Rubrik CloudOn Instantiation Demo with Andrew Miller

so we're now into the second part which is cloud on or instantiation because cloud on is our term that we love and instantiation might have a little bit more meaning and that is because I'm sure this morning that you woke up and thought I love archives people out there who love backups out there who love archives sorry that you insulted you and you raised your hand already but you know oh well no one ever I love data no my name is Andrew Miller happy to work with Rebecca and Chris I work at rubric you can find me on twitter i blog infrequently it's been a lot of blogging on rubric not as much my own blog that's hard and if you've seen Chris's intro slide where he has his spirit animal as a llama in Amsterdam and Rebecca gets a koala than I can jolly will have a kangaroo so we looked at first cloud cluster basically oh this is the foundation we can run in the cloud and then we're going to move on to cloud out which is a foundational part to get to cloud on so there's no there's no slide here for cloud out but this is because it's a foundational capabilities we've had for a long time in short you can take your backups that you've had on premises on premise on premises and some I'm going to get it wrong eventually on your hand I'm going to get a stink eye Prime but if you know what I do and and where everybody's everyone's safe like so that you can take your backups on primp and then instead of keeping them forever on rubric as Chris alluded to which is relatively expensive you don't want to do that we did actually add the ability to archive out the tape if you must but the way that we prefer for you to do it for years you can send that up to relatively inexpensive blob storage s3 Azure blob etc right that kind of thing however what you get out of that is not necessarily a lot right I'm not downplaying this because there's more recoverability there than you had out of tape at the same time also you can do single file restores out of the cloud not pull a whole 100 gig VMDK back but at core it's archived so now as we go into cloud on we're going to build on top of that and if you missed it this is in some of our webinars we called 3.2 hour more cloud release because of cloud cluster and for ATO we call it the even more cloud release and of course that's a pretty big build-up for cloud on so there's got to be something here other than more although if you would prefer to listen if that actually puts you in a happy place because whenever I think of that it's like I just makes you so happy so cloud on what is it is the ability to take power Don I take an archive snapshot of a VM that is living in the cloud and power that on that's simple but yeah there's got to be a lot of stuff going on it under the covers to make that happen so in when we're doing this we actually go and kind of troll through the dot vmx file because starting out today this is a VMware VMware - AWS specific feature there will be more in the future right the foundational concept is the same so give you a recommendation on the config file to click deployment with orchestration so that even goes toward that in the UI it's relatively simple and the orchestration is the aspect that if you're thinking about you're uploading all of the incremental VMDK backups up to s3 that's not necessarily consumable by anything on the Amazon side so we do stuff to take those merge them all together turn it into a flat monolithic the stream optimized pull it into the VM import service from Amazon which notably doesn't cost you money when you call that service by the way is that is that leveraging the relationship between VMware and AWS specifically not yet okay so it's not using that piece of it no not at all I was doing a conversion to an ami does that then take time or you it does take you bits underneath ok does take time that's that's a copy migration to get so that's even why over the last bullet point cuz that's a perfect segue on tobacco is that we can do this on demand when you do it on demand we'll see in the demo you've got to wait for however long that conversion time is we can't help the conversion time but what we can do is we can give you a constant conversion option that every time that a new backup archive is uploaded we trigger that off so that it's ready for you right you go use that for dev kit for example all for a dr or something ready to go just comparator you might be stealing my use case fly all right it's it's here to go looks live but yes you're exactly right exactly right UI integration launch power on it's in the UI and this is where this is this is now a little bit of marketing but when we talk with customers sometimes you see differences between sectors health tarah finance etc this was something that every single customer do we talked to wanted like this is something I want and as well because we talked about you know we started with backup in recovery as a use case but we consider that to be one use case this really aligns with the whole word cloud data management company we're looking at hybrid cloud enterprises right I mean this is not a traditional backup and recovery feature was that we are into demo time Andrew and with what's the you may get this into the roadmap later at what's the plans I'm moving back if you need right now nothing public how we're happy happy to go further into that in an offline session as we finish up today that's a very good question though and now who was I talking with Justin last night anytime you say very good question it's a tell that you're thinking about the answer so but I said it the after I gave the answer I didn't say at the beginning so I wasn't like stopping to think stopping to a delay for time okay so we have our rubric cluster here which due to the magic of timing has timed out on us and we can see that there so I will take this full screen so thanks to Michael fall who'll be coming up a little bit later part of the sequel live mounts section we do actually have a VM up here that we are using in various ways I just want to go there we go okay so there's there's multiple ways that we could go there for instance we've got this virtual machine it's called se dash M fall - Lin we have it protected with the rubric policy write a brief pause here the policy is our core construct that applies to all different data types whether it's Windows Linux sequel VMware and you set the core things you care about how often you backup how long you keep your backups for archive and retention right brief slight that's the core here if you're not comfortable that are familiar look back at previous tech field days but as you can see in here we've got one Windows host protected in this case just to kind of overload the demo I'm going to go to it via going to go to the via search pardon me one as I have most of the demo memorized but I also was good and had notes to make sure I hit on all the key points so as we go into this we look at this we have the regular console and let me make it a little bigger that work for everybody so here we can go and because we did just stand this up recently I actually actually frankly for cloud field day - you know an evergreen dot is when we have a successful backup and you can see here we've been doing a good bit of backup it's not not a lot in here just see : users but normally when you come in here actually I think I got the Windows instance so let me make sure to get the VM here we go here's the VM because of the things we wanted to demo and integrate we actually are backing this up at a virtual machine level windows level a sequel level so lots of different ways you can slice it so normally you would have a point in time you can see in this case we're taking backups about every four hours the traditional options around live mount and instant recover we've done for awhile single file export but you also have this neat launch on cloud option so this is keying off of the existing cloud archive locations that are set up I'm not going to go to setup screen but it's basically your AWS s3 setup we've got to choose your bucket that it's going to go to and your keys so it's encrypted you can choose where it wants to where you want to put it it actually goes and checks to find the security group so this is a live view of the security groups out there of course you can it currently this does put it in the default V PC and subnet you can change that afterwards and in a next dot release excellor will be the option to choose that in this screen if you want to then of course it's actually going and looking the VM and saying hey I think this is an m4 x-large based on what's in the vmx file mapping it out because I don't like keep m4 dot whatever and t2 dot I mean there's a lot of options there right kind of thing so we're trying to simplify that for you especially on on-demand scenario if you really want to you can go and do a custom instance type that case you type it in manually now for the sake of this I will go ahead and launch this but to the question earlier this will take a little while because even though it's been archived up it's actually going in calling a conversion process but I want to keep it focused for right now we'll go kind of under the hood back on the slides a little bit but I want to think about okay so what's the customer experience like I'm doing this as an end user as a customer whoever so it's either I go in here I see the one screen and I I wait period of time or let's say that you know I really don't want to do that so I've got this option over here for configure cloud conversion that is literally what it says can even make that a little bigger convert the latest snapshots into an ami and either delete the older ones by default or keep them so this is literally just a slider and turn it on as you want to keep the older ami so be aware this takes storage this has a storage cost because these are EBS backed a mis but there's no compute costs here when you're doing this constant conversion option that even goes right toward the dr use case you're thinking about it i'll steal from three slides from now about thinking about a computer CR option that's a little bit different in the spectrum of dr things so we have like priority tears in that connecting the basically from a storage cost perspective you'd be keeping two copies of the vm one would be the one ready to go as a new one comes in to do the conversion and delete the old one unless you choose that say you know i want to keep all of these because that's that's phenomenal a dev test scenario or maybe even in a scenario of dr where you know that maybe you want to keep it back further and then maybe you're doing some manual cleanup kind of thing because you don't want to just have the most recent copy for that so i can kind of play around with multiple ways I'd go at it when you come over here in the interface we have a new cloud mount section under live mounts if you've seen in some previous versions of rubric it was just live mounts now there's for hyper-v and sequel's is that in a minute and cloud mounts so this is actually one that we fired up the other day and you can see here I'll call out to the mm fall VM one all this make that a little bigger to kind of show that this isn't smoke and mirrors I mean we created this for the for the demo here today but if you look at this ami instance where it says you know 3b9 one we go under the a.m. is we can see all the a.m. eyes that are out there that have been constantly converted kind of thing so you know you can actually see that down here we've got that that's the one out there and there's actually some other ones out there because we have this continuously converting as well as their the other VMs that were in this policy some sent to SVM's one that had puppet all relatively small things just because it takes time to replicate up and down kind of thing but legitimate for this purpose where I want to go to finish this is that when you look over in Amazon and again just kind of show there's not smoke and mirrors here so we go here they real so if you remember that 3b and one we come over here into am is I might have kind of pre-populated 3bn one you can see there's an ami there that's available and then even more when you come over into instances because it's not a VM and Amazon it's an instance right because you're not supposed to keep it forever unless you do that kind of thing and actually we've got this one that's called the M fall on demand just scroll down and look at this you can actually see the ami that it's based off of as well as the public IP number so kind of to finish this out from a demo standpoint we have the RTP session up because we might have pre staged this a little bit just for the sake of of going quickly so if you look at this virtual machine actually this is actually literally the jump box that we're using for some of the administration so this was actually a cloud on instantiated used to be a virtual Windows machine on Prem on Prem and it went we took it up we converted it on-demand it took a little while right kind of thing cuz there's not from an Amazon perspective it's actually up here and running and now this is a little bit like if you've seen old disaster recovery or Site Recovery Manager demos where it's like wow I had all this stuff and I have Windows running out there that's really cool so to make it make it a little bit more applicable keep it from a user standpoint so this actually had sequel running on it at the time where we backed it up and we actually had several databases out there such as the classic adventureworks 2014 we've actually got a table out there at center another thing so this is a it was running sequel backed up archived over converted over it's running up there there are these backups happening to like a different AWS region like I'm thinking the VR scenario and this really to me the only likes the only time it's relevant right if there's a problem with your region you may want to bring your servers online different one like how are you protecting that data so in in this case the one of the most common ones would be that you are on premises and you want to have a cloud er option there's a multi-cloud aspect which i think is what you're thinking about is say inside amazon but this could actually be an option where your on-premises and looking it up and having it run up in the cloud or you could go seems like we're not hybrid cloud but multi-cloud more as much yeah I was thinking about the scenario maybe where you're just in the cloud and you're backing up to very very if you're already there the crossing region right either cross region or even cross cloud if you wanted to actually replicate using cloud cluster or you could have a cloud cluster protecting and then actually sending its replication because it is the same code base so a cloud cluster running an AWS could actually be archiving over to Azure not that that would make either Microsoft where Amazon terribly happy but your work out cluster in Northern Virginia and the cloud cluster California replicating and paint it up in the opposite region when it goes down it sounds expensive yeah it's clouding the cloud you know going in you're going out of one and into another the main over there is that we are working on and this is now gets into kind of sales and pre-sales type engagement around cost calculators because you're right if you if you approach it the wrong way it becomes prohibitively expensive it kind of falls down but you have a technical standpoint this is working we actually of cutters are doing sources region the region right within the same provider that would be what regional region costs what love our leads are you oh yeah an AWS Union but at least if it's done through rubric its be duped otherwise it's stored like in you know s3 and then s3 copied bring one to the other but it's not it wouldn't everything I'm going to change that would be it actually BD Duke to if you went like if you're going from AWS to Azure units D duped in AWS before it gets sent to as right right when we replicate you create the items yeah the replicated traffic if it's clustered a cluster that's always deduplicated it compressed if we're doing a cluster that's archiving just to a bucket basically or a blob we're only just sending the different the incrementals and that is an ax rubric format that's already been deduplicated and compressed as well so major impact there from a cost perspective I'm obviously dancing around quantifying that with hard numbers but this is why we why we work from a kind of a calculator standpoint with customers very requirements to deploy the cluster within the same AZ or can we span these for protection that way a single single cluster similar to similar to on-premises on-prem actually needs to be be be locally linked so on Prem it would be you know local 10 gig links so a single cluster would always probably in the same region but you could have different clusters and Woking retain a disease in their region as long as the links are fast enough but that's really the core technical answer I'm just saying from a protection standpoint by four nodes 1az goes down I saw up three nodes and it's surviving a Z so then in terms of the D our scenario where you said on from to cloud don't I have to have rubric on run yes not scenario yes Justin you could also have edge running locally on pram right and backup to cloud cluster as well if you don't have a physical appliance you have their edge virtual appliance doing that as well an edge is the single node remote office branch office that we provide specifically for that kind of scenario and and different performance characteristics just because frankly it's running not the code but it's running on top of whatever virtual infrastructure you happen to have out there versus running our own boxes with that are dedicated okay so then can you clarify the question he asked earlier where you said there's no fail back so we want this for dr i'm gonna want to feel bad so right now the fail back would be a manual process of ending frame so it's a one-way this is where i'll i'll freely say in 4.0 we have introduced cloud on the same comment of happy to go further in an off-camera session but today that is what it does a little bit under the hood I almost wanted to just jump to use cases and then come back what I think I'll I'll finish out the under the hood if I can if I can beg your indulgence hopefully make sense so what we're doing here under the coverage this thing from a user customer perspective very simple operations UI pretty transparent we're sending the archives up to s3 or AWS we're the point where you request cloud on or it's a go and making a recommendation and then what we're doing under the covers this is where we're actually doing a transient ephemeral dissolvable take your pick rubric node that we actually spin up in this case in AWS and that's actually handling the conversion process because it's having to assemble all those incremental incremental ii uploaded vm decays which is a serious thing because you've got to want to have your cake and eat it too of i want to be able to upload incrementally but the import service doesn't want it in incremental it wants it and stream optimized so you've got to go from incremental to flat monolithic assembled together then the stream optimized then you can pull it through the vm import service alright the conversion just always happens after you request a cloud on if you do the on demand option yes if you have the continuous option checked it's actually doing that every single time there's a new incremental push stop I missed that part thanks cool and finally uh actually almost there's but then either you're launching it on the cloud or the ami is sitting there ready for you so at that point there's no no cost except from a storage standpoint because it's an EB aspect ami and then we power down and we purge that temporary Ruppert node unless you have chosen enough options where there's constant you know cloud on conversions happening so it stays around so that we would know it is only temporary you're not you're not going to miss it where you not going to use your existing say for class AWS Rubik no this is just for that conversion correct correct very intentional design decision to enable this to work whether you're doing cloud cluster or not we'd love for you to do cloud cluster but you don't need to and this is actually something this just managed by whatever brick is calling that process and that it's disposed of we're not needed what what incentives exist of me to stop developers just leaving these things running forever because I need it which happens the way we would need to do that like a customer would need to have some sort of financial control it's what you need to do that man wearing a lot of money to no reason but one other piece of this has been on demand and I can even go I won't jump back into the demo I can pull it up later but there's a screen where you can see all the cloud mounts that have been in requested by rubric and so you can at least track and monitor all those and then delete those I'm going to go a little bit to API and other items Michael please added to that so rubrics API driven nice to talk about leaving stuff out and saying hey I want to clean this up I mean we have internally in our own labs and we can certainly help customers with this but we can develop scripts using our REST API to go through and do the cleanup for you so that it while it's not baked into the product we have all the tools there to enable it and make it happen I'm envisioning something like ServiceNow that we spoke with earlier today where it's like you you order it up and it comes up and then it just keeps billing that that you know for the cost code and so if you want it on that's fine but you've got to pay for it also at least from a financial incentive standpoint we don't have any incentives to make you keep earning that stuff out there so we're happy to help you clean it up from an API perspective or otherwise or even future product capabilities yeah but it's not like we are baked into trying to increase your cloud bills because we get a cut of that but it's also about help one of the things would be if it takes a long time to spin one up I mean that's one of the issues with why I would say I will over provision stuff because I don't know I may not be able to get it later or it takes ages and I need it so I want to destroy it which means I'll incur costs that I don't necessarily have to as long as it's very quick and easy to do and that's the exact reason behind that continuous option and you keep one ami or just keep all of them in okay it's actually almost the minute we have mitigate that kind of question yeah use cases are always hard sometimes I want to put them in the beginning because it's kind of why you care but if you don't but people may be getting too soon it's like well this doesn't make any sense think of the architecture so in this case use cases the end of all has to be done we talked a good bit about cloud sandbox for test dev I don't think I need to beat that to death anymore except that from a future standpoint realize I took Brian's question and put it in a futures category I will actually talk roadmap here for one minute we have a firm commitment from product management that in the year 2030 not only will have a cloud sandbox will we have a inn-yard sandbox option as well so this is for the main benefit here is that for all of us you can't explain to our kids what we do or our nieces that our nephews if he Rebecca expect this to spur IT adoption that didn't go over quite as well as I thought it was funnier woman my shadow cat let's go ahead but I mean that's just like happy memories right I mean for everybody to come back that's recovery so talking about this is a uniquely compute 'less option there's far more that we're going to keep going but we see this is a foundational capability because we are handling some stuff where if you look at what's happening under the covers there's a good bit of moving parts just to make this happen lift and the shift might be a dirty word but you know that this could be facilitate one-way conversion options this is where you're probably treating whatever cloud option is a little bit more of an eye as and not going into the full refactoring that kind of thing but can be used for that because we are helping with either on-demand or pre stage that even goes from a lift and shift standpoint to thinking about the windows of cutover windows and downtime and I'm doing certain things you have customers that are actually asking you for that we have customers that are very interested in reducing data center spend and thinking about how they can take resources and moves into the cloud specific were they going to use it this way I'm honestly not sure but I know that anything they can do to move things that direction they're good after them but I'm trying I guess what I'm getting at is kind of a more of a reality as opposed to a potential do you have customers that are actually doing that today cuz I have if the answer is yes then I have a follow-on question to it if the answer is no that we can move on I know so I know the answer is yes from a talking with our product management team I'm not going to pretend that I know some of the details about the customer scenarios but I'm having to take a shot at the follow-up I could call it no it doesn't out of it but so what are they experiencing after they make that move as an after they're actually migrated into the cloth yeah so in that model in this model it's they're usually picking up specific applications and this may even be a little bit more of an azure play than an AWS plan some ways they're trying to take certain application sets pull them over and not have to run them on premises sometimes triggered by the traditional I've run out of data center space I have to go to entirely new rack I've got it I've got an incremental cost that's triggering a big physical data center expansion absolutely but you are seeing this and I'm sorry I did want to make sure that we're real quick crystal-clear on this question but for lift and shift no refactoring they're moving an application customer you have customers that are moving an application to public cloud and they're happy with it and they they're not there's no further or the divet II happen systems and other purposes for the time being I don't think there's anybody who necessarily loves lift and shift is an option but sometimes it's better than the alternatives from a I've got to add this much more from the storage array standpoint or other on-premises type options full people you've got lots of small data centers which cost a lot of money and they can shut down half of their data centers by just lifting and shifting into her eyes I you know that again I think there's not a great not a potential but when I run the numbers that's doing that note is expensive yeah that's very expensive that's like this takes me back to my days when I used to do actually combinations of V plus C plus C stuff and we'd cost it out compared to AWS and say like you know these are these are the numbers so and I'm not saying it's like 20 40 percent more expensive I'm talking about multiples the numbers can out to be that is a little bit Mark's the usual thing is people have no idea what the on-prem costs often Wow because that - you know but I know I know people who are a very distributed organization who is running you know a single rack in I think it's sort of 7080 colors around the world that costs a whole bunch of money to maintain that - that a use case but yeah I would agree that it you listen shifters is not an ideal but people have strategic reasons for maybe doing it but I mean I don't know whether you necessarily use rubric for that because they're free tools out there it's a one time conversion I don't think this is your perfect or it's not fair understand yeah is there there may be a reason why I was number three on the slide yeah finally and this is a little bit sore Rebecca what we're talking about of customers and api's we have I mean Chris referred to customers you get excited about so what we're doing which is really cool and that anything you do here you can call from an API standpoint so we are frankly pretty excited to see what else people come up with we're not thinking of you may have some wheels turning in your head already if like they should have had some other things on the you skate slide or because all of this can be called for as a restful api this can be a foundational capability that may underlie some other orchestration automation products what I want to go back to these are second limit after these slides so dr if you have to do that conversion process yep either continuously or on-demand that's going to have an impact on the RT on the RP all right yes and you talked a little bit about what impact that's going to have sure so this is definitely within this spectrum of D are options which I'm not going to always the premise like it's synchronous or asynchronous or a base replication okay so the factors T that play and from and I'll start with RPO perspective because our Pilon aren't you so RPO perspective would be the frequency of how often we're backing up at a rubric level how long that's taking to go up to the cloud and then that continuous conversion process so generally when we look at the vm import service from amazon that the timing can vary but you can often get our PR POS that are converted to an ami like that final what I can use that can be in the 12 to 24 hour range depending on the size of the VM your mileage may vary default supply all right for an RTO standpoint actually pretty fast because I actually thought about going through and doing a manual launch of a instance from an ami I mean you can click through that in the GUI within 60 seconds if you know the screens look like that kind of thing so the art to you actually ends up being really fast from a OS running standpoint and then of course we get into what the applications are they need to run to the business level I'm on liner yeah especially if it's multi tier so does that answer the question yeah yeah it just wanted to get a feel for you know what what levels are you aiming for so there's going to be a fifteen minute version it's going to be 24 hour if we're talking kind of classic tier one two three four we are still operating in some of the lower tiers as far as being real about the capabilities just because I mean you think about moving that data up there but if this is built into your current strategy and it's cost-effective for some things and more reliable than other options in that out there and moving the dial further on that over time but I think I don't want to we are good on time but I want to keep going to the third section there any other questions before we keep going real quick do you guys this natively tag the instances when you're spinning muffin ect not yet okay is that gonna be an option to add tags because I'm kind of going back to thinking about Jeff's question like have you Eddie managed ongoing costs if you can tag that and run you know some lambda function to report on that then you can you can already track it by virtue of either looking at the rubric console you know cloud mounts screen an either frankly calling that and then making the correlation out of the scrip because you can call everything via an API there will be more options as far as tagging in the near future the other question I had is um using cloud cluster maybe going back to rebecca's it's it's still a client on the ec2 instance right or is that changing so Cloud Club measured and agent in the hits I meant yeah we supposed to put an agent on the on the instance to back it up with cos cluster right there future plans to change it and and to be able to select ec2 instances from the console yeah I'm a put that into the bottle up session Ashford's but yes okay

Channel: Tech Field Day

Views: 5,463

Rating: 4.7777777 out of 5

Keywords: Tech Field Day, Cloud Field Day, Cloud Field Day 2, CFD2, Rubrik, Andrew Miller

Id: Z_6whNYHjLU

Length: 31min 1sec (1861 seconds)

Published: Sun Jul 30 2017