A RisingTide lifts all clouds

by Robin Harris on Thursday, 30 July, 2009

Check out their homepage and, as of today, “This page is not here yet….” That recursive mindbender is an apt introduction. Who – and what – is RisingTide Systems? StorageMojo spoke to founder Marc Fleischman to find out.

The StorageMojo take-away: a semi-stealth mode open source storage software company focused on virtual DAS for Internet-scale virtual machine infrastructures. Let’s break that into bite-size pieces.

  • Semi-stealth. Marc spoke to a blogger without an NDA. On the other hand they are planning to get their web site up in September.
  • Virtual DAS. Virtual direct-attached storage for virtual machines and/or physical machines.
  • Internet-scale. VDAS for 100,000 physical or virtual servers – or more. They can scale down, but in the cloud space this is the current sweet spot.

Block vs file
Block service offers 2 key advantages over scale-out file systems: they’re bootable, enabling seamless virtual/physical migration; and, RisingTide supports >10k LUNs per target, i.e. one LUN per user, thus enabling management of storage attributes (QoS, IOPS, capacity, availability, etc.) on a per user basis.

RisingTide’s stack is Linux-based. It includes a comprehensive iSCSI target, an initiator, a generic target engine – they aren’t wedded to iSCSI – a very high speed (8 Gbits/sec on 10GigE) fabric module, snapshot engine and an HA option. The iSCSI initiator allows Rising Tide to deploy storage features such as end-to-end QoS for LUNs and custom protocol-level extensions.

A key feature is a highly scriptable management interface that enables ISPs to to build – or more likely, integrate – their own management console. Cloud computing is based on the idea that capital is cheap and labor dear.

A scalable management infrastructure is key to realizing the advantage. GUIs don’t scale.

ISPs, whose thin margins encourage OSS, using RHEL and KVM (kernel-based virtual machine), as well Xen VMs are RisingTide’s target. One prospect, who currently hosts 500 PB, is all physical DAS but wants to decouple storage and CPU. They think they can do it with an iSCSI infrastructure.

They also support FCoE, but aren’t seeing much interest in that. Given the economics of ISPs that is not a surprise.

The StorageMojo take
Cloud computing is all about scale, which is why the private cloud concept is suspect. Few enterprises can justify the scale to required to make the leap from labor-intensive to capital-intensive computing. And really, what enterprise vendor wants to help them do so?

This is a good thing, unless you are counting on selling private clouds. The tech is fascinating, but once the numbers are run, the concept will be done.

Which leaves the field open to companies that live and breathe massive scale: big ISPs; Google; Amazon; IBM Global Services; and the firms supplying kit to these folks.

Massive-scale infrastructure isn’t just more of the same. It is a brave new world with its own dynamics that are already reshaping the IT industry.

Courteous comments welcome, of course. Update: A few clarifications made about 2 hours after the first post. End update.

{ 4 comments… read them below or add one }

Wes Felter July 31, 2009 at 12:30 pm

Virtual DAS? Isn’t that called SAN?

I wonder why they chose Linux over OpenSolaris.

Dan Brown August 5, 2009 at 8:12 am

I really wonder if the private cloud is about massive scale or ubiquity of access and storage at a lower cost. If I’m at a reasonable 100TBs and have three regional offices (East, Central, West), a private cloud might make sense. Simple, direct access across each of my offices. Replication of data (typically file-based) from each to the primary data center. Access to all files from each location regardless of where created. And, an infrastructure that’s easy to manage, runs on a cluster of commodity servers, is self-healing, resilient, scales and protects my data. I don’t have PBs to manage. I don’t have thousands or millions of users. However, I do see a benefit to the private cloud for this company.

Jeff August 5, 2009 at 5:00 pm

Would be nice if the linked worked to their hompage

Sajai Krishnan, CEO, ParaScale August 5, 2009 at 6:54 pm

Nice Robin – good chuckle from the first coupla paragraphs of your article. And of course only courteous comments. Just because you have not bought into ParaScale’s brand of koolaid does not make you a bad person :-).

So you say “cloud computing is all about scale, which is why the private cloud concept is suspect.”

Scepticism is good. And Dan Brown makes my point about private clouds very well in the “micro” sense. Thanks Dan.

Allow me to take a “macro” view. With the explosion in unstructured data, the real question is whether a Google or Amazon type commodity, scale-out files-system architecture can be deployed in the enterprise. To us at ParaScale this is an cloud architecture that is deployed internally and provisionable on-demand just like a public cloud. Call it commodity-based, scale-out tier 2 NAS, with on-demand provisioning – and perhaps this is better messaging for the enterprise.

Relying on file storage methods of the last decade are not economical or scaleable going forward, especially for the 90% of file data that is rarely touched and mostly inactive. We say more about his here http://blog.parascale.com/?p=170.

I know you are an economist at heart, and so you will appreciate this. Sure public clouds provide massive economies of scale. But 2 things. One, unlike compute which can show highly variable use patters, storage has a moving average that trends upwards. So in a highly variable compute scenario you can get more benefits from a public cloud that are less available in a storage situation. Two, by using commodity servers in your internal cloud (including repurposing old servers) you really get direct benefit from Moore’s law, more so than any other part of an enterprise’s typical storage architecture. With architecture’s like ParaScale’s, a CIO can really say that he buys his storage at Fry’s or equivalent.

Sajai, CEO, ParaScale

Leave a Comment

Previous post:

Next post: