Is CloudByte Really the Future of Storage?
A few weeks back I attended Storage Field Day 4. One of the storage vendors we heard from is CloudByte. CloudByte is a startup that has been around for about three years. They are strictly a software company that works with commodity hardware and they’re product is available for a free trial download here. Their target customers are service providers and enterprises. They work with VMware, Citrix, OpenStack, and CloudStack on the application side of things with Microsoft on the road map.
As with many of the other vendors we saw that week, CloudByte is touting a scale out infrastructure with whatever kind of storage hardware you can get your hands on (as long as it’s on the Hardware Compatibility List of course. What they kept coming back to is that they’re offering ways to efficiently and predictably use the storage so that you’re not spending all your money on really expensive flash when perhaps your business doesn’t require it. They offer unified storage via CIFS (or shall we say SMB), NFS, Fibre Channel, or Direct Attached Storage while putting all your disks in the same storage pool to again use it more efficiently. However, there is no auto-tiering option so tiering is mainly handled by an administrator. Also, it should be noted the Fibre Channel component is a separate add-on and many features available with NFS will not be available using other protocols due to protocol limitations.
As you can see from the picture below there are two main components in this architecture: ElastiStor and ElastiCenter:
ElastiStor is the software you’re actually downloading and is where all of your data is actually stored. It’s capable of communicating and coordinating with other ElastiStor nodes. ElastiStor is a essentially an abstracted layer across multiple disks, and contains what CloudByte refers to as Tenant Storage Machines (or TSM). Each TSM has a controllers along with networking and storage to make this abstracted layer work across heterogeneous storage that may even be across multiple geographic locations. TSMs can be migrated between ElastiStor nodes should one have a very high workload.
As I mentioned above, though they don’t do auto-tiering between storage tiers, moving the TSM from a storage pool with lower performance to a storage pool with higher performance and vice versa is possible. Many of these features are possible through the use of ZFS, which is the file system ElastiStor is using under the covers. It supports up to RAID-Z3 which allows you to lose more disks and reduces risk of losing data during disk rebuilds, especially as disk sizes continue to increase.
The ElastiCenter component is the management piece, which is actually optional. Since CloudByte offers APIs to connect to ElastiStor companies can use their own management software to integrate CloudByte into their current environment. This is also where the analytics are captured and aggregated which can be used for capacity analysis and reporting. ElastiCenter is also used to monitor and configure things such as TSM details, where TSMs are located, limits and “grace” for usage, and so on. Grace allows for bursting within a TSM and can be configured with hard limits or by using policies.
CloudByte offers HA through creating HA groups with 2-4 ElastiStor nodes and can be configured via ElastiCenter. They’re all using the same underlying storage which is how one node can take over if one node fails. Each node would have access to its own primary pool. If a node goes down one of the other nodes would take over as owner of that pool. Within the management UI you can also create and schedule snapshots as well as create disaster recovery TSMs. This is not yet integrated with VMware, but this is also on the roadmap. You can use the custom APIs to take advantage of some of the VMware features currently, though. Keep this in mind if you’re using any backup technologies that utilize snapshots.
CloudByte is definitely worth checking out. They’re offering many features such as working with multiple protocols, QoS, and storage in multiple geographic locations that many other scale-out storage vendors are not offering. It also seems that some of the things brought up during the presentation that they don’t have currently are on the roadmap. Not to mention the ease by which you can evaluate the software by simply going to the website and downloading it.
Here’s video straight from their presentation at TechFieldDay: