How we Help Our Customers to Build Exabyte-Scale Clouds with Scality

Today I have the pleasure of presenting to you with a team here from Harris, France. This is the company's headquarters. Um, I'll give you a little sense of the lineup here in a minute.

Uh, we're gonna talk to you about a bunch of customer stories. Um, we are really in the thick here of cloud storage, object storage file storage to help our customers build these infrastructures for a hundred petabyte, 500 petabyte exabyte scale clouds. And I think, uh, and I hope you find this, uh, interesting.

Uh, so as Alistair said, I'm Paul Pali. I'm the head of marketing at Ity. Uh, I'll leave you to guess as to when that picture was actually taken.

It wasn't, uh, last year. I can hint you that. Um, I'm joined today by a great group.

Um, four of us are here in Paris, myself. I'm gonna provide you with an introduction to the company, a little bit about the market. Uh, I'll be followed by Ben, who's our VP of Customer Success.

He'll talk about one of the customer stories, which was is with a US Bank. Uh, he'll be followed by Nick, who's our product director, uh, here in Paris, working on the ring. He'll talk to you about a super exciting one within the space, uh, industry.

Uh, we have a sales engineer, Aurelian, who's gonna talk to you about a European bank doing some really amazing stuff with ring lots and lots of applications on one instance. Um, Jeff is my colleague based on the east coast of the us. He will talk to you about Iron Mountain.

This is one of the two that we can name publicly here today. The others, we have to keep a little bit masked. And then you're lucky enough to have Giorgio Reney, one of the founders of Ity in the room there with you.

He'll be the end and he'll talk a little bit about the underlying technology. Okay, so with that, let's get rolling. Um, so let me give you kind of the 32nd intro on the company.

I think a few of you have seen some familiar names. Know who we are. Uh, scalability is now a 15-year-old company.

We are still privately held. Uh, we've had an exceptional relationship with HPE for about a dozen years, actually closer to 11. Uh, they distribute the product, they resell the product.

We also go to market through our own network of distribution and channel partners. And what we're known for is really distributed file and object storage solutions. So we develop software.

It's for on-premises deployment onto the customer's choice of hardware. And that's one of the tenets that we've had since the beginning is to offer choice. Customers can put this on their favorite vendor, their favorite form factor of server.

They can grow along with it. And then at some point they have the freedom to choose, uh, to modernize and to go to something more dense and something faster. Uh, we'll talk about mainly our flagship product today, which is the scald ring.

That's the one that's been in the market since 2010. It's really aimed today at solving very large capacity use cases for cloud infrastructure, for data lake, and also for data protection for backup. Uh, we have a newer product that was introduced a few years ago called esca.

It's aimed much more at the mid-market. So it's intended to purely solve, in a very simple way, backup up cyber resiliency for ransomware protection. Okay?

That isn't the focus of today. So we're gonna focus a lot on ring, especially for cloud. Mm-hmm.

So what I wanna do is now kind of tie this in a little bit about when did enter the market and what was happening in the industry. And we all know this, right? Amazon came out in 2006 with a very, very successful storage service.

Amazon S3, if you, those of you that remember it only took them about five years to get to a trillion object stored. So from 2006 to 2011 or so, we knew that this was happening, right? We all saw developers jumping on board.

They loved the idea of having this scalable on-demand platform for storing lots and lots of unstructured data. It helped them build websites, cloud services, and it was just an absolutely fantastic solution. It took only 15 years to get to a hundred trillion objects.

So again, a testimony to the growth and the popularity of a solution like this. It's very interesting to note that when we started, it wasn't yet clear that Amazon, the S3 API would become the defacto standard for object storage. In fact, there were times where we were still looking around thinking, do we invent our own?

Do we jump onto something like OpenStax Swift? Eventually the industry, including the application developers, the ISVs, uh, really recognized that this is the one, right? We embraced it, the app developers embraced it, and we all were able to come together and use this as the defacto.

Now, while the developers were very, very happy and, you know, a certain GL class of, uh, companies were very much using this, there was another kind of influence here, and that was the service provider community. They were starting to get really nervous and think, what happens if these hyperscalers kind of take everything from us? So there was really a lot of pressure to start thinking about how can they respond in kind with their own cloud services that would compete in effect and let them keep their market share.

Um, I was actually part of this industry at the time. I was a consultant to, uh, one of the companies building a cloud vis. Uh, some of you might remember that they introduced something called sais Symphony.

It was a virtual private data center offering, uh, built on VMware, Cisco, and actually equal logic storage. There were many of these, right? They were all just trying to rush something to market, whether it was fast or vast, or an ISAs service, or even email as a service.

That's another category that we saw. Um, even stun Microsystems before they were acquired, offered a cloud. It then became Oracle.

Lots and lots of this was happening. So there was this rush, this pressure to start building cloud services. And these people needed a solution for scale out storage, right?

They weren't gonna go and develop something like what Amazon had had on their own. So what happened is S Scale E entered the market in 2010. Uh, the founders of our company actually had relationships with many of the largest telcos and service providers.

So companies like Comcast, Orran is the big telco here in France. Many of these needed a solution to start building their cloud storage infrastructure. In the end, over the course of the next few years, 10 of the top 20 global telcos build their cloud infrastructure on scalability.

So they were offering public cloud services using our solution to build the storage. Okay? So that was one big trend.

The second one is that now you have the enterprise market and the enterprise, of course, thought on demand, uh, computing, cloud computing, uh, internal chargebacks. This is a great model, right? So we started entering the enterprise ourselves and ultimately found that we, uh, found a lot of interest within financial institutions, manufacturers, other sort of global 2000 companies, building their own private cloud storage or their own private clouds.

They needed a storage infrastructure. Today we can say that seven of the 15 largest global banks have their private cloud storage built on scalability based on scalability. Um, so that's sort of the advent up until about three or four years ago.

Now, what are we seeing? We're seeing, really seeing that cloud is broadening. So cloud storage is now becoming of interest to people building things like data lakes, but also high performance analytics.

Um, and then ultimately data protection. If you look at the service provider community, the cloud service provider community, you see a lot of data protection offerings, disaster recovery as a service, ransomware protection. We have huge customers here.

This is where we starting to see a hundred petabyte plus scale customers deploying data lakes or aggregating data that needs to be prepped and cleansed before it hits the model stages. Cybersecurity, deployments, genomics, uh, within analytics, you'll hear about one particular use case here from Georgia later that has, uh, the need to store a petabyte per day of data. So that's really high performance on top of the object store.

And then we're truly seeing exit byte scale. So I'll leave the news, um, for our presenter on Iron Mountain. Alright, so how do we go to market?

We have two different differentiated solutions. Again, the one will that we'll focus on today is ring. Uh, this is our high capacity solutions.

So think about this as a very scalable object store that can grow to any capacity, need and also scale out, scale out performance. And as you'll see in, it can scale out in multiple ways. It's really a great fit for people that are thinking I need hundreds of terabytes, but eventually I might end up in this, you know, uber arena.

Eventually get the exabyte scale. Uh, again, esca is kind of the other side of it. It's the mid-market solution.

Very, very simple single use case purpose built for backup. Um, so the question then to ask before we talk about all these customer stories is, why is Ring ideal as a private and public cloud store? Uh, the first thing, of course is this API support.

We've spent eight years now on building really, really deep and broad and compatible S3 API support. And this means the basic APIs work just as they work in Amazon. Uh, moreover, we do all of the advanced APIs for locking, for cross region replication, for lifecycle management and more.

Uh, we decided to also build our identity and access management capability, our multi-tenancy on top of an a AWS compatible model. So what does it mean? It means that when you administer ring, it looks like Amazon.

You create accounts within accounts, you have users, you have groups, you have roles, you have policies. It all looks very much the same, okay? And that creates this sort of isolation within a shared system so that you can post multiple applications, multiple different use cases.

On one instance, uh, we do distributed data protection. It's part of the core IP of the product. Uh, Giorgio may reference it a little bit, but distributed eraser, coding, integrity, checking self-healing against failures.

That's all been in the product since day one. Uh, for a long time, we've offered a geo script solution that can span multiple data centers. So think of a single instance of the ring across, say three data centers.

If you have a failure of one data center or a network outage, the system stays up and running. We provide a hundred percent uptime SLA in that environment. Okay?

Uh, what else else do cloud builders need? They need utilization, tracking how much data, how much data's being consumed? How can I report it?

Do you have QOS? Do you have capacity management? All of that's in place.

Uh, we've invested in two special things as well. We believe that being rock solid from a cyber resiliency perspective, especially against cyber threats, is a big, big, uh, need today. That's called core five.

I have a separate slide on that. And then I wanna mention what we call multiscale, which is really our form of a dis disaggregated architecture. Okay, so let me hit that one first.

We believe that the real complexity in cloud, as you'll hear about it today, is that the requirements vary so much, right? It's really unpredictable. Are you gonna see a cloud environment that has 10 apps or a hundred or a thousand?

All of these apps have different workload patterns, different data, si sizes, different IO needs. It's really not even an IO blender anymore, it's an IO monsoon, right? So this ability to sort of disaggregate the system in every different dimension is what Ring is unique at.

And this has been in the architecture since the beginning. If you need to dedicate more S3 endpoints, you can scale it out. If you need to decouple those S3 endpoints and put them on separate servers, you can do that.

If you need to scale metadata for more buckets or more objects, you can do that. If you need more S3 hops per second, you can do that. And of course, the backend can be scaled out.

So think of this as this sort of ultimate m by end disaggregated architecture that achieves this kind of multidimensional scaling. Uh, the last thing I wanna mention, and then I'll introduce my next speaker is cyberresiliency. C.

So there's a lot of talk today about immutability, especially in this, you know, ransomware protection, cyber resiliency arena. We agree that that's absolutely needed. Uh, but there are so many other possible attack vectors for a dedicated, you know, a dedicated, malicious, uh, actor in the data path is sort of the common one.

People come in and they try to destroy data, delete it, overwrite it, encrypt it, uh, but there's also attacks on the human users, on the network, on the data itself, on the infrastructure, on the data center. We need to protect this at all levels, and that's something that we do. Uh, we have a branded term for it called Core five.

It's really zero trust architecture based, but then with multiple capabilities in the line of each level of the, uh, of the software stack to protect against these things. Hey Paul. So With Yeah, go ahead.

You mentioned, you know, scale of exabytes, hundreds of petabytes. I, do you have any, obviously you've got, uh, multiple ring, uh, you know, mirroring kinds of capability. Do, do you have any sort of offline backup capabilities?

Uh, Backup. Backup. Okay.

You're talking about backing up our system to a, an external target. Yes. Yeah.

What we support is async replication to another target. We could go to a cloud, that's one way to do it, but people generally aren't backing up at a hundred petabyte scale. We do support tape, and that's one of the things one of my colleagues here Nick will talk about.

Um, so you can have a cold store behind the warm store and you can also have a hot tier. Uh, but I'm not seeing sort of traditional, let's do a full backup or an incremental backup of a hundred petabyte plus, uh, object store. And then, uh, do you offer any sort of, uh, automated tiering, um, between we do current scales of performance Yeah.

Or economics? Yeah, there's an internal turn tiering capability that's twofold. It's number one based on the media type, but also the data protection, the durability algorithm.

So you could, for example, say I want a tier one hot on PLC flash, and I wanna replicate it so I don't have the erasure coating overhead. And then you could tier down to QLC. You could have the QLC be erasure coded, and you can have many of those tiers.

So ultimately it's not just bound to two or three.