BIG-IP DNS Load Balancing Introduction

hi welcome to this latest episode of light board lessons and today we're going to reshoot a previous video on the global traffic manager which is now big IP DNS and so let's before we get into all the load balancing algorithms which will consider for future videos want to go over just the basics of how the big IP DNS and the local traffic manager interact and so when you have the clients out here in the and the internet and you come in to say a data center and then say we have another data center over here we'll call that data center to of course come that way from the cloud in any particular data center you have your local traffic manager advices in either an H a pair or you have a device cluster where you have many active in a standby or you know whatever your scenario is and so you have your infrastructure layer here you probably have some routers and firewalls in here and until will have your L PMS and so those connect down through infrastructure to your web farms and so you know that's sideways so you know any number of web farms in here and and say this is hundreds of servers or whatever and so on the front side of these LPM you have virtual servers so you say want v1 v2 and then that's mirrored over here on this on standby or maybe you have a completely different business unit that maybe has an LP M over here and and that's in a sync group with this LTM in the middle but you have another v3 and v4 okay and so all these different virtual servers exist for any number of applications that sit here in the back end and over here in this other data center would be hopefully at least a highly available experience for those same apps so in this case we have say we have LPM and then we have another LPM for a different business unit and just assume this one also is an H a and then you have other virtual servers over here we'll call this v5 and v6 and then b7 and b8 okay and so when it comes to bringing these clients to these particular virtual servers whether it's IP or ipv4 or v6 that's where DNS comes into play and so traditionally before the world of load balancing it even with up you know some basic load balancing you can serve those records with one or more addresses and so in the old days you would say I have that services available on four addresses so I'm going to go ahead and send all four addresses out to the client that's still possible in the clients just going to pick one of those from the query that they receive and of course it'll cache it for however long your TTL is when that TTL expires it'll go out to the L DNS server sort it out here in the cloud so you have L DNS times you know however many l DNS servers that these clients have access to and that's dependent on their infrastructure that from their provider and then those L DNS servers will cache as well for as long as these records tell it to cache and and then when that cache expires it's going to come out and it's going to look for an address and that's where on our n the GPM comes into play and so if we extend this data center I'm sorry it's an old habit GTM actually f5 big IP DNS so we'll just clip the dns in there over here in this data center as well and in the GTM I'm sorry the up five big IP dns will respond to a particular query so saying this client is looking up dead central dot F five.com and so they're they're asking that question and so they're going to ask that of their own local cache if it doesn't have it it will come to the L DNS and at the L DNS doesn't have it then it will based upon how we've set up this start of authority for f5 comm it will eventually make its way to our f5 DNS device and in here this is where the configuration like for your local traffic managers you have your your pool your pool member your virtual server the DNS sign it starts with your server and so let me actually we'll start with data center and the data center you would define here and then the server would actually be this LPM you can use third-party devices as well but we'll stick with with f5 and so your server would be one of these LPM and then you have virtual servers from those servers that will then be either discovered or manually added to the configuration for GTM and then that cut becomes the constructs for pool so just like at the LPM level where you have individual services as pool members here you have a virtual server as the service in a pool and so if we have say we have one particular app and all four of these virtual servers are in at particular app we can say that will put these virtual servers in that pool and V 1 V 2 V 5 and V 6 and so this would be let's call that pool one and then pool one would be part of a white IP which we would define it could be a wild card you don't have to put dev central f5 calm but this white IP would be listening just like a virtual servers listening for a particular IP port combination the GTM listener is the highest attendance of listener is listening for this fqdn and so when it receives that there will be potentially a lv decision here so if I had more pools saying that in this particular one I have a pool to a pool one and a pool three at this level it's going to make a LV decision for which pool I want to virtual servers I want to send it to and then within the pool there's also an LV decision that is going to be made here and so it's a tiered decision whereas on LP M it is just a singular decision at the pool level with GTM it is the LV decision at the wider P level and then the LB decision at the pool level and so you know that's the the basics before we get into the LB decisions themselves on on the tearing of how those decisions are made and and how the overall architecture of the infrastructure works when it comes to global traffic management and so hopefully this has been helpful and we are going to get into the load balancing algorithms in our next set of videos thanks a lot and we'll see out there in the community you

Channel: F5 DevCentral

Views: 42,160

Rating: 4.4609928 out of 5

Keywords: f5, devcentral, lightboard, dns, big-ip

Id: 9fooqDbwJIk

Length: 9min 1sec (541 seconds)

Published: Wed May 03 2017