Caringo Inc. today teamed with Dell to introduce a disk-based appliance for petabyte-scale deployments of Caringo...
Swarm object storage software.
The new object storage server hardware system, branded as Caringo Swarm Server, is based on Dell PowerEdge technology. Dell sells Swarm Server directly and handles hardware support.
Swarm Server prepackages Caringo Swarm 9 software, which Caringo released Oct. 6 with upgrades for data protection, erasure coding, storage quotas and replication. Customers may still buy Swarm 9 as a software-only installation. Swarm Server is backward-compatible with earlier editions of Swarm.
Dell and Caringo have been longtime partners. Dell sold Caringo software on its hardware as the DX Object Storage platform through an OEM deal from 2010 to 2013 before discontinuing that product. After dropping DX Object Storage, Dell sold Caringo software as a download to run on Dell servers.
Caringo's object storage server rollout follows its August launch of Caringo SwarmNFS, allowing direct ingest and presentation of file-system shares to an NFS plane. In 2015, Caringo added FileFly, which plugs into Windows NTFS to migrate aged data from NetApp filers to back-end Swarm object storage.
Caringo storage cluster scales to hundreds of petabytes
Projected use cases for Swarm Server include archiving, backup, big data, bulk storage, content distribution and cloud storage. According to Caringo, Swarm Server is optimally deployed as a four-node hardware cluster, including one Swarm M1000 management server and three Swarm S3000 storage server nodes. That recommended configuration provides 288 TB of raw object storage.
Tony Barbagallo, vice president of product at Caringo, based in Austin, Texas, said customers frequently request Swarm bundled on hardware. Swarm Server nodes can be ganged together to scale to hundreds of petabytes. Barbagallo said one beta customer, an undisclosed government agency, has scaled its Swarm Server cluster to 16 PB.
"We get into a lot of enterprise organizations that just want a whole solution ... they want one throat to choke," he said. "We had to make some tradeoffs -- as an appliance, you want to minimize [the number of] SKUs. It's kind of the 80/20 rule. This object storage server model is going to solve most of the customer use cases we tend to see."
Swarm Server M1000 is a Dell PowerEdge R630 model. The 2U Swarm Server S3000 storage server uses Dell's PowerEdge R730.
A single Swarm S3000 storage node contains 12 slots for 8 TB near-line SAS disks for 96 TB of raw capacity per node. If purchased as recommended, four servers create a system with 288 TB of raw object storage with Swarm 9. The system supports NFS, SMB, Hadoop Distributed File System, Amazon Simple Storage Service and REST APIs.
Swarm M1000 management server is a 1U chassis with 18 CPU cores -- Intel Xeon E5 2695-v4; eight 1 TB 7,200 rpm near-line SAS drives -- 4 TB with RAID 10; and eight 16 GB DDR4 RAM modules. M1000 is preloaded with the full suite of VMware vSphere 6 management tools.
Device connectivity includes two 10 Gbps dual-port Intel x540 interfaces -- four ports -- and a 1 Gbps Ethernet management port. An internal module holds two SD cards to boost EXSi hosts.
Swarm Server does not come with an operating system. Swarm 9 runs directly in RAM. Swarm S3000 storage server is set to run a Preboot Execution Environment boot over a high-speed network. RAM updates the cache and makes searchable index data available to the storage cluster.
Swarm 9 release fortifies elastic metadata search, data protection
Caringo Swarm evolved out of Caringo CAStor, a content-addressed storage system. Enhancements in Swarm 9 include the ability to annotate metadata without having to create a new object, which is useful in cases when provenance rules exist to govern the original data. Authorized users with role-based access can attach additional metadata without modifying the original reference object. Swarm 9 metadata annotation works with versioned objects of structured and unstructured data. All metadata is indexed in Caringo's elastic search. Swarm 9 allows users to create an NFS mount point from saved metadata search results.
Swarm 9 data protection policy management encompasses data replication, erasure coding and unlimited object versioning. Permitted users can apply replication or erasure coding to different files residing in the same object bucket.
"Our goal is to minimize the learning curve of whoever is administering the storage," Barbagallo said.
Caringo added whole-disk encryption with AES-256 ciphers to protect data at rest across a storage cluster. Storage quotas are not new in Swarm, although previous versions required users to set quotas via programmatic HTTP calls. Version 9 is designed to help storage administrators delegate self-service to users or groups in specific roles.
Once logged in to Swarm, a user is allowed to provision networking and usable storage to a bucket, domain or tenant, within an upper limit set by the administrator.
Caringo's object storage server reacts to changing market demand
Object storage uses a flat address space to store files, images and data blocks as individual blocks. Each object is assigned a unique identifier for retrieving the data from any location.
Early object software vendors have been acquired by or partnered with hardware vendors. In 2015, Western Digital bought Amplidata and IBM acquired Cleversafe. Scality has hammered out reseller agreements with Dell and Hewlett Packard Enterprise for its Ring object storage. Cloudian added scale-out arrays as targets for its HyperStore object storage software stack that runs on x86 servers. Quantum sells Western Digital's object storage on Quantum Lattus systems.
Steven Hillsenior analyst for storage, 451 Research
Proprietary object storage platforms include EMC Elastic Cloud Appliance, Hitachi Content Platform, NetApp StorageGrid, Data Direct Networks' Web Object Scaler, Exablox OneBlox, Huawei OceanStor UDS, Nexsan Assureon and NEC Hydrastor.
Software-only object storage startup Storiant had a short-lived run before shutting down in April. In May, newcomer OpenIO joined the fray with object storage software that integrates industry-standard server capacity as a grid of capacity. Open source contestants include Basho Riak CS, OpenStack Swift, Red Hat Ceph and SwiftStack Swift.
Steven Hill, a senior analyst for storage at 451 Research, said giving customers an optional appliance makes sense amid the shifting market.
"There are always going to be companies that want to run Caringo software-defined storage on their existing infrastructure. The appliance model is perfect for turnkey customers who don't want to get involved in the minutiae of streamlining and optimizing storage architecture," Hill said.
"Partnering with Dell is a pretty good choice. Caringo has basically built out everything it's done in storage and put it in an integrated module."
Caringo said a four-node Swarm Server cluster sells for $164,000. That breaks down to $48,000 per each Swarm S3000 storage server and $10,000 for the M1000 management server. The price also includes three years of maintenance and support, plus all VMware licensing.
Your primer on scale-out vs. object storage
Software-only or gateway appliance for object storage?
Caringo recasts CAStor object storage as Swarm