadn't been very long but it's good

to see you hey thanks for letting me uh pop in appreciate it y'all absolutely have fun today you guys it's gonna be exciting is any is it gonna be videoed like do we get to watch you somewhere you just going to report back uh well we'll see what we can do I'll find out sometimes they do sometimes they don't I would assume the one there is probably going to be is this a job for my flip phone again there you go is it fixed now it is it is so what is it though Colin what's the name of the speaking engagement that you have at

C2C so this is It's actually kind of a big deal so this is kind of Google's first official post next event um and it's one of the first ones I can really think of uh at the Google campus uh here in Seattle um it's one of the newer buildings it was opened right before the pandemic um and there hasn't actually been that many uh events held there uh because it just reopened you know within the past 12 months and so uh this is held by C Toc shout out to Marcy and the entire team there for all their help we do a lot of these

with Google Cloud at multiple campuses and sites um it's a gathering for both customers and partners and googlers to join and they have a curated session list of I guess speakers that I'm humbled to be a part of um that kind of focus on specific aspects and so I'll be covering both uh Cloud Primitives and optimization as well as recapping some of the major highlights uh from Google next uh our friends at Splunk uh as well as a number of customers like Alaska Airlines will be presenting and walking through their gen case um it's be pretty cool and we'll probably do a few more of

these throughout the year so be on the lookout for any of these kind of Road shows that uh might be in your neck of the woods awesome well thank you thanks for uh go representing ciq at the Google Play appreciate all of that good I'll be I'll be in my my Rocky swag and Air Force One so I'll be showing up cloudy as ever nice that's amazing all right you guys so uh you know if you have any interest in a um kernel an upstream kernel and Rocky n we would love to chat with you more about that about optimizing your performance and your security

and your what's the other one compatibility is that the other word ease ease of use ease okay easy that's sort of yeah that that it works it works all right you guys so we would love to chat with you go to C iq.com you can fill out anything there and it will come to us and we'll get back to you we just absolutely appreciate you thank you so much for being here thank you guys for showing up and all the good work that you do and everyone behind the scenes making this happen because not easy not easy so we appreciate it thank you thank you

we'll see you next week same time same place thanks everyone [Music]