Bestand wählen
Merken

DevOps for Networks, NetDevOps, NetOps or Whatever: Get Your Network Cooking With Chef

Zitierlink des Filmsegments
Embed Code

Automatisierte Medienanalyse

Beta
Erkannte Entitäten
Sprachtranskript
I'm excited to be here this is my 1st chef golf and how
the Orissa networks as had a series of integrations with shelf over the years but as we're trying to take a new tack on that now and that's been a really fun thing for us to do because chef makes it incredibly easy but the real thing I'm
here to talk to you about today is is not sales or marketing around what we do but is how we can use the self community and the experiences around Dev Ops and to bring our network folks in the fall the so versus a little bit about me Gergely and accessibility engineer which means I get to play with a lot of our cool automation a lot of our API is improving are API as an abstraction layers and working with a lot of integrations with different different tools and and chef is 1 of those now I've
managed networks before I worked for several network vendors and in my free time I like to do crazy things like run towards burning buildings a sort of away from the so you know and climbing steal both arenas in such along and contrary to some network folks I prefer to spend my weekends and my evenings my holidays outside in the freezing cold data center of like to be growing with the family of like to on boat so hopefully we can give some ideas here that might help you or your network team yeah to do the same thing so if you do everything in the cloud that's also in you might not be a lot about what a lot of this presentation and if you really rather spend the you your your holiday weekends attached to a Council cable 1 data center that's cooler respect that and I'm not gonna stop you but there some other alternatives but I'll bet that maybe not a lot of you or full-time network folks but maybe have a network or you have some people that said hey can you figure out what can we do anything to about this network equipment on so hopefully you guys can have something to take back just a curiosity how many people in here actually manage network equipment excellent that's actually of far larger number than I really expected I know a lot of times ended deluxe environment it's it's all server enact folks in the well yeah Manila guy he's he's these back home somewhere all undergo passes the from so that's great but having as here you the that's a great question OK alright cool we got 1 sort of alright that wasn't very funny
memories of of color sort if so a was starting by talking about some factors in fiction around at work the love you will think that Knock engineers when we talk about automation that they have to become developers that's absolutely not the case don't get me wrong there's some places where you learning some Python learning some Ruby understanding some basics of things like get for version control and that sort of thing can be really really helpful but you don't need to become a developer that doesn't need to be a full-time job there are a lot of people also think that the network it really playing style lots world so the equipment's totally different totally different user interface from your typical servers your application configurations how we do that or they think that the equipment just can't handle and then the most damaging 1 of all there are a lot of people out there that say if I only my network that I can't do that only network interesting 1 that is a very key mistake because the truth is that none of this works without a solid network design network engineers are specialists and we need to make sure that our network is resilient they can handle the bandwidth requirements build the latency jitter requirements and the time requirements of our business is not just about getting data from point a to point B we do that and hopefully we did really well but it's about supporting everything else their business needs to communicate and to do its job and to make money the the so we also can really benefit from revision control around the network you know or configs a little bit different than being what you typically see answer server world still if you take nothing else away from this please take your network configs and go put them in a revision control system have a
backup have the ability to go back to what did it look like last week what did it look like your last month or at least what it look like last nite but when we start learning how to represent the network as you've heard the phrase infrastructure is carried the network is part of the structure when we do that it becomes code code automatically belongs in a revision control system that gives us the power to control who has access to that who can promote things to the next level and it gives us the ability to go back and look at our very easily see what changes being requested exactly what z effect and and and build to approve that an applied in a consistent manner 90 talked a lot a network guys make the copy and paste in a little tweak here that's that's pretty close the automation right I'm not taking it all by hand anymore OK some people get that I mean yet I'm pretty sure we've all been there but that's not really the best way to do this if you look at the the tools the technology that the we've been talking about here all week you look at the ability to go and will say to a of configs session on a switch using that version of software running on that's which you can verify this desire configure will parts correctly it's not quite as elegant as a nice linter for another language but it absolutely works in absolutely can be automated similarly with good know his equipment we got the ability to spin up entire virtual environments including switches routers load balancers servers all that there's something as simple as a vagrant up command the run your automation against that verify that that what you Bill is going to deploy successfully among pull it back out you really go up to your next level of tear your staging environment to production environment what the same concept applies the servers applied applications can work you know world you know and I'll show you some examples of some steps of how to get there but the key here is to take your network and make it more consistent make your changes more rapid and improve your customer service you have internal customers which winning you improve the service to those customers you improve the responsiveness those customers you accelerate your business I worked plenty of larger price customers large financial customers that it takes 18 months to roll out a new application and MIT 6 weeks just to roll a the land into the network between the approval cycles and in all of the different and blessings of they have to get and scheduling of maintenance window but with modern network designs and especially when you look at things like How restore and build applications to to tolerate the putting applications behind load balancers distributing them across multiple racks multiple machines multiple network devices so now let's say you have to take out 1 network device for the complete upgrade and depending on your device that could take 20 minutes it could take 50 ms and depending on what you got but we now have the control 3rd to 1 lecture to really go and remove a set of systems from a load balancer take that stick that rock offline do all the upgrades you need make all the changes you need bring them back online Abbas back into the load balancer automatically your customers now had not noticed effect and there are other designs we can use to the same the same kind of thing same kind resiliency the key is good network designs working with your application teams have good application designs tableau hearty environment so how we get there what what is
the network it is the thing that transports the bits sure is the thing that we all complain about when we can't get to whatever our favorite website or application is but for your business is how your customers interact with the it's how you make money you to your applications don't work without it your databases are really great but isolated and on an island without it you can't I mean how how well how long does your business operate when the network staff How much money do you lose In lost opportunities when the network is down so it's a critical resource so we have to be mindful of so for those not familiar this is your typical network in Fig on cut short but the the best come over going typically we have 1 monolithic configuration file is not even really a file but it's this virtual struct that we interact with like it so far but then there can be challenges with staging that are changing different uh different services at different times of such but when you look at the network operates to serve different services just like your service to some of them are gonna be familiar to the server teams some of those might only be recognized by folks that are actually network practitioners but they're just services and we can break those down In the things that we managed individually from there we can take that thinking in step back and look at our environment and figure out how do we find commonalities across our environment use that to build a data model for what you actually need in your configurations I know what I mean by that things like into k syslog can you guarantee me that every network device in your environment is singing information to the right syslog server has the right time configured on it that sounds simple and even using Scheffer while this but for those people in our world were still not quite there yet so find those commonalities you you might have a set of racks then it are a four-year be aware cluster for example all those hosts have a consistent configuration or most of them have a consistent configuration from the network side use that consolidate that into a data model that then you have 1 place to make a change when you need to make a changed your entire environment that's a really key piece
because if you get that right and then then you use good cookbooks to implement that now when you do need to make a change is undergoing a change the configurations which you have 1 1 of data to change now that could be in your you're can system that could be in some database somewhere could be entered Jason young will follow but were network I so it's probably in spreadsheet somewhere was be honest but yet 5 minutes on StackOverflow and a couple lines Ruby your chef cookbooks can pull that data right of that spreadsheet that's really what works for you that's not a problem you wanna make sure you protect that data but your data can really live anywhere that works best for you you just need to use your chef cookbooks to get that from its home the on 2 devices so now once we reduce the scope of the change down to just was say a lot of data we've also reduced the scope of approval we reduce how difficult it is for someone understand what this is going to affect when they say 0 you're you're not changing you know 5 lines of conveyed on the switch on each of these 300 switches you're adding 1 be land to the environment that's it and you're doing in a consistent manner the way everything else has been done it's low risk so that's much easier to get through our approval processes and we can start working to break down these massive approval processes that we had over the years when you reduce the scope of your changes he also reduce the risk again this with experience gives you the power to help breakdown these massive change management words that so many of us our our involved the and quite frankly they're 1 of the largest impediments to us moving at speed they're there for good reason there was a problem and that problem calls an outage that outage cost money and we want we want to prevent that both good automation good consistent environments we can do that with out all these manual steps and actually we can operate much faster and in a much safer way this quite frankly as a network I I'll tell you we are all worst enemies the majority of network outages out there are not caused by softer failures are not caused by hardware failures those do happen to know that the majority of network outages that monetarily impact of business are due to us making a mistake or dealing with an unexpected interactions when trying to match together too many changes in that 1 change window that we get for the whole quarter the absolutely Larry fellas so on the steps so we can take to to get there so just like a server you got a bootstrap network devices now this might be something you doing as a coming out of the box in a green field environment this could as easily be something that you're doing around film of environment and very commonly ears you gotta get basic management can that you got interval your API is on the box and depending on your server you may need to put the subclass on this on on the network devices there all are actually some very interesting options available with certain cookbooks now to where and certain vices a have a good extremely accessible API is to where you can actually avoid having to place that's of client directly on the device you can actually had that run on sort of a proxy noted and in your recipes specified what device for which devices that applies to the the I just what the and of 2 and the well it depends on depends on the upper existed so long so yet so the question is is it even possible if I summerize is even possible to put the Klein on optimized and it depends on device I so with uh with some the Cisco products they did some of those do you have the chef-client available for them on what is a custom package and uh you know there's a there's a process for putting on the arm when you look at something like Aristo work most some were based on Linux we're just their Linux systems with accustomed set of processes custom set of CIA lies they're sitting on top so for onerous device for example you download the standard Linux chef-client and install at the API and assault right on your device and your reader from that point on it operates just like a server with anywhere from the 48 a couple hundred network interfaces but yes and this on of I I do the OK so that so there's a there's a good possibility as long as they support learning third-party factors there's a good possibility use that the the of so we have a couple arousal we can go to with managing the network for folks that is seen the server world wrong in use whichever you're probably familiar with discrete resources these resource types manage if a is small subset of things after a very precise grouping work work work type of thing and this is really great this is really was really powerful it's very granular it gives you the ability to start by just managing some pieces of your device and gradually expand as you improve your processes is as you improve your experience it also is a great way to just demonstrate what I was talking about about possibly being all box so if if you're able to build that into your cookbooks as as we have then you this indicate the device name by default it's going to try to communicate with the local box running but if you specify a different last then you'll have tell your client on each individual box now we started down this road back in 2013 working with Juniper Networks and started a cookbook called net the the pRecall concept and back in that time chef resources were made up of 2 parts there was the resource type and I was the provider and you get a single resource type of multiple providers because the provider was the implementation for a particular operating system what that meant was we were able to define a single the land resource type and Arista could create a provider Juniper compute could create a provider anyone else could as well the for some reason that didn't really take off and there there certainly are advantages to having you your your custom resources for your own devices but so we were a little disappointed that didn't take off better in you know we were we were I guess we're a little bit too early to that for the network side and a a little bit too much sharing for the network effects at the time but is a good concept but also the network folks sometimes will come and and we saw what the traditional network configure looks like and when you try to abstract that out of this the there is a big learning curve there so another way we can do this is with templates if you imagine just taking the standard running configure on a device whether so Cisco device and rest device whatever I take that standard running configure the network team is comfortable with they've been dealing with for years and think it is translate that into an URI template now we have something that's comfortable for them to transition to and brings them in the power of automation of the power of the testing the power of the all the control we have from chef from del ups team more things but at the cost as you get more comfortable with that is so just for placing a few things with some variables you can start building logic and others templates and with that logic in those templates you can think now drive that with a fairly simple datamodel but you can simply list ear layer to porch 3 parts the NTP servers name servers and your template drives that into your standard Running configure output they can then be tested with the path from the box validated automatically and then deployed at which 11 which brings us to the continuous integration this works on the network to most modern network vendors now having a a virtual version of the for some vendors it is a new V M that resembles the operating system on the devices for other from other vendors it's the full-blown copy of the operating system the only difference is just a device structure and the fact that in a b and you'll have those there's a 684 dealing with so there's a little a little bit of fudging not area but as a purely the hardware abstraction at that point anything that you can do on your physical not devices you can deploy in the virtual world for test environments simple they up can bring up this this 5 diems pre connected run chef cookbooks honors the news inspect to verify that things converge the way you thought they would that the right relative there that the right interfaces are up bright settings are in place and they were destroys gone and as writing your CI pipeline that comment gets added up to get help with your pull request and now before anybody has to review any code changes is already been tested it sorry there is validated you know that it'll work again something we've seen from the network so our network side for years but it can be done with the network as well so or back to why but and you see played before praises for those Net UPSnet develops it's just the much happens we know what people using but as a community if you're not if you're network person and you're here today I applaud you also you've taken a huge step if you have now a team that's back home and you're trying to figure out how to improve things for them how to improve things for you the perhaps the impediment to what you're trying to do hi you're trying to improve business let's bring them into the fold this is a community Dell UPS is about community let's reach out to them and was seldom understand that the tools are there the components are available all we need to do is start building these pipelines bring things together getting everybody talking working together and we get the network as a key player at show cop next year I hope so same things again that applied to the server world act world start small don't try to take over the world 1 step worth is selecting the right tools are you going to use are you going use Jenkins or some other continuous integration environment what and you know water is a control system the worries in the grand scheme it it doesn't matter as much but it matters that you pick something you start working down pat takes a low risk and high value configuration items and get them under automation control then when you already we started getting comfortable with those when your management the rest of the team is comfortable with that start expanding that out and continuously growing the the I got love what every licensing all week about how hideouts so I know we get frustrated with some but go hug a network engineer go tell me it's OK to come out and be part of this style of world yes and no shut up maybe what do you have any questions thank
Reelle Zahl
Reihe
Neuronales Netz
Integral
Neuronales Netz
Nichtlinearer Operator
Server
Subtraktion
Bit
Kontrollstruktur
Abstraktionsebene
Gebäude <Mathematik>
Gefrieren
Zahlenbereich
Quick-Sort
Integral
Rechenzentrum
Software
Datenmanagement
Ablöseblase
Server
Äußere Algebra eines Moduls
Addition
Neuronales Netz
Programmierumgebung
Streuungsdiagramm
Neuronales Netz
Stereometrie
Punkt
Virtuelle Realität
Formale Sprache
Programmverifikation
Versionsverwaltung
Kartesische Koordinaten
Datensicherung
Übergang
Online-Katalog
Softwaretest
Prozess <Informatik>
Code
Autorisierung
Bildschirmfenster
Router
Neuronales Netz
Softwareentwickler
Güte der Anpassung
Bereichsschätzung
Teilbarkeit
Softwarewartung
Scheduling
Dienst <Informatik>
Menge
Einheit <Mathematik>
Rechter Winkel
Festspeicher
Server
Versionsverwaltung
Schlüsselverwaltung
Programmierumgebung
Subtraktion
Kontrollstruktur
Mathematisierung
Dienst <Informatik>
Datensicherung
Code
Lastteilung
Stabilitätstheorie <Logik>
CLI
Physikalisches System
Virtuelle Maschine
Multiplikation
Software
Fokalpunkt
Endogene Variable
Ordnungsreduktion
Softwareentwickler
Datenstruktur
Konfigurationsraum
Leistung <Physik>
Soundverarbeitung
Benutzeroberfläche
Mathematisierung
Physikalisches System
Quick-Sort
Summengleichung
Mereologie
Dreiecksfreier Graph
Gamecontroller
Stereometrie
Bandmatrix
Kantenfärbung
Bandmatrix
Neuronales Netz
Virtualisierung
Datenmanagement
Service provider
Client
Typentheorie
Regelkreis
Neuronales Netz
Gerade
Schnittstelle
Softwaretest
Hardware
Güte der Anpassung
Bereichsschätzung
Ruhmasse
Kontextbezogenes System
Biprodukt
Dienst <Informatik>
Menge
Rechter Winkel
Server
Client
Ultraviolett-Photoelektronenspektroskopie
Routing
Programmierumgebung
Subtraktion
Wasserdampftafel
Mathematisierung
Dienst <Informatik>
Mathematische Logik
Unternehmensarchitektur
CLI
Variable
Lesezeichen <Internet>
Netzbetriebssystem
Datentyp
Vererbungshierarchie
Installation <Informatik>
Datenstruktur
Konfigurationsraum
Soundverarbeitung
Green-Funktion
Datenmodell
Elektronische Publikation
Netzwerktopologie
Gamecontroller
Demoszene <Programmierung>
Wort <Informatik>
Neuronales Netz
Bit
Punkt
Prozess <Physik>
Gemeinsamer Speicher
Bootstrap-Aggregation
Versionsverwaltung
Gateway
Kartesische Koordinaten
Template
Online-Katalog
Datenmanagement
Bildschirmfenster
Kurvenanpassung
Default
Schnitt <Graphentheorie>
Figurierte Zahl
Diskrete Untergruppe
Funktion <Mathematik>
Datenhaltung
Template
Abstraktionsebene
Nummerung
Teilbarkeit
Konfiguration <Informatik>
Teilmenge
Datenfeld
Tabellenkalkulation
Information
Schlüsselverwaltung
Standardabweichung
Proxy Server
Web Site
Quader
Stab
Gruppenoperation
Interaktives Fernsehen
Implementierung
Code
Stabilitätstheorie <Logik>
Multiplikation
Zusammenhängender Graph
Ordnungsreduktion
Widerspruchsfreiheit
Leistung <Physik>
Einfach zusammenhängender Raum
Zusammenhängender Graph
Kontinuierliche Integration
Mathematisierung
Einfache Genauigkeit
Mailing-Liste
Physikalisches System
Quick-Sort
Inverser Limes
Mereologie
Modelltheorie

Metadaten

Formale Metadaten

Titel DevOps for Networks, NetDevOps, NetOps or Whatever: Get Your Network Cooking With Chef
Serientitel ChefConf 2016
Autor Julian, Jere
Lizenz CC-Namensnennung - Weitergabe unter gleichen Bedingungen 3.0 Unported:
Sie dürfen das Werk bzw. den Inhalt zu jedem legalen und nicht-kommerziellen Zweck nutzen, verändern und in unveränderter oder veränderter Form vervielfältigen, verbreiten und öffentlich zugänglich machen, sofern Sie den Namen des Autors/Rechteinhabers in der von ihm festgelegten Weise nennen und das Werk bzw. diesen Inhalt auch in veränderter Form nur unter den Bedingungen dieser Lizenz weitergeben.
DOI 10.5446/34628
Herausgeber Confreaks, LLC
Erscheinungsjahr 2016
Sprache Englisch

Inhaltliche Metadaten

Fachgebiet Informatik
Abstract The cloud players get it (often with custom automation by a huge development team), and the people that do everything on AWS don't care. But what about the rest of us who manage real networks everyday? What does DevOps mean to us? Is it just the latest fad? Where do I start? Come see why there's never been a better time to start taking advantage of DevOps practices and tooling to make it easier to run your network.

Ähnliche Filme

Loading...
Feedback