While searching for storage to digitize and protect large data stores, the Balboa Park Online Collaborative's IT director came upon a beta copy of an object-oriented storage application from a startup and determined it met his needs better than a traditional SAN or NAS.
The San Diego-based nonprofit online collaborative provides IT and Web services for a group of museums in Balboa Park and a few organizations outside the park. Those services include digitizing and backing up collections from its clients.
Jason Quinn, director of IT for the group, said he seized an opportunity to evaluate a beta copy of the Exablox OneBlox, a scale-out, object storage appliance, in early 2013 to address its big data needs. Quinn said OneBlox gave him the scalability to keep up with quickly growing data stores, allowed users to handle their file restores, and built in storage management features such as snapshots, deduplication and replication.
In this podcast interview, Quinn discussed the pros and cons of object-based storage, the challenges in implementing object-oriented storage, the performance of the Exablox system and the value that the collaborative is receiving from its use of object-oriented storage.
What problem does object storage address at your IT organization?
Jason Quinn: We have a few large projects. We service about 13 different museums and other institutions, so we have a few storage-related issues that are going to require solutions that maybe the museums aren't in the best position to solve through traditional means. We have lots of data. We're digitizing collections, so we have huge amounts of data that are easy to solve with large enterprise-scale systems, but we don't traditionally have the budgets for those.
One of the things that we are doing is just a simple centralized backup service for all the institutions. And the types of storage that we're getting into, they make it difficult through traditional servers and SAN and NAS [network-attached storage] to get to the kind of sizes that we're talking about. One of the beauties of object storage is that it's flexible. You can add capacity to it as you go along.
Another use case that we are looking at is we are constantly digitizing media for all the museums and putting it online, and object storage is flexible for the different types of media that you want to store on it. You don't necessarily need to think about different storage solutions for different use cases. You can just have a one-size-fits-all solution that you then expand as you need it.
Can you describe the evaluation process that led you to object-oriented storage?
Quinn: We realized that we had a few storage issues coming up so we started looking at more traditional SAN and NAS, as well as direct-attached storage. We weren't specifically looking at object storage at all. We were just looking for a solution to suit our needs, and we had a unique opportunity to beta-test a unit from Exablox. So we got a unit in-house, and we started testing it. And we realized that object storage answered a lot of questions that we weren't ready to answer yet, and it solves a lot of problems that we weren't really expecting a storage system to do.
One of the things we like about what it does is the copy-on-write kind of instantaneous snapshot. Anyone who is familiar with traditional files servers from using Microsoft Windows is familiar with Shadow Copy. So, it's kind of that same technology, except it happens for every file as soon as they write, and while they write, actually. You can get fragments of files if you really wanted them.
One of the best features along with that is that the users can access those files if you choose to give them access. So immediately, we were able to stop doing restores for simple 'Hey, I deleted a file, and I can't find it now' [kinds of problems]. You can access all your files from the last day or week or month or however long without needing IT services.
Some of the other features that are rolled in are dedupe and replication to other nodes. It's all built into the system. We see those features in other devices, but generally at much higher price points and much more complicated systems, at least from the startup. The prices generally get out of our depth pretty quickly. So the Exablox business plan, where you can start off one node at a time and then fill each node up, really suits our needs to a tee.
You've talked a lot about what you like about object storage. What do you not like about it?
Quinn: My only real reservation about it so far is with the Exablox device, it presents a different layer of management. And being that we are an IT services company, it's one more utility between the techs and the clients and one more thing to troubleshoot, one more place for a password to be wrong, one more place for a path to be incorrect. I think that's more of a concern right now because we've only been using it for a short while, and the more we use it, the more that'll disappear in the end. When we first started beta-testing the Exablox, performance was still a moving target. Lately, with their last few updates, it's become much more stable and much more enterprise-worthy.
How would you characterize the performance of object storage compared with traditional NAS or SAN systems you've used in the past?
Quinn: My experience is specifically with the Exablox. The performance for the Exablox product started out pretty slow, but understandably I was early on in the beta testing of the software. They've been increasing the speed with each update. In our main use case, we use it as a backup target for [Symantec] Backup Exec. We also use it for some file servers, but those are for mostly smaller institutions, so I'd have a hard time benchmarking that. But for Backup Exec, it is beyond the speed of a single USB 3.0 drive directly attached to a server. So the throughput for it is very good so far.
The couple of institutions that were using it for file servers are smaller institutions, so I don't have a good sense of the performance specs of it in that type of realm, because we've only had a few people using it at the same time. I'm interested to get it in one of our larger organizations, where we have 50 to 100 people touching it all day.
What were the major challenges associated with implementing object storage?
Quinn: The first major challenge was moving files from the file server, where everybody was used to accessing them and putting them on a different device and moving them through a different path. We originally had thought, 'Well, we'll just start from the ground up and have everybody access the devices directly.' But it's harder to manage the devices through a typical Active Directory workflow and scripting because it does have a separate database. So everything doesn't always match up. And you can bridge some of the gap through scripting. But because we have so many different organizations, and they're at different levels technologically, it was a little bit of a challenge for us to absorb the object storage devices directly into our servers.
So what we decided was that we still have our Windows file shares, and we'd present the file shares to our clients through those. And we still let Windows manage all of the permissions and all the access, and the clients don't even know what the back-end storage is. That's been the way we've handled that so far. I don't know if in the future we'll have more integration with Active Directory than we have right now. Hopefully so. But for right now, it's been working great.
What kind of return are you seeing from object storage?
Quinn: It's hard for me to put an exact number on it. As far as our day-to-day, week-to-week kind of return, it helps us in so many ways. It makes backups so much easier. Our biggest backup servers are always our file servers, and we were getting to the point where we needed so much performance that we were going to have to spend as much on a backup server as we would on a live server. And we would be backing up about a half of one single file server in a whole weekend.
So with dedupe baked in and with replication already online, we're able to take the backup load of the file servers off our plate. Now we only have to back up our application servers and our domain controllers and things like that. So it's appreciably made our backup windows much, much smaller, and going forward, the yearly cost that we have to associate with backup servers and storage will be much lower than it has been.
What advice can you offer to other IT organizations about object-oriented storage?
Quinn: My first thought would be: Evaluate it fairly. It seems like a new term, but it's really well-thought-out, and it's the way of the future. Look at the solutions that you're evaluating in the same way that you would another solution, a more traditional storage-based solution, and then think about things object storage does automatically, and try to extrapolate how that's going to affect your business in the future. Like I said, it answered a lot of questions that we weren't hoping that the storage itself would answer. We thought we were going to have to put extra work into backup and redundancy and things like that, and it's one of those situations where you're pleasantly surprised.