Videos

Hello. My name is Matthew Monday, and I’m an engineer here at Nfina Technologies. Today I’m going to give you a brief introduction to our ZFS based storage solution known as Edgestore. The solution we can run it either hyper-converged under VMware, Windows or on a traditional converged architecture on bare metal hardware. I’m going to go over some of the features here of the actual storage appliance itself.  

I stated before, this is going to be a ZFS based solution. So with that, we’re going to get data and metadata and metadata check summing which means that every single block on the array is going to have a checksum that silent data errors, bit rot, things of that nature can be detected via the scrub utility, which will actually go through and read every block and compare the block’s checksum against the checksum on disk. It’s also going to include atomic transaction writes, which means that your writes are either going to occur or they’re not going to occur. There’s never going to be a point in time where you actually have to run something like check disk or FSCK to repair the file system because a write was acknowledged before it was actually committed to disk.  

As far as protecting your data, the array is going to be able to do two-way mirroring, four-way mirroring, single, dual and triple parity RAID which is going to be known as Z1, Z2 and Z3. Those are going to be the equivalent of RAID five, RAID six, and the nonexistent RAID seven in the hardware RAID world. It’s also going to come with native backup and DR built in. So, we’re able to actually replicate the data from your iSCSI or your fibre channel LUNs or your NFS or SMB shares offsite to a secondary location or onsite to a backup box that you have in the same location.  

It’s going to be a highly available solution. So, it’s going to be a dual controller, which means that one of the storage controllers can actually be powered off, shut down or completely fail, and the other controller will be able to take over the resources and continue to serve IOs to the client systems. So the system can be configured active- active or active-passive, depending on the set up, there’s different architectures where you can actually have a shared storage architecture where all the disks in the system are SAS and the two storage controllers are able to access all disks in the system via dual ported SAS drives. Or there’s a metro cluster option where the data and the two storage controllers are actually kept in sync via a Ethernet link that copies IO in between the storage controllers. This link is either 10 gig, 25 gig, hundred gig, depending on the actual setup. We can deploy this as a virtual storage appliance under VMware, Hyper-V, or we can deploy it directly on bare metal hardware.  

As far as accelerating the storage performance in the system, we have a couple of mechanisms built in. So, the first is going to be what’s known as the write log or the ZIL. This is going to accelerate random writes through the system. So essentially any random writes are going to go to the ZIL before they’re actually flushed out to the pool as a sequential transaction. The ZIL is typically going to be a high speed SSD or NVMe device. Read caching is going to be twofold, so you’re going to have first level and RAM, which is known as your L1 arc. So essentially all the RAM that you assign to these appliances, the majority of it is going to be utilized for a read cache here. So, the more RAM, the better. Second level read cache can either be an SSD or an NVMe device and this is essentially going to be an extension of the RAM cache. It’s going to be the caching mechanism there on the cache is a combination of your most frequently and most recently used. We typically see very good performance out of the read cache there. One thing to note here on a rebuild that I like to note is going to be that when you have an actual disk failure in the system, the rebuild will only rebuild what’s the data that’s actually on the disk. If you’ve ever rebuilt a drive in a hardware RAID controller, even if you have a drive failed within, let’s say a few minutes after you created the array and you put the disk back in the array and there’s, there’s really no data even on the array. The rebuild is still going to take quite a bit of time because it rebuilds the entire drive. The hardware RAID controllers aren’t intelligent enough to know where the data is on the actual disk, whereas ZFS will be able to detect that there’s very little data on the disk if there is, and only rebuild the data instead of the whole disk.  

Compression de-dup are available on the system. We do support thin provisioning on the actual LUNs on the array. We have the ability to take snapshots and clones of the volumes on the array. And as far as our access protocols from the client systems we support, iSCSI, fibre channel, NFS and SMB as the actual client access methods. With that I’m going to jump into what a typical architecture will look like for the full data protection here. So typically you’re either going to have a converged or hyper-converged production dual node system here on site, and you’re going to have what’s known as Zpools that are actually created within the system. And these pools are going to be able to reside on either of the two nodes that are on the production site. These pools can failover from one node to the other. An active-active setup. You would have one node that typically resides on each system with the ability for both pools to move to one system and then in the event of maintenance failure, anything like that. With that, we’re going to be able to replicate the data that resides on the production system here to a backup pool that resides on a separate piece of hardware here at the local site and then we’re also going to be able to replicate that data to a remote backup server that is offsite. So, with that, you’re able to protect yourself from viruses, data corruption, disk failures, rebuild failures, system failures, a natural disaster, theft, human error or downtime here. So as far as the demo here, what I’m going to show you is close out of that.  

I’m going to show you here the Hyper-converged solution. It’s going to be I have two hosts here in the demo cluster under VMware and also have these two storage appliances. So, the storage appliance here is tied to 17216 128. And this storage appliance here is tied to 17216174. These nodes are actually in control of the storage that is inside of both of these hypervisors and these are going to be the management interfaces for those particular VMs. So, as you can see here, I already have one storage pool available which is actually providing the storage for these four failover tests VMs here.  

So, I’ll just give you a quick rundown of what the solution looks like as a whole. And we’ll go through and create a pool and a volume and setup of a backup plan. So, through the management interface here, we’ll go ahead and take a look at the failover tab because this is actually how the nodes are clustered together to begin with. So essentially, they’re joined together, as you can see, I have edge one and edge two. They’re able to reach each other. My failover status is started here and whenever you create the pools on top of the cluster, they will show up under your failover resources and this is this cluster. Here is what actually controls the movement of the storage pools between the nodes. So, we have rings which are where our cluster heartbeats and information is exchanged between the nodes and then we also have our mirroring path.  

So, as I mentioned before, this architecture here that I’m demoing for you today is actually utilizing the cluster over Ethernet architecture. So, my data is actually mirrored between the storage nodes via this link here. So, what I’m going to show you is I have this storage pool here where I’m actually utilizing or providing this Windows Server datastore. What I have here is you’re able to come under the pool level. So, you create a centralized pool. You have 1.45 terabytes of storage that’s centralized. and then from there, we’re able to create what’s known as Zvols. Zvols are going to be utilized for block storage. So that’s going to be either your iSCSI or fichannel volumes that you want to actually deliver to your clients. Your NFS or SMB shares are going to be on what’s known as a data set. So, from a data set, you create the data set, and then you’re able to actually create either the SMB or NFS shares on top of the data set. The virtual IP is an important concept for the IP based access methods. So NFS, SMB and iSCSI because this is going to be how the dual node system actually provides a singular IP address for the clients to access over. So, no matter which node the pool actually resides on. So, with that being said, these virtual IPs map to an interface on both nodes. So, you pass this IP to the client system and then no matter whether Edge one or Edge two actually is in control, the storage pool, they’re still able to stay connected through this singular IP right here. We do support Active Directory as well as LDAP for authentication on the SMB or NFS shares if that’s necessary. As you can see here, I’m connected to an Active Directory server in my lab and then if you don’t have Active Directory, there is LDAP for available to create a local-users in groups. 

As far as authentication on the iSCSI side CHAP and mutual CHAP are available for security and there are IP and user restrictions on this and NFS shares as well. So, I’m going to create a new pool here just to give you a quick demo of how the centralized pool was actually laid out. So, I’ll go ahead and create the new pool here and just click add new Zpool all the disks that are in your system will be available at that point. 

What I’m going to actually use is a four-way mirror here. So, I’m going to utilize these four disks because these are all the same disks. They’re 1.46 terabytes. I’m going to put those in a single mirror. I’m going to click next. Then you’re going to be able to add your write logs in here. I’m going to click these two disks, one local, one remote. I’m going to add those in as the mirror. And then you’re actually able to add in an SSD read cache here, which I’m not going to do because these disks are actually already SSD. So, reading directly from the disk would be the same as reading from the cache. There’s really no benefit in adding that in we just for spares on the actual pool. So, you can go ahead and assign spares to the pool for in the event of disk failures. And then I’m going to call this pool demo pool and we’ll go ahead and create that pool there. We click add wait just a second for the pool to be created and then the user interface will reload and we should see the storage pool available here. So, now we see our stored our demo pool that we just created as well as the VMware pool that I already had on the system. You know, this pool is going to come in, we’ll be able to see our disk layout on the pool we just created. So, it’s made up of two local and two remote disks and as well as the local and remote for the actual write log here, we don’t have any volumes created for our Zvols calls for iSCSI fibre channel or anything set up for the shares. Just I will go through and just go ahead and show you how to create an iSCSI target. It’s fairly easy. We’ll just go ahead and click next to create a new volume. You would create Click Add New Zvol. I’m going to call this demo vol and then you’re able to add the sign of the size here.  

So, I’m going to show that we can do the over provisioning as well as thin provisioning here. So, I only have 1.4 terabytes of space on this, but I can assign five terabytes to this volume. You have our settings as far as compression block size. So, you have some options here for block size, depending on the system, you’re actually going to attach it to. Typically, the defaults are fine here. So, we’re at 4.88 terabytes will click next. As I said before, we’re able to set up mutual chat IP restrictions on that, which I’m not going to do for the sake of this demo. So, at that point, our volume would be created here on this new pool that we had. And if we had a client system that we wanted to attach to it via iSCSI, we could here, as you can see, I said, you know, we created this to be five terabytes, roughly. But you can see the physical size here is only actually utilizing 56 kilobytes here on the array. So that’s going to be your thin provisioning here versus thick that that space, the logical sides would need to be available here on the actual storage capacity. As far as the native backup and disaster recovery options, I’m going to show you how to set up after you create a volume here. What you would want to do is create a backup and recovery plan. So that being said, we’ll go ahead and click Add Replication Tasks. You come to backup and recovery, add replication task, you’ll be able to click Browse. And we created this in demo pool. So, this is our resource. So, we’re able to see all the Zvols in the system and you can see I already have one that’s already configured in this task so it won’t allow me to do so. We’ll click this demo pool, demo Zvol, and then we’re able to create a plan here to actually take snapshots and replicate the data on that volume we just created.  

So, the way it works here on the retention rules is you take a snapshot every X amount of time and you keep that snapshot for a Y amount of time and you can have many rules here. So, you could say the default here, you’re going to have an hour’s worth of snapshots that are taken every 5 minutes. You’re going to have three days worth of snapshots that are taken every 15 minutes, and you’re going to have two weeks worth of snapshots that are taken every hour. So, we’ll just blow a couple of those away and we’ll just roll with an hour’s worth of snapshots that are taken every 5 minutes here on the source, and then we can actually go here and replicate this off to a secondary location. 

Now, if I want to go here to, let’s say 175, which is another system in my lab, then I would be able to go here to that system and actually replicate to one of the volumes that resided on that system. I’ll click Apply and then you’re actually able to set up a new plan to replicate those to that system. So, let’s say you had a secondary site and you actually wanted to replicate that just daily. So, then you want to keep that for one month. You would be able to replicate the source volume too, you know, a secondary location here at a different interval than you had on the source. I’m not actually going to replicate to that volume because there’s actually some testing going on that particular volume at the moment. So, you’re able to as far as the you click next, then you’re actually able to register vCenter if you want to include VMware level snapshots and store inside of the array level snapshots. Typically, this isn’t necessary, but some users may require it. You’re able to enable a mbuffer, which is a network buffer that can improve the send and receive performance. When you actually replicate to a secondary location and you’re actually able to add in a task description to identify here. So, we’ll just go ahead and add that in and we’ll wait a few minutes and it works on the clock here. So, every 5 minutes we should see after 5 minutes, we should see our snapshots populate here under the snapshots tab of the demo pool.  

While we’re awaiting there, I will go ahead and just show you the actual failover here. So, I know both of these polls are on the same node, but the cluster is running here. So, we’re able to freely move this pool here. So, demo pool, it’s showing up as active on edge two but I clicked move so this poll should export here and be imported on this node. Now as you can see, it is imported on this node. So with that being said, that backup task now moves over to this node and no longer shows on this node. So while we’re waiting on that to populate, I will show you, give you a quick demo of a restore here under the VMware pool, because I actually already have a task running on the Windows Server LUN, which is this datastore under VMware that houses these four VMs. If I were to go to this Windows Server datastore, I would see that it’s running 1,2,3,4 VMs right here. So, to actually do the restore here, which you’re actually wanting to do, is go to your snapshots tab and you’ll be able to come here and view the actual snapshots that are available for the given volume. So, I can see I have snapshots that date back to 12/8, which is about six, seven days ago. And then I have some snapshots in the 30-minute range up here from today. So, to actually restore from one of these snapshots. You have a couple of options. If you come to the snapshot itself, you’re able to click the details here and see what’s written and what’s used on the snapshots and then you’re actually either able to clone the snapshot or roll the entire volume back to the snapshot date. So the roll back is going to be a drastic operation that would take all four of these virtual machines back to this given point in time versus if you needed to do a file level restore inside of the virtual machine or you needed to only restore one virtual machine, what you would want to do is actually clone the volume. I’m going to demo clone and then you can directly attach it here to your target, which I have. This target is already connected to my VMware environment. So then at that point I have that demo clone that I just created. It’s showing up here, it’s attached to my target. Now I’m able to actually come over here to my cluster just to rescan storage on the cluster level. That’s going to go ahead and scan both hosts for the storage device that we just connected to. The target, it was that 13. So, I can see this, it’s actually the same size as 14. So if we just go to new datastore, whenever it scans for the new datastores that create, it’s actually going to detect that there is a UUID conflict with a datastore that’s already mounted in the system because the clone that we just created is an exact copy of the Windows server datastore that’s already mounted in the system.  

So, with that being said, whenever you want to mount a clone volume while the original datastore is already mounted in the system, you need to re signature the volume in order to get it to mount. So, we’re just going to click assign and you signature here. Then we should see this datastore mount after the volume is re signature here all we’re waiting on now. Let’s go ahead and show you what’s inside of Windows Server so we can see that there’s failover test, one, two, three, four and I can see that the snapshot did mount here. So, when it comes into the system, it’s going to show SNAP and ID and then the original datastore name. So, if I take a look here, that’s also going to contain those four VMs from the point in time of the snapshot. 

So that being said, I could either completely restore all four of these VMs or I could individually add one of these VMs in the inventory or attach a VHDX file to VM to gather information off of it if necessary. So, I’ve covered iSCSI storage as it pertains to running virtual machines in a VMware environment. Now I’m going to jump into the file sharing features of the actual storage array itself. So, I’ve already mentioned before we support both SMB and NFS as our file sharing protocols. We have the ability to make those shares directly on the storage pools within the array that allows you to get rid of your typical file server that may be running. Typically, this would be a Windows based machine, so there’s not going to be any need for that one anymore because you can deliver the file shares directly from the array. As I said before, we support Active Directory and LDAP for authentication on the file shares themselves and so you actually create the file share. You’re going to come to the storage pool level and click the shares tab. After you click the shares tab, you’re going to create what’s known as a dataset by clicking add dataset, you would give it a name, choose a record size, cache settings or typically okay, turn access time on or off. We support quotas and reservation which can be set here on the dataset level and then you click add and we add in a dataset, which I have an example here, test two, which I’ve already created. I can see the A next to this, which means I’ve already configured a backup task on this given dataset. And then once you have a dataset, you’re able to create the file shares on top of a given dataset. So, I have the shares below here. I’ve already got a share named edge file share and it’s being shared out over the SMB protocol at the moment. If I take a look at this file share, I can see that I have a backup and recovery task already set up on the file share. Its name as Test two and the retention plan on it is to take a snapshot every 5 minutes and hold the snapshots for one hour.  

It’s important to note here that we want to access this fileshare using the virtual IPs that are assigned to the pool. So typically, you’re going to add another virtual IP to the pool where your users would actually be able to access the file shares and map those to their local PCs. So, in this case, I’m using 172161253 as my virtual IP that I want to access the fileshare over. So, let’s go ahead and browse to that location using File Explorer. If I go to here, I can see I have edge fileshare. So, it’s access much the same way that your typical SMB shares are accessed. If I come to the snapshots tab here. I’m able to go down to the datasets and see that test two here has a task and it’s taking snapshots at a rate of every 5 minutes and holding those for an hour. The nice thing about the housing the file shares on the storage array is the snapshots are also viewable by right clicking on the fileshare itself, browsing the previous versions and there you will also see the array level snapshots populate within the properties tab. I would go into the fileshare. I can browse around and let’s say I jump into the schematic folder and let’s accidentally delete the archive file from the schematic folder. I browse to the schematic folder and I say previous versions. I can see that one minute ago there was actually a snapshot taken. Let’s see if that file resides in there. So, one minute ago there was actually a file called Archive which I just deleted. So, at this point I can actually copy that file back into its correct location or I can restore the entire folder to that point in time. So, I’m just going to choose Restore here at that point, it’s going to roll back every file inside of this folder back. And as you can see, we’ve gotten our archive file back in the folder.