Scality RING Powers Iron Mountain’s 500PB IronCloud 2.0 with Scality

Hello everyone. I am Jeff Mack. I'm the VP of Customer Success for the Americas, and I am going to talk to you about Iron Mountain and their Iron Cloud, uh, offering for public, uh, cloud storage.

So, iron Mountain has been around for, uh, for a long time, since 1951, and they offer storage solutions for their customers. So it started with, um, physical assets, uh, from documents to artworks that customers wanted to keep safe. So they had these giant vaults built in the side of mountains to to store all this, uh, comes the digital age.

Of course, customers started wanting to store their data, uh, safely away from their own, uh, sites. And I started offering, um, an early version of Iran Cloud that, uh, customers could use to send their data, uh, to, to, um, to Iron Mountain that included also actually services to scan, for instance, these physical documents that, uh, the customers had sent. And so every month we would take care of scanning that and storing them on their own servers, uh, for, uh, future retrieval if necessary.

0. Um, the, the requirements already at the time were they wanted scalability, so they wanted to be able to grow the, the platform they wanted, uh, security and compliance, so everything had to be encrypted, uh, whether it was in flight, uh, stored, and they wanted to, uh, implement cyber resonance as well. So they wanted to make sure that the data was safe.

Um, and last of course, they needed that infrastructure to be available 24 7, 365, uh, no downtime. They had some trouble with the first version of Iran Cloud. And so in 2024, they decided to do a reboot, uh, and come up with what they are calling internally, Iran Cloud two.

Oh, but they still, of course call it Iran Cloud. And they selected skelity, um, ring to implement that second version, and they decided to go big immediately because they knew that they would need a lot of, uh, space to migrate the data already on Iran Cloud version one to the second version, and to onboard new customers that they were planning to, uh, to, uh, to, to sign up. So right now they created two clusters, touring clusters in the United States, one on the east coast, one on the west coast, and each of these is, has a, a raw capacity of a quarter of an exabyte.

Uh, a little bit more than that actually. So the, the usable capacity of what the customers actually will be able to address is about 170 petabytes. Uh, they have object locking so that they are, um, able to enable the, the compliance mode that they require.

They also have bidirectional replication, as has been described by Ian and Ben before me. Uh, so we do CR between these two sites or the customers that decide to do it. It, it's not, they, they can choose whether they want to do that or not, obviously, because they're going to store the data twice.

They're going to pay more for the, uh, the additional storage and of course cyber availability. Um, KT ring, like we explained, everything is online. Uh, there is no downtime, even if you lose some components.

And, and the, the main use case for Iron Mountain obviously is backups, but it's general purpose as three storage. So they have some customers that also write data, but retrieve the data as well. Um, they only charge based on the consumption, not the egress.

So it's cheaper for the customer than going to alternate solutions, um, in the public cloud. So right now, they started ingesting in the east coast side. Um, their main customer right now is doing Veeam backups, and they have 60 petabytes of, uh, backups to, uh, to upload.

So they started doing that, uh, in early February. That's when the, the cluster was, uh, was put in production. Next slide.

Hey, Jeff. Yeah. Um, so I, this is cross reaching.

Okay. This is the question I had. You know, iron Mountain is everywhere, so, um, Yes.

And then, and they do plan to create more. Did, did you plan to create more rings, uh, down the line? They started in the us but they want to expand to, uh, other regions as well.

So that's the architecture, uh, very similar to what the previous presenters were, uh, were explaining for, for their customers. So they have an F five load balancer in each site, and the data then gets replicated to, or it gets transferred to, to, uh, the cdt, uh, S3 connectors. The, the connectors like, uh, again, Ben and Ian were saying are separate.

So there's a compute parts. They are separate from the actual storage, which is 120 servers, whether high density, super micro servers, 90 drives each, uh, each server can hold almost two petabytes of data. Um, and then we, similar to Nest, as described by Nick, uh, we can also, if the customers choose that, uh, replicate or transition the data to tape for cold storage.

Next slide. So we had to adjust a little bit to meet the needs, uh, of Iron Mountain. There are two things we need to do.

One was that they wanted to do, of course, uh, address encryption. Uh, pretty much everybody wants to do that now, but they did not have, uh, the capability of doing, um, address encryption through the rate controller like some, uh, servers do on the HPE side. Um, and they did not have cell encrypting drives.

So what they wanted to do was use red ads lux. Um, it is much cheaper to do it with standard drives rather than buying the, the self encrypting drives, especially because they have about 11,000 drives in each data center. So that, that would've been a quite a, a, an additional cost for them.

Um, we supported Lux out of the box with the, the ring. However, we did not have, uh, support for the, the automated disc replacements, uh, for Lux. So this replacements when this would be failing would've had to be manual.

So we, uh, we updated that feature of our product to support locks as well. So now when a disc fails, the replacement handles the fact that it's a piece less, the new driver has to be lock encrypted, and that's the first, uh, feature that we had to adjust to. And then the, the second, uh, requirements was replication to tape.

So what, uh, Nick described ness is transition to tape at Iron Mountain. They want to both support transition to tape, but also customers want to both re uh, keep their data on the hot tier and back it up to, to, uh, to call storage. So we had to also do replication to tape.

And, uh, at the time, we only had support for EMF, but they wanted support for aemo. And so that, that's what Nick described. We added support for AEMO in our products, uh, at that time.

So they started ingesting about four months ago, early February. Uh, the, the ingestion rate from the application is about 500 terabytes per week. So we're around a little bit short of nine petabytes, uh, today stored on the, on the platform.

Um, 6,500 operations per second, uh, on the cluster right now, which amounts to about half a billion per day, uh, operations at the S3 layer. Um, and then at the ring storage layer underneath, uh, just to give you some, some perspective. Uh, so that means everything that's, that's done after we do the protection.

Uh, we are at, at the moment we have about 119 billion objects stored, and that's only 6% of the clusters capacity used 60,000 disk operations per second. Um, they write around four gigabytes per second. It's mostly backups again.

So there's, there's not very much reads happening, um, but around one gigabyte per second, um, all the bottleneck or on the end customer side, it's not on the Iron Mountain infrastructure. And as, as of today, the, the, the feedback from their customers have been, uh, overwhelmingly positive. They are migrating more and more customers from Iron Cloud one to ironclad two, uh, whiskey.