Network Design : OSPF Scalability and Multi Area OSPF Design

[Applause] [Music] so how many routers I can place in one or SP area right this kind of things I will talk about in OSPF frequently asked questions but if I want to it is mineral tourism as I want as many link between the routers as I want in one OSPF area then I am talking about scalability ok so we always talk about material design right but when we have lots of routers in one ways bavaria so we will have some flooding right and depends on the failure rate between the routers link failure node failure depends on the rate of this failure we might have some problems on low-end devices if we have in the same OSP area so in order to reduce the flooding impacts we can have some something like multi really designed to provide stability but multi area OSPF design is not the only thing for scalability right they have hardware guys hundreds of OSPF routers can be placed in one OSPF area in fact on horan algún net i talked about a prefix operation feature which i will talk about now which protects traffic separation features an operator cell providers they placed i think 200 250 you can have even more routers in one oil spill area without even single sorry without even multi or US v design they are using single OS bavaria big finale OS f design so as i said o s v Monteria is not the only future you know SPF scalability and we will talk about now other things as well but scalability how we can basically place more routers more links in OS bavaria how we can have less CPU less memory usage on the routers not only those but also configuration wise how we can have basically less configuration and in for the operation how we can reduce configuration complexity etc so we will talk about them these all things ghost scalability OSPF scalability when we talk about BGP we will mention other things like route reflector and other things how we can place more routers how we can reduce to number of sessions here how we can place more routers in OS bavaria how we can basically reduce the control plane load ok topology information etc we will see now so one feature is OSPF prefix suppression ok I mentioned in rattle Ellis a type analysis each and every router in OS period they create type 1 essay and they identify their connected prefixes their neighbors and between neighbors the OSPF metric so all this information is sent and not only for basically con point-to-point links between two routers but also for loopback but also for any other stop subnet so but for OSPF for SPF algorithm we don't need this IP information of the point-to-point link ok it might be just unnumbered but still OSPF basically can calculate the SPF can calculate the topology so with prefix operation future we are removing IP address information from those point-to-point link so two routers you have and connecting wire one link and you assigned on this side on router a 1x to 160 of 0.1 and other side 198 0.2 you don't need those addresses you can have do peg interfaces for many other reason like BGP over name and IP cetera but you don't need this point point address and that's why when you reduce this when you remove this information IP address information on point-to-point link and you remove from the routing table and instated database basically you have less information and basically it is not only for scalability of the rocking table and your link state database also for your troubleshooting it is good right less information to deal with good for troubleshooting if you can have you can do the troubleshooting faster which means availability also increases because for the high availability in network design tools we talked about what are the attributes mean time between failure and mean time to repair right mean time between failures not related with this but mean time to repair related with troubleshooting time right how long it will take to repair the failure so troubleshooting if you reduce the time for troubleshooting you increase the time for availability it is good so that's why this prefix operation feature is the important feature and operators they are using by the way this feature of course if we are talking about just couple routers 5 10 20 routers we shouldn't deal with with prefix operation or multi or away or SPF design as well we are talking now large scale that's why scalability large scale OSPF design and this is one of the features also for full mesh topologies ok let's say you have you connected 10 routers 10 data centers in each data center you have just two routers for high availability right five data center and you want to connect those five data centers together and you receive VPLS service so 10 routers now will be in the same OSPF area and 10 routers now will be in full mesh topology if you are receiving VPLS in that chest so what you can do you can basically with OSPF there is a feature called database filters it is very similar to mesh group I will explain I say as mesh group or you can jump just to assess mesh group and you can check that but basically the idea is you will have two or three daughters which will do the flooding on behalf of other routers so in full mesh topologies then this database filters this future provides flew the introduction so then it will also add extra scalability to full mesh networks that say full mesh topologies so even though with the prefix operation and this the database filter in the full mesh topologies still we might have some problems because number of routers in OSPF area there are some limits not maybe numerical limit I cannot give you numerical number you should place 50 routers maximum in one OS paper etc we cannot say that but the problems comes with the router LSA size we have linked empty you write 1500 most of the time by default and you increase if you have some encapsulation some VPN even MPLS at extra 4 byte for each label so when you have link MTU size 1500 let's say and router LSA I think 24 byte for each LSA and what's happening with each and every extra router inside the say single OSPF area they will create route release a and your router LSA size might be bigger than link amp you when it's bigger than in came to you yes it can be fragmented but problem is fragmentation because fragmentation creates CPU issues on device resources will be used more and other device also will do the reassembly so they will basically put in an order so it is not good fragmentation you should avoid as much as you can that's why the idea with the number of routers in one OSPF area number of routers in one respect area scalability limit most of the time we talk about router LSA sighs okay just keep on keep an eye on that and because each additional link and subnets as I say we'll be added to the router LSA and you don't want to have redistribution also we talked about some Aria types right special area types like stop totally stop in SS a total n SS a I told you between those backbone and non de canarias there is no topology information exchange so in one area and SSA area let's say the routers in an NS si area doesn't know the topology information of the area zero which router is connected to which other router and metric information etc in the array zero so in that case when you have this kind of special areas there is no also some some of the reachability information out I should say some of the reachability information will not be exchanged so in stop area just as a stop area but we don't have type for n type five right so in that case what you are being fault isolation so in if link or not failures inside the array zero or in the stop area then this information will not be propagated between the areas that's important they will not have full SPF calculation in inside the area if something happens in another area fully SPF calculation versus partial route calculation basically has different impact on the resources on the CPU and memory okay so that's important when you have full SPF it will create more CPU it will consume more CPU resources that's why you want to avoid that one as well if you have lots of routers why you want to place in some routers especially low end routers more resources routers inside this nomicon area such as set up totally stop so you want to avoid topology information and you want to send is less information as you can inside the dead stop areas or low-end devices when there is some problem inside the backbone area or any other number Canarias they will not basically run full SPF etc they will just change the metric if there is a change in the topology etc but not fully spare calculation which would create normally a lot of CPU resources on low-rent routers and they might even crash look at this topology we talked about area types we talked about LSA types now we are talking that areas different areas basically used for the scalability right and Levant routers we can place in Nam backbone area and we can make that man weapon area set up totally stopped in SSA and we don't need to send each and every prefixes from the backbone area but just summarized maybe not type 4 type 5 not type 3 so on and so forth but let's use all of our knowledge now and let's understand what's the problem with this design basically I just intentionally created a problem but let's understand what's happening okay so let me just explain what is this topology we have area 10 20 connected at the top and in the middle we have backbone area and area 30 area throat is connected to a IG IP domain what I am saying at the right side area 10 is regular area that's why all the LSA types including type 3 4 5 are a lot so as you can see type backbone area standing type 3 4 5 if it is normal area regular area so we are not talking about stuff totally stopped in SSA but just Aria 10 different an area 0 but just a regular normal area 10 yes everything is a lot type 3 4 and 5 from the backbone area would just entered in area 10 right but backbone areas type 1 and type 2 LSA how they will send they will be sent to the area 10 they do not send of course - and type tube those Ellis's are special to the dead area but they are summarized in to type 3 and send to the area 10 that's why normally of course type 1 and type 2 of backbone area will go to the area 10 is Type three okay I am four and five from the externals domain they should just come they should come from external domain so what are also we are saying ABR's create high policy into an area 10y for each and every type file we have also not each and every type five sorry for just a spirituality we have time for a lesson in this ABR creates this type for NSA and send into the area time area 20 I am saying stop area that's why only type 3 LSA is a lot is it true so if area 20 is stopped so we don't have basically they said area 20 stop saw a real 20 stop then type or n type 5 is not - and type 3 LSA from the backbone will be sent by the way even if you connect area 10 and area 20 routers directly they cannot send the LSA directly because those LSA need to come true area tech area 0 sorry they probably right so we know that directly that they can not communicate but they should communicate through backbone area what else saw area 30 is there a problem here what I am showing here gij Arceus Anand is type 5 LS a so if y IJ is standing try 5 elicit an area 30 can not be stopped or totally stopped right because in total stop and stop we cannot have type 5 LSA also if it is type 5 Allah say it cannot be a necessary or totally necessary because we can do redistribution Romeo J our people are you 30 with an SSA and totally necessary ax but those NSA would come as type 7 so since we are showing type 5 Allah say this area 30 should be normal area regular area but I am also showing from area 0 type 3 LSA only if it is just regular area normal area Y type triela say only and I didn't mention any filtering etc there is no field trainings that's right then why type three this alone it's the problem right if it is Norman area then it should be type three four and five right so far so good so either than in this problem in with this topology with this design either I should say type 3 4 5 from backbone into area 30 so then it should be from the EIG IP type 5 a let's say or if this is from the back one type 3 LSA only then probably here area 30 can be what if type 3 less a only top it can be or it can be an SSA but if it is stopped then I cannot have Romeo GRP type 5 LSA I cannot a basic 3 distribution that's why it cannot be stopped can it be an SSA type 3 let's say on the NSS a yes it can be right but in that case e IG I pitch and knots and s type 5 LSA and I should change the NSA type from ERP from type 5 to type 7 LSA so NSSA Arya Arya 30 Candide but and in that case type 7 NSA I should show or area 30 is regular normal area and then in that case type 5 LSA can come from EIGRP but from the backbone it should be i 3 4 4 & 5 similar to area so ABR has to have two interfaces at least one of them should be in area zero I told you before so even if you create the page interface and place that loopy interface in area zero makes the trout Rabia but AVR one of the interface of ABR array aboard router has to be has to be ended RS 0 backbone array okay it's the rule and I am also showing you here a chart again summarizing the which area allows which LSA types you have to know this by heart and as I said in many exam in interviews in real life you have to know this our speed is very common in an enterprise and service provider networks I will show you inside water and data center in this course the used cases of OSPF they are very common OSPF is very common in those networks in those businesses and you need to understand this backbone regular stop totally stop and not months not so study areas very fast again let me summarize in backbone everything is there of course in regular normal area also all LSA types are allowed but there is no topology information exchange okay rich ability information is exchanged through type tree let's say between the backbone and normal regular area and stop area doesn't allow type 4 and 5 external prefixes totally stop in addition to basically in addition to 4 and 5 it doesn't allow type 3 LSA as well but when there is no type treat when there is no inter area LSA inter area rich ability information then how that totally stubby area routers reach entire domain through default route of course and ABR sense the default route ABR sense the default route is typed realization only type 3 LSA in total suitable area is default route okay that's why I am saying default 3 there and not so stubby NSSA area we have 1 2 3 not five okay we don't have also for here but we have type seven so externals Allah says is seen as type seven redistributed Allah says distributed preface information is seen as type seven and of course the AVR an SSA AVR translates type seven to five although I don't have topology for that but there will be habeas in the connecting NSSA area to the backbone area d'etre to Vindoo the translation will win receive the type seven LSA from the NSS area and in order to central backbone area it will do the translation from seven to five so in the backbone area and any other place in the network those prefixes redistribute prefixes will be seen as five type five not seven only seven inside the NSS area [Music]

Channel: Orhan Ergun

Views: 2,417

Rating: 4.8048782 out of 5

Keywords: ospf design, ospf multi area, ospf area design, network design examples, ospf scalability

Id: CLG-mfYEGVI

Length: 19min 39sec (1179 seconds)

Published: Sat Apr 21 2018