01 - High Availability Architecture

okay Anthony so what is the topic of the day today well let's go through our high availability architecture okay sounds good we start out in this environment we'll have two boxes it's our traditional cluster so we have heartbeats in between we still need to connect our storage to the boxes along with storage we have our network so now that we have our environment we're protecting an application application fails from one node to the other all with shared storage this is on one site site one so this is woke away CheY a local high availability right correct local high availability moving from their customers expanding they want to get a little bit further out so another site another node so we're working with campus cluster okay so what's the requirements here campus cluster can work off of the same heartbeat infrastructure so you still have to be able to connect to them you need that same level of network so you're connecting in as well and now you also could have local disks to that site the disks will be mirrored so you have dual access making sure everyone can see all the disks the latency has the requirement said this isn't going to be an issue but you're in an extended distance while still maintaining that same local cluster okay when you say in an extended distance about how far it needs to support with latency the ability to do mirroring without impacting the application okay so relatively short distance correct okay if customers don't have that relatively short distance but still would like this clustering removing the the connections to both sets of disks you still need those disks but now we're going to talk about a replicated data cluster so rather than the Mearing where everyone's connected to all the storage now you still have that same connection just locally you can either do that at the array level or at the host level this is the replication you're talking about correct okay the replication replicated data cluster this replication requires that Yuri have synchronous replication it has to be close enough that you can write from the application directly through the array to the other array come back and respond or through the hosts okay but it is still a local cluster all failover goes from node to node to node our final architecture is dr so disaster recovery in this instance we no longer have one cluster we have two separate clusters per site one cluster locally here one cluster locally here you still have replication either at the host or the array site but it's no longer synchronous because it's not close enough now you're doing a synchronous it's different in that we're going through and writing down from the application to the storage array and the storage array will replicate out or through the host to do that same method it writes out as I can write so it does the right and in asynchronous mode it does it when it can do it it doesn't have to do it immediately once it's written down to the primary disk absolutely and the difference here is that with each cluster being separate they both understand the application but only one can be on line at a time so if site one goes down situ comes online side two goes down site one comes online when you switch over you're also changing the replication okay so basically what I'm hearing from you here is our VCS product can handle all four of these different types of architectures with a little bit of different configuration right absolutely customers can move from a local AJ configuration to a campus cluster or replicate a data cluster as well as extending it out to dr fantastic well thanks for explaining the different h.a architectures that we have that our customer is going to be able to see so thanks a lot thank you

Channel: Veritas Technologies LLC

Views: 50,084

Rating: 4.6014233 out of 5

Keywords: 4609117465001, infobits-cs, edu-customer, youtube, edu-infobits

Id: lJLk3yPRV1w

Length: 4min 38sec (278 seconds)

Published: Thu Mar 10 2016