Cloudian's Samurai S3-clad object storage warriors

The software side

Besides this enormous hardware change, the HyperStore software development is relatively minor. To put it in context, the potted HyperStore version history looks like this:

• 2011: v1.5 – Peer-to-peer scale-out, multi-tenant storage software
• 2012: v2.3 – S3-compatible API and multi-region storage
• 2013: v2.4 – Billing/chargeback by region/tenant, server-side encryption and Citrix CloudPlatform connector
• 2013: v3.0 – Compression and a virtual appliance
• 2014: v4.0 – Erasure coding and OpenStack connector, NFS support and archiving to any S3 cloud
• 2014: v5.0 – More scaling with vNodes, and monitoring/reporting
• 2015: v5.1 – Hadoop integration, federated geo-replication, end-user credentials, hybrid cloud chargeback reporting and OpenStack Icehouse support

The latest v5.2 adds user self-service storage policies for quality of service (QoS), data protection and durability at the bucket level, plus, obviously, FL3000 support. The QoS involves monitoring, on a user and groups basis, the amount of storage bytes consumed, objects, requests/min, inbound and outbound bytes/min. If limits are reached then requests are held back until the next service window.

The software is available to run on commodity servers by the way.

Cloudian’s CMO, Paul Turner believes that S3 is going to be the way to access object storage. He says there are 50-60 applications using the Atmos API, very few using SWIFT and none using CDMI, compared with more than 4,000 S3-based apps.

He said: "5.2 software will enable a Cloudian admin to manage double the capacity of the previous HyperStore software version."

Overall Cloudian pitch

Cloudian’s marketing pitch says your data is “forever live”, hence the FL moniker. It centres on HyperStore now offering 100PB+ scale computing with no downtime, at a fraction of the cost of traditional storage arrays, and it being suitable for small, medium and large enterprises.

“Large” means large, as Japan’s NTT has 2.3 million users storing data on a single HyperStore system, and service provider Nifty has 3,500 businesses storing their data on a single HyperStore as well.

No downtime means the system can tolerate component, node and rack failures.

Here is a canned quote from Turner: “Cloudian HyperStore is the only fully S3-compliant, multi-tenant, multi-data centre solution that is ‘forever live.’ Enterprises require always-on access to business-critical data and the FL3000 series delivers on this. An enterprise can deploy once and manage its data, on its own terms, forever.”

“Forever” is rather a long time, Paul, but let’s enjoy the hype without carping.

All object storage is scale-out: it’s the nature of the technology. But it’s possible that Cloudian has the most scale-up nodes available. With its NFS and S3 interfaces, it can say it offers a better filer and/or a better Amazon S3 system on-premises than anyone else, and on which you can run Big Data analytics.

Turner says Scality doesn't even come close to scaling up to 384PB and Cleversafe definitely doesn't scale up to that level. The 7U starter system has 1 cent/GB/month pricing. Compared to Isilon, Cloudian costs 65 per cent less on a $/GB basis and needs 25 per cent less cooling and power.

Is Cloudian better than Scality?

Turner claims the following advantages:

• App support because we have S3
• Faster data access because we have a distributed Cassandra database look-up, rather than hopping around the nodes of a RING
• Our converged appliance model makes it simpler to buy and operate
• Multi-tenancy QoS to avoid any tenant starving out other users

He claims that Cloudian also has a QoS advantage over both Scality and Cleversafe.

Cleversafe's tech is not as good as Cloudian's for small objects. It can only do erasure coding, doesn't do flash caching, and is not a peer-to-peer system like Cloudian.

There is no quantitative performance data but, judging by the hardware spec, the FL3000 ought to have decent data access latencies.

Cloudian has got itself an army of powerful Samurai object warriors. Can it make progress in the market, competing against other suppliers such as Cleversafe, Scality, HGST’s Amplidata, EMC’s ECS, Caringo, NetApp’s StorageGRID, HP’s object offering and more? Will its arguably better hardware tech enable it to build a channel capable of competing against these players? ®

Bootnote

* These are drive-managed shingled disks, not HGST host-managed shingled drives.