Theoretically, almost any storage system and software can be used to build a private storage cloud. But over the past year or so, vendors have come out with new products or re-worked old ones
Bycast Inc. StorageGRID
Object-based StorageGRID started out being used mainly for medical archiving, but Bycast added support for clustered NAS, security partitions, chargeback and virtual servers last year, and now markets StorageGRID as a building block for private and public clouds. Adding discrete security partitions provides multitenancy and lets IT departments support multiple internal customers.
StorageGRID cloud storage software virtualizes storage across locations into one pool. It runs on most vendors' storage systems and is sold through OEM partners such as Hewlett-Packard (HP) Co., IBM Corp. and Iron Mountain Inc., as well as VARs.
Bycast has two pricing options for StorageGRID. Customers can buy a perpetual license based on managed capacity or an annual license for a pay-as-you-go model. Perpetual licenses typically cost approximately $2,000 per terabyte. Annual license fees per terabyte decrease as system size increases. As an example, Bycast said a license for 500 TB is typically about $1,500 per terabyte.
Caringo Inc. CAStor
Caringo was among several vendors to tweak existing platforms to make them more of a cloud play last year. In Caringo's case, it added support for virtual machines and a Cluster in a Box feature that lets customers assign one or more drives to each virtual CAStor node in a physical server chassis using multicore processors for better performance.
Caringo also added a free CloudFolder feature. CloudFolder is Windows application that lets customers drag and drop files into a folder on the Windows desktop. The files will automatically be added to a CAStor cluster, either at the custom site or at Caringo's own test cluster at its data center. If the data is sent offsite, it's sent without encryption, although Caringo officials said encryption is planned for future releases.
CAStor also supports replication among clusters.
Cleversafe Inc. Distributed Storage Network
Distributed Storage Network (dsNet) lets service providers build public clouds or organizations set up a private cloud using an information-dispersal algorithm that "slices" data for distribution in different storage nodes. The algorithm contains added code that scrambles the data for encryption and allows the data to be rebuilt even if some nodes go down.
Although its value lies in software, dsNet consists of three hardware components: the Accesser, the Slicestor and dsNet Manager. The Accesser provides an on-ramp to the cloud by slicing data before sending it over the wire. Slicestor appliances sit on the network and can be geographically dispersed or centrally located. The dsNet Manager is an out-of-band appliance that monitors the dsNet system and rebuilds missing or corrupt data.
Last year, Cleversafe added a Smart Client feature that works as a content delivery network (CDN) and takes into account network performance before retrieving slices over the network. The dsNet file system is accessed through a WebDAV interface, which allows customers to create and allocate partitions and pool and redistribute storage. dsNet supports object-based storage and iSCSI, with plans to support CIFS and NFS. dsNet with a 1u Slicestor with support for up to 4 TB raw storage, an Accesser and dsNet Manager has a list price of $9,996.
DataDirect Networks Inc. Web Object Scaler (WOS)
DataDirect Networks launched its Web Object Scaler (WOS) system last year as a direct challenger to EMC Atmos. WOS scales into the petabyte range -- up to 6 PB in a 100-node system. However, the minimum capacity of two 7.2 TB nodes is smaller than the Atmos starting configuration of 120 TB.
The 3U WOS 1600 is available as a 7.2 TB SAS node optimized for performance or a 16 TB SATA node optimized for capacity. The 4U WOS 6000 can be configured for capacity, performance or both, and can hold up to 60 TB. The WOS nodes make up the cloud, and WOS appliances can be clustered across geographically dispersed sites with a global namespace. WOS also supports file replication and can distribute data across sites for load balancing and redundancy.
DataDirect Networks offers a software agent called a WOS-library (WOS-LIB) that sits on a customer's application servers and translates between user upload/download requests and placement of data on the back-end system. Customers' applications must write to an API in the WOS-LIB to access the WOS cloud.
EMC Corp. Atmos
EMC Atmos is software that turns commodity servers and disk into a storage platform, although EMC only sells it bundled with hardware. Atmos cloud storage software runs on x86 1U servers and SATA JBODs. Atmos is available in three hardware configurations:
- The WS2-120 has eight servers and eight enclosures, each holding 15 2 TB SATA disks, for a total of 240 TB.
- The WS2-240 holds 16 servers and 16 disk enclosures, each with 15 2 TB SATA drives in each enclosure, for a total of 480 TB.
- The WS2-360 is more of a capacity configuration, with six servers and 24 enclosures, each with 15 2 TB SATA drives in each enclosure, for 720 TB.
Early this year, EMC introduced GeoProtect, a RAID-like data protection option that splits fragments of data within each data object among geographically dispersed locations. With GeoProtect, data can be restored if fragments are lost.
EMC claims Atmos scales to manage petabytes of data, using a unified namespace to treat multiple systems as one. Atmos uses object-based metadata to allow users to set policies that determine where to store information, which services to apply to it, how many copies should be stored and in which locations. REST and SOAP Web services are built in, as are capabilities such as replication, versioning, compression, data deduplication and disk spin-down.
Hitachi Data Systems Content Platform
Hitachi Data Systems last year recast its archiving platform as a cloud storage platform built on its Universal Storage Platform V and VM storage systems. With HCAP, Hitachi Data Systems added support for logical partitions to segregate data and administration for multitenancy, access rights to prevent unauthorized access, and encryption. It also includes a REST interface. Secure multitenancy allows customers to provision services to multiple business units for private clouds (it can also be used by server providers to set up public clouds).
IBM Smart Business Storage Cloud
IBM packaged its Scale-out File Services (SOFS) and storage systems into pre-configured storage cloud bundle options called Smart Business Storage Cloud last October. SOFS is a management "wrapper" around IBM's General Parallel File System (GPFS), which was designed for high-performance computing (HPC) environments.
Beneath the SOFS scale-out NAS cluster, customers can pick of any of IBM's enterprise data storage systems, including the XIV Storage System and the System Storage DS3000, DS4000, DS5000, DS8000 or DCS9900. They can also choose to integrate Tivoli Storage Manager (TSM), TSM Hierarchical Storage Management (HSM) for tiered storage data migration and LTO-based tape libraries for backup.
Instead of basing its cloud offering on object-based storage with management based on API integration between applications and storage repositories, IBM is using standard network file protocols NFS, CIFS, HTTP and FTP.
ParaScale Cloud Storage (PCS)
ParaScale Cloud Storage software turns commodity Linux servers into a parallel NAS farm. ParaScale CEO Sajai Krishnan calls PCS the "first shrink-wrapped cloud solution." PCS supports virtual servers, multitenancy and user self-service through the SOAP and REST Web services APIs. ParaScale Cloud Storage can also serve as a backup target for virtual machine images, and can boot and run live virtual servers from the cluster while maintaining data consistency for attached applications. By supporting synchronous replication to up to eight targets, PCS can continue replicating as the virtual machine is being written to. Synchronous replication between nodes keeps all data current within the virtual machine.
It's available as a download from ParaScale's website and runs on any commodity hardware. Pricing starts at $.095 per gigabyte.
Symantec Corp. FileStore
Symantec repackaged its Veritas Cluster File System into the more cloud-friendly FileStore last October, adding support for file-based replication and CIFS, HTTP and FTP to Cluster File System's NFS support. FileStore is designed to let organizations add new storage or replace nodes while staying online. Symantec claims FileStore scales to 16 nodes and 2 PB of total storage, and can handle file sizes as large as 256 TB and up to 200 million files per file system.
Symantec uses FileStore for file-based storage in its Symantec Online Backup Service, but it also sells it to organizations looking to build private clouds. FileStore pricing starts at $6,995 for two nodes and two CPU sockets. Symantec sells it without hardware, but partners such as Fujitsu and Xiotech Inc. will bundle it with their hardware.
For more on cloud storage:
3. We explain how internal private cloud storage makes its way into larger enterprises
This was first published in February 2010