Unboxing 3 PETABYTES of storage!!

In fact, it's surprising how many wires were crossed on this project, but we're finally doing it. We're building petabyte project number two and at some point not too soon we may not even have enough space left on our servers to download the stuff we're on. recording right now and linus, you could say that you could stop being a digital hoarder and, oh, I don't know, delete some damn data, but I actually have the perfect counterargument for that, you sound like my wife just lets me have fun and Let's Let's have some fun today, ladies and gentlemen, because I accidentally have over three

petabytes

of hard drives.

Intelligent deployment makes it easy to handle daily tasks, such as patching Windows images, updating applications, and migrating user data. You can do this on all your existing files. network or cloud without leaving your desk get your free offer at smartdeploy.com linus okay, so the first problem was totally my fault. In fact, I told Seagate that our goal was to show a petabyte of usable space in a single 4u enclosure instead. in two enclosures like we did last time and I told them that to do that I would need 75 of their 16 terabyte hard drives to account for the space we would lose formatting general and parity data, so that's true on five zfs raids . z2 we could lose up to two drives per v dev, so that's up to a maximum of 10 of our 75 drives before we actually lose data and that would still produce over 950 terabytes of accessible space. 75 unit custom chassis I thought I ordered 45 units.

They said yeah, yeah dude, your server is in the mailbox, you're welcome dude, they're from Eastern Canada, they really sound like that, it's amazing and I was like, Is that right? 75 custom drives and they say what are you talking about, man, apparently I never clarified that I needed 75 bays, so it has 60. So it looks like we're going to have about 750 terabytes of usable space, but wait, wait. Guys, the title is not clickbait. I'm still going to have a petabyte of raw capacity here, the difference is that we're going to make up some of that shortfall with solid state, so let's take a look at the drives Seagate shipped. over here so there's actually more paintings here than I expected which is interesting so we're all learning things today um this can go here so these are the right ones these are the ones we can do first this is How funny oh Oh my goodness, there are so many of them it's actually like this is adding up quite a bit.

I'm not really sure how it happened, but this was mistake number two. Seagate Iron Wolf Nas drives are designed for network attached

storage

use. respect a million-hour mean time between failure and 180 terabytes of access per year, have a three-year warranty and feature the Seagate Agile Matrix, a combination of hardware and firmware features that make them perform better in the raid arrays they have. RV sensors and better vibration tolerance to improve performance and reliability in multi-drive arrays and a combination of solid performance and power consumption in a wide variety of workloads, including video editing, which is our core concern, here we end up editing videos directly from the vault more often than you probably think, either because it's a big project and the wanik server doesn't have space or because a sponsor wants to change after the fact or whatever the case, so thank you seagate, I appreciate you family, these are great units and we have recommended them many times, except for one small problem.

I don't really know where the communication cables were crossed, but they are rated for use in cabinets of up to eight units at a time. Well, then Seagate sent some. Their Iron Wolf Pro drive cases give us a total of more than two

petabytes

of

storage

, but they are also designed to have up to 24 drives in one case. Honestly, I would have been perfectly comfortable with the ironwolf pro units. They have an extra two years of warranty compared to the regular Ironwolf, have data rescue service included, and are rated higher for both annual usage and mean time between failures, but we're supposed to set a good example for you and when I clarified, hey guys, the plan is actually to put all the drives into one system that they sent to the big dogs, meet the exos 16 at its current maximum capacity of 16 terabytes, each of these is rated for a 550 terabytes per year of access and two and a half million hours between failures with all the vibration detection and mitigation technology at Seagate's disposal to then qualify them for an unlimited number of drives per enclosure, oh, let's add them to the stack right now.

I have never seen so much storage in one place in my life. This is more than three raw petabytes of storage. 225 drives for 16 terabytes each. The bad news is that Seagate says I have to use previous shipments of our drives for other things or return them. so make sure you are subscribed so you don't miss out on more nas creation collaborations with other youtubers. That's one of the ideas I had anyway. Let's take this opportunity to take a closer look at our venue. Now I am very excited about this server. So this is an early model.

This is a prototype of their next-generation Storinator. Here we go from outside. It looks like a normal old airplane Storinator with 45 units. Typical sheet metal construction. All that good stuff, but oh, that's different. They have improved. their game, so they actually went from a cable based backplane system where each port was individually connected to these PCB backplanes, so this dramatically simplifies the wiring as they are only running a four port sas cable to each of the four bays and In theory, they should also improve reliability. They've also introduced some logic to stagger the turns of the unit so that you don't get the same kind of power surge when you first turn on a storinator and all the units, up to 60, are just like starting.

It's like accelerating really well, although I'm not ready to build this yet because there's another surprise now. I don't know how much of this came about because of my request or how much they were already working on, but what is this? then 45 units finally joined the red team, that's right, we have an AMD Epic processor here. I'm actually not one hundred percent sure what exactly the model number is and then it's equipped with what we're looking at here. 128 gigs of RAM now there was a little mishap on our unit and this is like an engineering sample board that the gigabyte provided so I couldn't really get a new one.

Two of these memory slots are dead, but this is based on a hard drive. storage system, so I'm not really too worried about the extra couple of memory channels killing our system performance like it did on all of our NVME NAS. That said, I alluded to the need for some SSDs to make up the difference in capacity between the 60 drive and the 75 drive which I thought 45 drives were running on, that's where they come in and I need to figure out what the heck they are, here we go, so there is a zfs feature called adaptive replacement cache or arc and essentially what it does is it takes the most frequently used data from your hard drives and then stores a second copy in your system memory so it doesn't have to go up your spinning disks to access and that's especially important for something like running vms or a database where a lot of the searches will be done on the same few entries that we're already using in the existing petabyte project, you know which one you have the two bays.

Not using is something called l2r or level two bow. Sure you can add hundreds of gigabytes or even terabytes of memory to a system now, but the cost can be quite prohibitive, so that's where SSDs come in, so this one here is a 7.68. terabyte ssd, that is almost 8 terabytes ssd per micron, yes it is only sata 6 gigabit per second, but although it is obviously much slower than system memory, it is much faster than spinning disks and easily enough to saturate our 10 gigabit or even 40 gigabit network connection. and then, more importantly, it's much cheaper than simply adding more RAM to the system.

Something to keep in mind here is that we ended up putting six of these here so that our raw capacity reaches the petabyte that we promised but l2 arc. It doesn't actually scale particularly well with a lot of capacity, so it's possible that what we end up doing is just using part of it by cheating a little bit, but for the moment, we have a petabyte of capacity too, just like has a copy of the data so it doesn't really count towards your total capacity, but these are just minor details, like knowing which direction the drive is going for example, I know I'll be judged very harshly for this, but while I might put a dual 10 gigabit midcar right here in this open space in the back.

I actually decided to use one of our old 40 gig infiniband connectx 2 cards because it had a slot that didn't have a cover and Anyway, I don't know what else to do with this old thing and I mean, it's 40 gig even though it's a card older, it's enough for hard drives, so it's there with its 3D printed support from that video from long ago. a really cool feature so i have actually done this in a silly way like hot gluing or double sided taping a fan on top of my hba cards so these are the controller cards for all of these drives that we're going to connect, but hey, now there's an official way to do it, so we have a cooling fan to take care of all of our add-in cards here, very nice, now that they have so much space here that they could definitely modify. the chassis and make a third fan here if they really wanted to, that's incredibly awesome, uh, this is an example of a really old storyteller, this is from about five years ago, you can see there's a lot of things about it that are a lot worse so this horrible horrible mounting system for the drives with these rods holding the thing up wasn't great you can see there's a lot less space here of course I haven't even talked about the new mounting system for the drives okay, let's do that, but first I want you to see how they used to wire them, what a nightmare, we've actually done a swap for a dead port on one of these things and it wasn't a It's so funny, this right here is apparently a guy of 3D printed friction mount and then works with the spring mount they already had on some of the newer chassis along with that PCB backplane which apparently makes it easier to line up the slots perfectly. which is easier to put drives in and out so let's see if that actually worked well so it used to be fine anyway but oh oh okay that's not bad so all that's left now is to install 60 units, well, 58, I already made two.

You can tell this is a prototype, it has a lot of scratches, dents and stuff. I think this was actually their work sample before they sent it to me. You know, you could call it used or you could call it pre-tested at home. Stretch and 900 60 terabytes of raw spinning storage along with 46 terabytes of SSD storage for a total of one petabyte in a single chassis, but I'm not done yet, the only thing missing here is a working device, so I've already talked about it. about read caching, which doesn't really have any danger associated with it because you're just making copies of this data to put in your RAM or your L2 arc, but proper caching and write caching is something that you can do with zfs and you can use. your memory for it, but the problem is that in the event of a sudden power loss, which who knows could happen, any in-flight data that is stored in RAM but has not yet been assigned to your hard drives will be lost. , so it might be worth it. add something like an octane ssd to one of our pci express batches here to handle the caching data being written so we don't lose it in case it's in ram limbo and hasn't been committed to the persistent storage completely.

That, however, will be reserved for part two, where Anthony and I are going to team up to get this up and running on the network so we can start downloading some of the data from the original vault to the new consolidated vault. Yes, my friends. 2 volt capacity is now 1 volt, isn't that amazing technology, talking about amazing technology, Pulseway is real-time remote monitoring and management software that helps you troubleshoot on the go by sending commands from any mobile device, it's compatible with windows mac and linux and pulseface unique app gives you remote desktop functionality. You can get the status in real time.

Resources of thesystem registered in users. Network performance. You can manage Windows updates and more. With Pulseway you can create and deploy custom scripts to automate your IT task and you can scan and install. things update all your systems on the fly, it's super cool and super powerful and you can try it for free at pulseway.com or through our link in the video description, so thanks for watching guys, if you're looking for something else to watch, why not? Look at the epic saga that was our NVME storage server upgrade. Yeah, we're basically replacing the entire server room right now if you didn't notice that.