The benefits of object storage are numerous: Its modular design makes it highly resilient, and its flat structure...
makes it easy to use. It's able to scale to petabytes of data -- which makes it a good option for public storage -- and that level of scalability makes it more affordable than traditional block- and file-based storage. Based on these benefits, it's no wonder the open source world has taken an interest in object storage, with Red Hat Ceph leading the pack.
Corporate storage area network (SAN) devices struggle to keep up with the scalability demands of modern data centers. Ceph open source software satisfies these demands with highly scalable object-, block- and file-based storage under a unified system. Ceph object storage writes binary objects to storage layers. Object storage devices (OSD), of which Ceph has hundreds -- even thousands -- provide the necessary storage. Ceph has already caught on with cloud providers and is poised to replace corporate SAN products in enterprises.
Ceph integration challenges
Ceph object storage uses a multitude of storage devices to store small binary objects with a maximum size of 4 MB. Applications write and replicate storage to the OSDs in the current storage group, and then, when you need to access the data again, the OSDs serve the data. The combination of small binary objects and replication makes Ceph a very fast storage offering, because multiple servers work to deliver all of the binary objects to the client.
It can be difficult to integrate Ceph object storage in a VMware environment due to the way Ceph accesses storage. Most Ceph products offer three interfaces: the Ceph Object Gateway, the Reliable Autonomic Distributed Object Storage Block Device (RBD) and the Ceph file system (CephFS). The first of these interfaces is highly cloud-oriented, whereas the latter two are dependent on the Linux OS.
The Ceph Object Gateway provides an Amazon Simple Storage Service-compatible gateway, which allows clients to access storage with RESTful API calls. RBD is a Linux kernel-level block device. CephFS runs on top of Linux as a Portable Operating System Interface-compliant file system. Neither of these allow for easy integration with VMware vSphere and related products.
A fourth gateway
Though it isn't included in the upstream Ceph project, there is a fourth option that makes it easier to integrate Ceph with vSphere: the iSCSI gateway. SUSE created the iSCSI gateway and includes it in its Ceph-based SUSE Enterprise Storage. Currently, the iSCSI gateway is available as open source software, but isn't generally available in the open source project.
The iSCSI gateway turns any Ceph cluster into an iSCSI SAN. This iSCSI SAN has all the benefits of Ceph object storage and is easy to access because any OS includes iSCSI clients. This applies to vSphere, as well. VSphere makes it easy to create a data store on top of an iSCSI volume. The iSCSI gateway is completely invisible behind the storage offering -- for example, vSphere doesn't see the difference between a Ceph cluster and a NetApp filer -- so administrators can continue to do business as usual.
VMware's object storage alternative
If you're still wary of integrating Ceph object storage in your vSphere environment, VMware offers an alternative. VMware Virtual SAN (vSAN) is an object storage product that stores and manages data with flexible containers, known as vSAN objects. A vSAN object contains all data and metadata, which is replicated across the vSAN cluster. When administrators provision VMs on a vSAN data store, vSAN creates an object for each virtual disk and distributes these objects over the SAN.
Although both Ceph and vSAN are object storage offerings, they're completely different. VSAN creates and replicates one binary object for each disk file, whereas Ceph chops these large disk files into many binary objects and stores them on multiple servers, which boosts performance. Even though vSAN natively integrates with a vSphere environment, Ceph's superior performance and attractive open source licensing model make it worth any administrator's consideration.
Object storage set to dominate web and mobile
A beginner's guide to evaluating open source object storage
Which is superior: Swift or Ceph?