Bare-Metal Meet Cloud-Native

29:35 · Watch on YouTube ↗

Transcript 4,766 words · about 32 min to read

Auto-generated captions from YouTube, not hand-corrected, so names and technical terms may be imperfect. The video is authoritative.

m. I basically an intelligent platform management interface IP mi and it turns out that that's a one standard but there's other standards as well and when I was in the lab 20 years ago it wasn't actually this technology be used we use wake-on-lan which is it predates it the way con man is kind of like very very early days just I'll send you a special Network packet and you turn yourself back on or if you're suspended you come up this one is pretty pretty awesome I first time I use it was with an impaired server the guys over there they sent me a server like this big the photos in is in this actually it took a family-size dining table just to plug it in much like your ones I don't have a rack I don't even if anywhere to put it I mean it's super noisy but it comes with this and so downloaded a Java client I connected over the network and then I mounted an ISO from my computer to it and I installed Ubuntu and that whole experience was like something I never done before and immediately saw that a value in that out of band console and management right we we we live and die by it in the data center you there used to be a time where we would have a KBM keyboard video mouse adapter with you know the fancy ones would have LCDs that lip up but basically we'd have a vga keyboard and monitor connected to each computer and that's just an inefficient way to run a data center when you have 42 when you can you now practically get 42 use of storage me up use get into a rack that you're not going to get 42 cables now you know that is a key technology and that is an advancement it's no by no means is it new it's just a newer and there are other ones from other manufacturers which kind of sort of conflates the thing and I bought a HP micro Pauline a few years ago and actually turns you've got to pay to access the out-of-band console a license fee which isn't which isn't great but anyway if you combine all of these booting services like your DHCP your NFS your TFTP your I pixie you can get a workflow like I showed you on the Raspberry Pi the challenge is how do you make that secure for the enterprise so if you're secure for production something that cloud native and it feels cognitive I think one of the keys is it should feel like you have a pool of capacity that in theory could be infinite if you can add to it and at any time you can use it as if it was all homogeneous right despite it being actually heterogeneous we've got Raspberry Pi Intel Newark Dell servers and PI R on servers the whole works in there we want to treat it as a pool right and we want to be able to say I've got this capacity now I need a spark server now I need a kubernetes master now I now I want to run and get lab CI runner and to be able to request that by API or by some kind of workflow and have it declaratively brought into existence by the system that's that's kind of how I see the cloud native approach on it yeah that makes perfect sense from a data center as me as a data center provider that has always been the nerve on the the in unit like we tried to do this with virtual machines basically the concept that hey you know what we can deliver api's to the hypervisor from the hypervisor to the to my developers and then they can provision machines and practice that just never got adopted because managing virtual machines this to your server list concept who wants what developer wants that meant servers none of them it turns out containers passed pipe solutions become much more digestible from a per perspective however it puts stresses on the operators and this is where we get into this developer versus operators model with the CNC f it puts these stresses on the infrastructure folks that we just don't have the tools to give developers the the type of control they want over the infrastructure or this friction has experienced this is again another promise to that mix of saying hey you know what as an operator I can I can deliver a match and deliver an experience that developers are wanting to consume alright this is you know I have you know that I probably love-hate relationship with two minutes I love that that's the potential of the project and when you layer something like open pads on top of it is it's wonderful my problem is the merging of these two worlds oh my goodness that is that's that's the TT I think you know when you look at it there's probably more levels to this tonight than I imagined to begin with this there's a level of I'm an IT operator with deep knowledge into bare metal and provisioning tools how do how do I give my developers what they wanted a the other side of it is how can we make our existing estate look and work like a cloud developers worn infrastructure as code I've sort of highlighted digital rebar here one of the things they they're really keen on his infrastructure as code maya iook for sure this idea that you can use a tool like terraform from actually court to declaratively right I want this server this memory install this application you run a command and it gives you it you then want a little bit more or another server you just add it in your file we tweak a value you run it again and at any point if you lost the servers of what happened you have to switch to dr is when you script again that's it and all times there's no clicking in you eyes there's no sort of manual steps we're not tied to a particular server we just got this pull and we've got this code and github you know and I'll get up and we just run a command you get back to where we were and so digital rebars trying to do a little bit of that and you've got tools like Mars from canonical that's just sort of automate what's already there ironic I've heard really good things about from the OpenStack foundation despite means that tooling sort of been sort of known now for being maybe a little bit over complicated people saying that it failed for what it was designed for things like ironic or sort of a phoenix from the ashes where there's actual value there that you can bootstrap your cluster and then get kubernetes on it and use your top level projects uncle mob yeah a couple more there and plunder is is quite a fun sort of community project that one guy sort of scratching an itch in his spare time Dan but Tinkerbell was kind of this one's caught my eye is from packet and I've been playing around with it and what they're trying to do is sort of take her take the experience of the last six years of doing bare metal by API and open source it and it's anything but smooth and there's a few teething problems as to be expected that it's a genuine right they do have the experience they've done this stuff and so I'm quite I'm kind of very interested in that I'm going to be speaking on it next week it's software circus and then other tools that are interesting like metal case metal stack metal cube there seems to be a next wave of projects so unlike these general-purpose the only goal is gimmick kubernetes on bare metal that's it but I want the my I don't want a hypervisor I don't want OpenStack I've got these computers I just want to keep an eighties on it for the best possible experience yeah Alex I really appreciate you coming on and sharing this practical experience for those people who are thinking okay well I have it now I can go out and build a data center in provision this there's still a lot more to do Alex connotated into it especially if your focus on kubernetes but there's real challenges because the infrastructure and most enterprises is either based off of I study fibre channel or NFS obviously how how do you leverage that existing infrastructure to deliver permanent storage precipitant storage to kubernetes there's that challenge to overcome and then how do you do it in a way that it's friendly next to your Oracle and VMware workloads etc so taking this generalized works that in applying it to a cloud native environment in delivery method that is that's the work and it's good work to do you know I pick one kubernetes a lot is not because I dislike kubernetes it's because right now we don't have the tooling and a skill sets in the open market to to fulfill the dream that I love sponsoring folks like Alex and this is a plug to sponsor Alex over on get up there's a couple of things that people could participate one is if you've got a brand and you want to connect the developers get a link back from a high-ranking site come to open files calm put up a home homepage once a ship that's one thing definitely good for big big brands the second one is that there's an insider's program on running and I'll send an email out with updates and all the open source software but also in Cuban AES on Raspberry Pi and go and do a little editorial last week was all about content marketing and what you would learn from that how you can apply even on an everyday job as an employee just go on there see it tear that's that fix what you want feel free to sort of join in and you'll get that premium newsletter you also get to sort of contribute towards the safe install software and the future of it yeah we have this saying in our community different community either go or you pay to have someone go so I fer to pay to have Alex do the hard work of making kubernetes consumable and this is a great example of that work out states for the great work in the community if you want to learn more about the CTL advisor in what we're doing with the data center you can go to the CTO visor comm we will launch a microsite CTO advisor DC dot-com our CTO a DC in the next couple of weeks so you can learn much more about the data center this content will probably be there until then talk to you next CTO DOS thanks Kay thanks everybody