 Yeah, it's on you guys hear me good now. All right. There we go. All right. Hey guys. I'm Bob Callaway principal architect at net app Got Sean Cohen here from Red Hat and Marc Sturdevant from HP Today, we're gonna give you an update on what has been going on with the Manila project During the kilo cycle so to give you a quick agenda just to know what to expect For those of you who aren't aware of what Manila is will go through a little history of why did we start it? Who what are the corporate companies that are involved? We'll look at like why would I want shared file systems in a in a cloud a lot of people to kind of Throw their eyebrows at it and kind of say well not really cloud Appropriate we actually think that there's a lot of use cases for Manila So we'll go through those and we'll talk about the new features that were introduced in the keeler release in the context of Those use cases so not only will we tell you why you should care? We'll tell you things that we've been working on to make those use cases a reality Then we'll go through some other features that we've added for and consumers of Manila as well as for the development community for Manila to make Things easier and be consistent with the rest of what open stacks up to we'll talk a little bit more about What we're currently working on in the Liberty release what we're going to be discussing at the design summit here this week And then we'll talk about really briefly. How do we get started with Manila? So a lot of people hear a lot about it But want to know how do I actually get my hands on the code start exercising the API? There's a variety of different approaches, but we'll go through probably the quickest and easiest Way to get going and then finally we'll just give you a reference to some other sessions that are going on here I think we have ten other sessions here at the summit talking about Manila So where do you learn more information? So for those of you who aren't coming here and say oh Manila. There's this cool city in the Philippines Why are we talking about it an open stack? opens, you know Manila very simply for those of you who are familiar with what cinder is for block storage It's a center's a self-service provisioning and management API for dispensing block storage volumes out to consumers or tenants of a cloud Manila is for shared file systems. What cinder is for block storage? So Manila dispenses in a fully self-service open rest API shared file systems out to tenants of a cloud So literally the interaction is a user comes in says I would like a 5 gigabyte NFS share And I would like this particular network range to have access to this share or I would like a 1 terabyte sift share I want to do authentication with an active directory resource And I want to make sure that these tenant networks have access to that share so Manila orchestrates and manages that that level of interaction of Storage artifacts that are provisioned by a you know particular vendors technology Whether that be something from a net app something from red hat HP and a variety of other vendors that have integrated with it But that's that's core kind of its core purpose is to dispense shared file systems One of the things that before I jump on here that I like to note is that you know cinder Is a very complex service as it is even though its goal is to dispense, you know block volumes Manila actually has that networking component, which isn't always present in a block storage environment So you have to think about how do I get shares exported into the tenant specific overlay network So if I go in a neutron and I create a particular network I may want to export a shared file system only to that Neutron network or only to that nova network that was created So there's a little bit more magic that has to happen behind the scenes to make sure that the storage That's providing the shared file system can access that neutron network or that nova network And to make sure that all the plumbing is taken care of behind the scenes to maintain a secure connection For both the clients and the provider of that shared file system It's a little bit about who's been involved in Manila You can see the long list of drivers on the left side of this chart We entered the release with four drivers one from red hat one from net app and one from EMC We've added eight additional drivers and I won't go through them one by one But obviously you can see that there was an explosion in terms of the interest In the companies that wanted to come and make sure that their products were exposed through the middle API We saw the core team grow. I think officially double. Keep me honest here mark But double in size. So we added new representation from Sousa HP Mirantis in that app a large increase in the number of reviewers committers and a pretty substantial line of code Count and the thing that reason I put those stack a little charts up there just to show the diversity This isn't just a single corporate entity that's behind this service like you sometimes see an open stack While net app did start this project back in 2013 We quickly saw a lot of other vendors get interested in this and it truly is a very diverse community of a lot of different companies A lot of different contributors. So I just wanted to call that out here So again like a lot of people still may be scratching their heads and going well, okay I heard you Bob, but why should I care about shared file systems in the context of cloud? You know, we started this project back in 2013 because You know net app heard from our customers that look we want an open standard way of provisioning shared file systems Your vendor specific capability is great but we really need an open standard way to move forward and and to enable really a heterogeneous infrastructure Capability so in 2013 we started this project in mid 2014. We saw Microsoft actually go Public with hey, they wanted to introduce Azure files as kind of a first-class entity within their cloud deployment and Amazon actually followed suit This April in announcing the elastic file services capability on their cloud It's not often that you actually get to out innovate Amazon So I think this is a really cool point that this is the open source community as a whole got in saw a problem Saw a bunch of use cases that they wanted to address and really jumped on this project in earnest And we've seen it really pay off quite a bit So when you think about the kind of the context or the table here that we depict in terms of what are the different storage services that are out There on a cloud Open stack, you know is aiming to fill those gaps just like the major hyperscalers that are out in the market today So in terms of again, so now I've talked about you know, what Manila is why why does Manila exist? Think about it in the context of these use cases and we're going to go through some of these in a greater detail Unfortunately, we've only got 40 minutes with with you guys today So we'll we'll go into a certain subset of those But you know things like big data. I want to provision the underlying HDFS file systems for use for a Hadoop cluster I want to provision an FS share to underpin a mysql an oracle deployment for a database as a service pattern I may want to move existing legacy apps onto open stack to get it to gain the advantages of perhaps a better cost Metric that I get versus a proprietary virtualization system being able to share data across tenants This is something that we added in kilo to make sure that data Actually can be visible and mounted and accessible from different tenants inside of a cloud And making sure that that's fully integrated within the keystone constructs of federated identity that you guys heard about at the keynote this morning Thinking about bringing up shared file systems as a source of data for continuous integration devops types environments to make sure that you've got the Right data at the right time for the workloads that you want to spin up in the cloud And then finally you think about really that that picture of the globe that Jonathan had up on the keynote this morning of all of these different Open-stack instances and thinking about the truly hybrid cloud Having the different endpoints with workloads spun up at the right time at the right place with the right cost and Performance characteristics shared file systems play an important part of making that a reality For customers out there today So all of these things are where Manila can actually be applied and provide real value to open-stack employers So to jump into a couple of these in a little bit of detail One of the things as I travel around the world and talk to NetApp customers is we hear that a lot of customers have started and either You know they've got one guy in the back office That's written a set of pearl scripts that when he gets a ticket open He'll go in and he'll run his little pearl script and that will dispense a sift share a new home directory for a user And there's that one or two guys in the company that know how that works And if they decide to go take another job or they want to switch to a different vendors technology They got to reinvent that entire infrastructure not to mention that they break the consumer interface for that So we see a lot of people very interested in Manila for replacing those homegrown or legacy systems with an open standard API And so this is again when we thought about how could we introduce this capability into the market Being an open ubiquitous API was absolutely a must-have And so Manila really does provide that into the market that same level of self-service Create a share delete a share manage access to a share all of those primitives in a completely vendor agnostic framework That's what Manila is all about Some of the features that we added in order to support kind of replacing these Homegrown systems was to be able to operate in a variety of different network environments So you think about for that those variety of systems provisioning home directories or provisioning NFS shares underneath databases In the kind of existing IT environments of of yesteryear Maybe it was a simple flat network topology or it was a I need to carve it off and put it on a particular v-lan Manila has to be able to interoperate with those environments whether or not Neutron is deployed and whether or not that share should be accessible on a given neutron network Neutron may not even be in the picture in this case if I'm replacing a homegrown system So during the key to release the Manila community added effectively a plug-in model such that Regardless of what your underlying network topology is and what segmentation technology you like to use whether it's a completely flat network Whether you want to aggressively go towards a VX LAN or GRE type environment Or if you're good with just flat out in a regular old VLANs Manila actually can adapt to any of those different network environments and make sure that the shares that it creates and manages are Accessible on the right network at the right time. So this is a really powerful capability that we we added during the kilo release One of the other things we did in terms of addressing that kind of heterogeneous requirement From customers is we added the capability to specify for drivers to specify the presence of resource pools So recognizing that even you know cross vendors all storage isn't created equal and even within single products from from different companies All storage isn't created equal. So adding the pool support in this release allows individual products to expose a consistent definition of what is a resource pool An individual vendors can can add their kind of enhanced capability list to what the scheduler learns about But effectively at the end of the day It's a standard interface for products to come in advertise available capacity as well as additional capabilities and to make that ubiquitous across all products That want to integrate with Manila Another use case that we wanted to address was the movement of existing legacy applications on to open stack Another thing I hear about from customers all the time is I may be you know deployed on a leading Virtualization technology. I won't name names, but you guys can probably guess who I'm talking about You know it works. It's great. A lot of good function, but really really expensive So part of the reason that people look at open stack is yeah, it's self-service. It's great. It's an equivalent for Amazon That's wonderful. Frankly a lot of it's just cost, you know You look at IT budgets aren't growing a huge amount You got to figure out how to do more with less and open stack provides a pretty compelling value proposition To make that a reality But just because I stand up additional infrastructure doesn't mean I have to be able to move the workload onto that infrastructure And so making shared file systems of first-class citizen in open stack really enables Workloads that are built assuming the existence of shared file systems to be able to move over to to an open stack deployment Such that I can get that lower cost all of the benefits out of this new virtualization environment that perhaps an open stack or KVM Might avail for me all of that now becomes addressable for the applications that that I'm running today So I don't have to rewrite those apps to use an object storage API perhaps over in the next five years ten years 20 years I'll rewrite those apps to a swift interface or an s3 interface and maybe I'll change the way that they handle Transactionality and a whole slew of other things There's a lot of apps out there that work today and if it ain't broke don't fix it So I have to figure out how do I move those apps on to the right infrastructure to leverage the economies of scale and the cost benefits that are there So when you think about You know the storage that underpins these legacy apps today and even for new apps You know all storage is not created equal I've said that a couple times now and whether you think about that from for example like a performance point of view a Flash drive is not equivalent to a SATA drive or a SAS drive in terms of the number of IOPS the type Latency profile that I get from that drive the physical hardware is Differentiated and I need to make sure that when consumers use a storage service in a cloud that they're able to access those differentiated tiers of capability So I want to be able to say look for database workload. That's highly latency Sensitive, I want to make sure that you provision that workload on to an all-flash Storage product and for you know home directories that you know, there's got to be a lot of duplication And frankly the capacity planning exercise and those things that can sometimes be a little hairy Maybe I don't want to I'll use technology like deduplication or compression or thin provisioning to mitigate kind of the You know the capital outlay at the outset of that deployment So I want to make sure that I can take advantage of the storage in order to address that particular use case So during the kilo release one of the things that the team worked on was really enriching the concept of a share type So this is analogous to the concept of a volume type That exist in center, but it's a way it's for an administrator to create a Tier of service and to map in individual Capabilities into that tier definition So if I wanted to create a gold silver bronze type hierarchy and have gold be all SSD Silver to be maybe SSD but I'll turn on dedupe and compression and then bronze or best effort is just you know It's going to use SAS drives and be cheap and deep from a provider point of view I want to make sure that we we avail that capability to an operator such that they can slice and dice the storage that Exists to match the workloads So Sean's going to talk a little bit more about how we want to rather than just from a performance perspective We actually want to map that to the application workloads itself. So Sean's All right, so I like I like the segue of comparing to sender volume types And what it allows us to do in the block storage is actually to map the different capabilities We have when we connect any back-end to open stack open stack is all about being a pluggable infrastructure and when we take Any type of back-end that we want to expose and we need to expose it all the way to the tenant flow Right and the way to do it with we've done it in the blocks for example using volume types So share types has actually pretty much an equivalent But it also allows us to bring this richness of heterogeneous as you saw we can actually already today have 12 different back-end supported in Manila allow to create the service catalog to your IT but more important to the Application the workload as you know open stack is the infrastructure layer and we are serving applications And this is a key part to connect between the different workloads. We have and segregate their needs using foushers Service so here we touched upon some examples. So one of the example, which is We see typically in Swift like heart archiving right so I can actually have different class of storage map To my shares archiving now remember your use of as IT users, right? A lot of the day-to-day work you're doing can be is archived because it's documents. It's like raw fuzz You don't want to put it in a SSD's right because that's a different share type So we are able to actually to provide you the service when you need it but also make sure it's being stored on the right back end and Using the Manila API we can do it pretty much transparently now the key years is using Manila opensack is also not just a pluggable infrastructure But he's a scale out Infrastructure so we can actually use it for analytics and you just heard like HDFS has a driver now in keel release I can actually scale out my analytics workload running on Manila using the and we actually have a talk this summit With IBM that actually showcase one of the example. I think is this one Using even containers in the context. So if you need to expose a file system to containers, right? How do you do it? This is where Manila comes into play today if you want to do the same thing with block storage How do I actually take the specific volume from sender and muppet to different? Even containers right? There's no way for me to do it today Manila actually brings it out of the box already So we can actually combine the goods of all worlds. We already have Sahara for example that has Hadoop as a service We have the HDFS driver. We have support so out of the box support for Containers using file systems, and if you bring it all together guess what we can rock And this is exactly what's the number one use case in opensack who wants to raise your hand and say DevOps right so Manila actually mops beautifully into the DevOps model and it's also war for development and See I and testing right so think about it all the way from your build environment And you need to create the version control right you can use clones of Manila in Snap and give it to your developers or testing so it speeds up the life cycle Pretty much heavily and it's already there right. We don't have to do a large investment in the code just to enable this Moving on to another example the shared file system on the map so we heard Bob mentioning the progress we've done in kilo release to To accelerate it but but I think if we look at the what brings us to the table when we talk about the shared file system Right, it's pretty much three things. It's the ability to deliver in speed right faster time to value I can provision in minutes and seconds and and get the data where I need it in my cloud second the flexibility and I have a self Service and if you're going back to the DevOps Use case as just mentioned if I'm a developer. I don't need to even go to it I can easily get my shares instantly right So I have all the flexibility I need to provide to my end users in the cloud and lastly and key important is the elasticity right and one of the things we've done and we have Started to doing killer now. It's going to be a focus in the Liberty release is actually at the ability to expand the share now Bob mentioned earlier Amazon elastic file System that was added the EFS this April and one of the key things of EFS is the elasticity And this is exactly what we're missing right now in vanilla to get to that point to get to parody if you want with Amazon So the ability to actually both grow a share on demand and if you connect that Message to what we talked about earlier with the different back-end support with the different share types Boom if I if my back-end can support like the growth of storage or we're going to deal with fin provision in the next release As well, this is all ties to the third part of being elastic So When we talk about another use case, so pretty much tailored to DevOps use case But the automation integration with an open stator and API this goes back to today's keynote about having one Open API which is standard across projects and I think if we zoom into Manila specifically We actually leveraging and our investment in automation that we've done already using the the rest API But we also be able to leverage the automation tools for finite further integration So if think about any tools you currently have in your environment that helps you manage your cloud infrastructure Connect the two of them together with Manila actually allows you to connect using the open API to do so So we talked about the elasticity increase of flexibility faster value and now let's look at another use case Which is done actually in the keel release is leveraging heat As well, and if you look at the blue box at the top now We have the notion within heat to utilize them in a templates for provisioning All right, so going back to the use cases. We talked about big data We talked about database as a service a bit, but let's talk about it just more So it gives us the ability to actually use file shifts Fashes with file systems, but and Bob actually mentioned all the if you will legacy Application that we're not being able to on board open-stack tail now So I think that's first of all one key use case that Manila allows in the database as a service world, but it also adds the Transparency and convenience of the database management, right? You just map a share That's a this is where it starts and it's where it ends the next level and this is Liberty and beyond And and and mark will go in more more details is integrating with other services such as strove So it's not just about yeah, we can run any database That relies on sure all file system But you always be you be able to tie Manila to this new cloud as a service offering We have an open-stack and then you have Just like I mentioned there with the DevOps example The last one is the hybrid cloud chairs and this is actually have two levels One of them is the external components of share, right? So we going to this morning Keynote about the ability of getting external resources in time etc So I want to actually expand that in a hybrid cloud when you know hybrid cloud is the ability to us on-premise Private and public and one of the things that we need is actually to me great workloads And and how do we do it today in open-stack? Can anyone refer me? How do you do it? All right? So that's actually the future. We and this is we're not inventing and will I mean Microsoft are doing it and that's in Amazon will be able to do the same thing you can actually mount the The the the specific shares from external resources may be from your own premise to my public or my own parents to my private cloud So that's a very key use case that we we're going to see open the doors for open-stack to on board And allow migration in And now with that I'm going to hand it off to mark is going to walk us through Some of the muscles we've hit in the kilo release as well as what's expected in the Liberty or I can yell so One thing I get from this picture is we did a lot a lot went into kilo and Really six month releases that come and go pretty fast. I don't know about you But to me it seems like just yesterday someone brought back a t-shirt from Paris for me Thank you But if you go through here you see we've already talked about about a bunch of these things we talked about share types We talked about public shares going cross-tenant We talked about things like plugging in with neutron and Nova and heat and Sahara and trove I'm going to talk a little bit about plug into horizon and dev stack You see Oslo up there with the underscore le l w li who recognizes what those mean any developers out there? You have minus one right then they pay it forward a couple of times Well, we're doing the things that open-stack projects do we've integrated with everything well We've integrated with a lot and we're using the Open-stack way of doing localization. We're doing open-stack tempest tests. We're fitting in so over six months Manila's gone from being incubated not even part of open-stack not that long ago to being integrated integrated integrated integrated all across the board One of the big things that we did this time for integration is the Manila UI now towards the end of kilo We managed to get this to become an official project. It's part of open-stack. It's actually owned by the horizon team But the core approvers are core from Manila and core from horizon So this is the horizon approved way of making Manila plug into the official horizon UI Prior to this we had a fork. We had a separate UI We didn't really have a good answer for what if you want the latest horizon and the latest Manila But with kilo we have that so if you're using kilo get the Manila UI and That'll put us in a good position going forward to get all the latest features of Manila in the same project as all the latest features of horizon And if you're working on horizon and you make changes Certainly you won't need to manage the plugins if you're working on Manila You might be contributing additional functionality in this project if you're out there a developer your free agent looking for a project to work on Feel free to jump in and work on Manila UI Speaking of developers another thing we did is the dev stack plug-in very convenient thing for developers So instead of copying files around is just a one-liner and Manila plugs into dev stack I don't think I need to talk more about that but you can go to the wiki if you're using dev stack fire that up And don't forget to bring in horizon in the UI while you're there one of the features is more functionality is Manage and unmanage now for that hybrid cloud for those legacy applications You really you're going to have existing shares that you want to bring into Manila So everybody knows manage existing shares with Manila is a huge feature and what we managed to do in kilo is we made Significant progress on this enough so that it got in kilo. So we have the API We have a reference implement implementation and we certainly have a lot of focus on this what we're going to have to do Is for liberties we need to focus on getting the vendors to implement this So with that I'll move on to liberty so plant work for liberty and beyond One of the main focuses for liberty is going to be third-party CI We need to focus on quality and we've seen from other projects cinder neutron that as much Testing as we can do with unit testing as much gating we can do with reference implementations We need to have the vendors run their tests also and it's amazing now I haven't worked on cinder in a while So when I go there now and I see how many tests get kicked off all over the place with every commit It's it's there's a lot of testing going on. So this will be a big focus for for Manila We're not talking about the design sessions. We're just going to do it But it's out there on the mailing list. We'll we'll be discussing the specific milestones for requiring vendors to participate and Make sure someone's assigned to make sure they're making plans and make sure that they're able to start reporting The results of their tests So that's something that will decide as a community what those milestones are but there's a very good proposal out there And we'll be talking about it shortly For other focus areas for liberty and going forward. It seemed very convenient for me to just grab the list for our design session topics those obviously aren't It's not everything that we're working on. We already mentioned extending shares, but These are really the top of mind ones so mount automation. Everyone's wanted that for quite a while So we'll be discussing it again. We want to make sure that Self-service shares once you add access to a guest. You really want the share to be mounted. You want it to be there share migration administrators need to be able to move shares around replication for failover we need to make sure we have the consistent and Flexible way of doing that that all the vendors can work with consistency groups whether it's for replication or for snapshots if you have a group of shares that need Consistency in that in that format. We just need to design that we need to agree on that and do it cross vendor Then provisioning is one of those things again We need a consistent way of doing it and a flexible way so vendors can do it the way they need to do that So we'll have a talk and get that design nailed down so that we can implement that in liberty Version to objects and API versioning that's about smooth upgrades So if you're developing interested in versioning version objects API versionings, I hope you can contribute come to our design Sessions or at least follow our etherpad if you're not a developer. You're probably glad to hear we're working on smooth upgrades Fault feedback is something that we're all interested in Relationships between shares and snapshots is another one of those things we need consistency across vendors or at least agreement in how to expose those API's and service images used by drivers that use You share servers. We're doing some work to design to share that to make sure that's usable leverageable So quickly on mount automation This is something This is something that everybody wants and it's been talked about in previous summits The concern here is that there are there are security reasons why we can't just do it and there's also resilience And what do you do with reboots and things like that? So we've talked about a variety of ways of doing it in the past and probably didn't get enough attention during kilo At least that I know of but now in the design session, we're going to have some proposals. We're going to bring it up I think we have some very good possibilities for how we're going to tackle this and I expect we'll make significant progress on this in liberty Replication replication is obviously a difficult one to challenge across vendors So again, we'll have to get together and we'll have to decide now We're not talking about replication at the block level. We're talking about replicating a share or a group of shares But really under the covers vendors might be doing replication at the share level They might be doing replication of volumes So we have to figure out how do we expose that so that vendors can implement it? How do we make it so that? Users can get some have a reasonable understanding of what they're going to get when they ask for a share and they ask for a Replicated share and then of course when you do fail over you want a consistent result. So that's where consistency groups I think we have probably a better handle on that again. We'll find out later this week when we're designing all of this But we we need consistency groups for fail over It and for snapshots if you're writing to a group of shares and you have to fail over or you're taking a snapshot You want to make sure that right order is consistent. You want to make sure that you've got a consistent Look if you're moving that from that application from you know one copy to another So again, we'll design that we'll talk about that Wednesday or Thursday The last one I wanted to bring up is share migration That's my Canada tie-in much like the Canadian geese For some reason have to migrate to California golf courses Manila shares do something very similar, but it's administrators Sometimes you have to move a share from one pool to another from one back end to another maybe even across back ends In this case, we've got some work already in progress We've made good progress on it But we do need to get together the design summit is perfect time to do that to make sure we have some agreement on What vendors can do what how are we going to do this and in this case? It needs to lead probably to if the vendor can't do it like migrating from one vendor back end to another vendor back end We need general solutions that can do that Migration is a great area because it can lead to other things like data copying for other purposes it'll probably lead us to other features in Liberty or beyond like retype and Again, it's something people expect. I think we'll make progress on that in the very near future With that I'll hand it back to Sean. Thank you mark So I don't have to tell you about the manila momentum because you already saw that right? I mean the number of features that we are able to add in the last cycle the number of drivers we had In the last cycle it's tremendous and it almost say one thing the need is there and now open stack aligning right? So all the work beautiful work we're doing and you're seeing here It's actually the fruits of the work that align to the need and you saw that earlier as well from from the marketplace as well with Commercial offering so how do I get started with manila? That's my next thing So I'm glad to say that The manila is already available in the rdo distribution, which is the community Version already from the general issue you can go ahead and deploy it using pack stack. It's pretty easy We're also going to include manila as a tech preview in the upcoming Rolo SP Distribution this summer so it will be already there so customers will able to start test it and work with it and Open bugs and help us get the mature point that manila wants to get and one of the things I Wanted to show you Speaking about the manila momentum is how to get involved as well. So you heard mark There's tons of areas that we need working hands To make it to make it happen that that's the great way to get you involved. So please go on Both to get up and a wiki. That's a great starting place. We have Open-seconded on RC weekly meetings Feel free to join us and of course we have the Mailing list as well, which is key right if you want to impact if on it even if you user the features that you need Please let us know right the operation side of open sec is very important on with the development He has to go hand-in-hand together and Going back to some of them Sessions so mark mentioned all the different Designs talks we have around manila, but I just want to copy show you how many talks We have which is general talks about manila and some of them are really cool, right? Take big data analytics and dockers the frill and manila. This is an IBM session How you can actually scale out? Hadoop with manila and containers. We also have Deutsche telecom Taking open-structure file storage to the telco club. So that's a different use case that we haven't talked about yet The telco needs rates not just enterprise application Telco also has a big interest in manila So I think today we're able to taste some of the key use cases of manila But this is only the start and as you can see from the numbers of design sessions and talks we have on manila It's no doubt manila time is now and with that Let's open the mics for some q&a. We still have I believe five minutes. All right. Yep And we've we've got a couple folks from the manila core team mark Jing Clinton probably a couple more than I don't see yet, but any questions you guys have? Do you have a microphone? Yeah, one there So just for the folks on the the video or dick didn't hear it the question was There doesn't seem to be an API today to roll back a share from a given snapshot So basically to restore a snapshot into an existing share. So yeah for that. I don't know you want to Yeah, you're right. It's not there So you can create a share from that snapshot. So you you have one, but it's a new one and And what you described is exactly what people would want to do So we should we should add that we should put in a blueprint and and see if we can make that part of the core API I don't see why we wouldn't want to It can be vendor specific like an extension. Yeah I mean it makes sense to for that to be a core API I mean again, yeah, look at the parallel with cinder cinder has a snapshot API and it also has a backup and restore API Does it make sense to have an explicit backup and restore API for shares as a whole? I think you could make a pretty strong argument that that does make sense And then it becomes a question of do I restore from a backup or do I restore from a snapshot? It kind of comes into the semantics of what's a backup versus what's a snapshot and won't go into that But that that type of a operation I think makes perfect sense. So yeah, so definitely Jump into the design sessions this week jump a throw a mail onto the mailing list jump on the IRC meetings That's the right way to engage and get us to start looking at that. All right. We have another question Yes, I wanted to know whether you have any kind of SDN integration in the works and if so with which as the SDN So the question is do we integrate with any SDNs? So we integrate explicitly with the neutron API so we don't have a specific hook to any other vendor specific technology So whether you're using Novenette or you're using neutron if you have a particular ml2 plug-in that leverages Open daylight or you've got an ml2 provider from from Cisco or any Juniper or Whatever that provider may be that's all abstracted by neutron at the end of the day So in some sense we support all of those a SDN providers that supports neutron It's a bit of a pass the answer to you But but but effectively we think that's the power of integrating in with the other open standards Is that that community will innovate at their own pace and when things come along? We'll just figure out how to take advantage of that Thank you. Sure The next question Hi there you guys talked about Sahara and HDFS fair bit does Manila support With provision for data locality ensuring that your Nova compute nodes and say a Hadoop map reduce are going to be Physically co-located with nodes of your shared file system So Manila today supports the concept of an availability zone much like You know sender and the other services do in terms of ensuring valid export locations from multiple availability zones and managing that Access explicitly in that context is something that we're we're figuring out in the context of the discussions around replication Because again you want to make make sure that the share is accessible in the particular cell or region or availability zone of the Cloud that I have my workload deployed in and I want to have to cross a boundary to get access to a share So adding in the the additional management APIs and objects in the model to be able to support that is something That we're actively designing in the context of replication So it's coming I guess is the short answer to that and during kilo one thing that was added was Allowing a share to have multiple export locations So in the past there was one export location for a share, but now at least having the possibility to have more than one You can have some that locality. It's just we don't have a good way of communicating it So that's not made general, but at least the different mount points are visible, right, right? All right, I assume we need to take the last question because we need to respect your So if I understood your question correctly neutron perhaps is complicated some folks don't want to go down that road quite yet Do we have the ability to just access a share over any layer three routable network? Is that a fair characterization of it? So so like a vert FS driver Effectively yeah, that's something that we've talked about in the community and it goes back to that mount automation theme of how do we provide access? You know certain vendors may not be able to interoperate it and put their full support statement from a client point of view behind a vert FS driver But there certainly are advantages for an individual You know Guest VM on a hypervisor node to have that access intermediated in that way So short answer is there's some trade-offs there We haven't seen anyone in the community really come forward with that But it's something that we have talked about in the past in general our network model allows for you to create a share If it's on a IP address that anybody can access anybody can access it if you want to filter that down to You know specific subnets or specific IPs and security groups you can do all that as well But you know like a vert FS style driver is something we've talked about but haven't implemented yet So I think we're out of time. Thanks guys for coming if you've got additional questions Like I said, we'll be up here at the end. Appreciate it. Take care. Thank you