Evaluating Storage-As-A-Service Options update from May 2009

Disk storage is already one of the data center's biggest expenses, the cost of finished arrays keeps rising, and IDC predicts that companies' storage infrastructures will grow 300% by the end of the decade. Amid this backdrop, there's growing concern that businesses will drown in the expense of storing data.

Enter cloud storage--network-accessible storage infrastructure that evangelists promise will prevent the apocalyptic scenario envisioned above by IDC in its "Exploding Data Universe" report (sponsored by EMC). Proffering largely untested claims about huge capacity cost reductions, the elimination of labor required for storage administration and maintenance, and just-in-time provisioning of capacity on a pennies-per-terabyte basis, cloud storage providers are getting the attention of businesses large and small.

To storage industry veterans, cloud storage doesn't feel like a brand-new model. The core concepts are derivative of service bureau computing paradigms advanced by IBM and others in the 1970s, of Sun Microsystems' "the network is the computer," and of application service providers and storage service providers of the 1990s.

What's different this time? The economy, of course. With on-premises storage costs already high and growing in many IT departments, interest in alternatives is strong. Storage-as-a-service vendors say they can lower costs by taking on the burden of storage management and insulating customers from related costs like hardware upgrades. If cloud providers can capture economies of scale by using the same pool of storage capacity to meet the needs of many customers and pass along the cost savings, the price of spindle capacity should be well below what companies would otherwise pay for their own storage.

The earlier-generation application and storage service providers made a similar argument, but they ran into problems that contributed to their undoing, including the networking costs of connecting users and applications, and customer concerns over sharing infrastructure with other companies. Once customers started demanding that their services be provided using physically separate components, ASPs and SSPs were forced to violate their own economies of scale by deploying infrastructure on a one-off basis for each customer.

Cloud storage vendors say they've gotten past those obstacles. Network connectivity is faster, cheaper, and more secure, and virtualization technology is being used in corporate data centers to provide segregated storage environments, minimizing concerns over data infection and demands for dedicated infrastructure.

Cloud storage folk point to the high cost of storing data that's retained for business or compliance reasons, yet rarely re-referenced by the organization. This data, they say, is ideal for cloud-based storage, and doing so will free up the spindles that companies now use for primary or live data. This argument has some merit and will likely become a mantra: the cloud as a low-cost repository for compliance and archive data.

DIG DEEPER

Private Clouds

Wary of the risks of public cloud services, some companies are deploying cloud-like environments internally.

Download this
InformationWeek Report

See all Informationweek Reports

Delivering such services, however, is again a two-edged sword for cloud providers. To address governance rules, including privacy laws, that apply in different geographies will probably require cloud storage companies to have multiple data centers with different operating procedures, audit standards, and certifications. This is no small issue and may quickly determine a pecking order in the universe of cloud storage services.

To better understand the cloud storage options available to businesses, InformationWeek issued a request for information and invited all storage-as-a-service vendors to participate. We did so under the auspices of a hypothetical midsize company called DIY Marketing Services, with nearly 100 TB of data in its storage infrastructure and volume growing quickly.

Five vendors--Caringo, IBM, Iron Mountain Digital, Nirvanix, and Zetta--participated, and our analysis of their proposed solutions is included in this report. The RFI can be downloaded in its entirety here.

Not All The Same
Cloud storage vendors come in different stripes. We divide them into four categories: pure-play companies focused on providing external storage services; generalists with broader cloud computing offerings of which storage is only part; specialty providers that offer a discrete set of services, such as remote storage for disaster recovery; and internal cloud vendors that sell platforms for cloud storage inside corporate data centers.

For all of the purported economic advantages of cloud storage and the improvements over earlier-generation services, some hurdles still remain. In April, McKinsey & Co. reported that, based on its analysis, cloud computing in general is more expensive than on-premises computing in some usage scenarios.

McKinsey's findings have merit, and they extend to cloud storage. The issues that make storage costly aren't directly addressed by the cloud model. From a cost of ownership perspective, the challenges of managing storage infrastructure are a key cost driver. Most storage array vendors differentiate their products with value-added software, embedding features on proprietary array controllers. A byproduct of that reality is that it makes storage management more difficult, especially when the infrastructure is heterogeneous.

Cloud providers may shield customers from the hassles of storage management, but the hassles remain. Most say that cloud storage management is a still-evolving story and one that they currently solve through the deployment of homogenous arrays that can be managed more simply using the selected vendor's tools.

This theme has helped define a second branch of the emerging market--technology for building internal storage clouds, offered by vendors such as ParaScale and Bycast. Internal cloud advocates assert that the security issues associated with storage across a wide area network haven't been mitigated as conclusively as external cloud service providers would like us to believe.

A Lack Of Standards
What's missing from the cloud story, whether external or internal, are standards. Most cloud storage providers present access to storage infrastructure via proprietary application programming interfaces.

Web services could be harnessed to facilitate the monitoring of resource status, including workload, performance metrics, availability, and capacity, so that policies could be crafted to allocate resources (whether physical, virtual, or cloud-based) to the right I/O. Such an open standards-based resource provisioning model has been embraced by Xiotech and a few other vendors and is on the drawing boards in many product development shops. However, no cloud services companies, including cloud storage startups, have thus far announced support for the strategy.

Asked about standards, most cloud storage providers reference standard certifications such as ISO or various security standards or the name-brand vendor whose storage gear is being used in their infrastructure. These may be standards of a sort, but they're not the kind of standards that promise to become as ubiquitous as TCP/IP, for example.

The absence of real standards for cloud computing generally, and cloud storage specifically, frames the key question that many prospective customers need to ask themselves before embracing the trend: How will I be any better off with clouds over the next few years than I would be from implementing a homegrown strategy of infrastructure right-sizing, improved infrastructure management, and intelligent data management?

The fact is that cloud storage appeals to many companies for the simple reason of cost savings. If vendors can deliver even a percentage of the cost reductions they promise, and if you can feel comfortable that your data will be secure and well managed, then the strategy could be a win-win.

Service-level agreements are very important and may determine the outcome of the cloud experience. An SLA is the vendor's promise regarding the availability and performance of a solution. There's a tendency to set the bar too high when entering into a new outsourcing arrangement, to hold vendors to service levels that would be impossible to deliver under the best of circumstances and regardless of price. Sometimes vendors sign off on impossible-to-meet SLAs just to get the business. When expectations are set too high, however, there's a tendency for the arrangement to sour within a year or two, as the record of traditional outsourcing shows. A clear understanding of both requirements and measurement metrics is important.

Plug Into The Cloud

or more on the storage-as-a-service trend, visit
plugintothecloud.com

Methods for signing up for a cloud service vary from vendor to vendor, as do contract lengths. While some vendors provide a stock SLA and contract, and a few permit sign-up on the Web, most want to meet with a potential client directly to assess the current situation and to arrive at a service offering that's tailored to its needs. In response to our mock request for information, several respondents preferred to go this route.Questions In Need Of Answers
Preliminary issues to resolve in moving to cloud storage are commonsensical. The questions that IT pros should ask as they move in this direction include:

No one, especially the vendor, much likes to consider the final question, but customer experience with the application and storage service providers of the late 1990s makes it essential to ask. The simple fact is that many vendors are appearing in the cloud computing market, which in the best of times probably can't sustain more than a handful of players. That means that some vendors ultimately will shutter operations, so IT departments that give cloud storage a try should have an exit strategy just in case.

In the InformationWeek request for information, our hypothetical company, DIY Marketing, was a composite of many companies we've interviewed that are looking into cloud storage. The structure of the RFI was intended to capture the current thinking and practices of vendors working in this area. Of the companies that responded to InformationWeek's request for information, three--Iron Mountain Digital, Nirvanix, and Zetta--are cloud storage pure plays, offering services for hosting and managing DIY's data. One, IBM, offers a set of discrete services for data protection. The other, Caringo, offers software for creating well-managed storage infrastructure internally, creating a so-called internal storage cloud.

Invited, but notably absent from the list of respondents to the request for information, were Amazon.com and EMC. Also not responding were champions of internal cloud technologies, ParaScale and Bycast. Most of these companies responded that they were unable to meet the time frames for providing a thorough response to our query owing to resource constraints.

In reviewing the RFI responses submitted, and from interviews with respondents to the RFI and other players in the market, my takeaways are mixed.

First, the positives. I was impressed that many of the vendors interviewed seem to have learned some important lessons from the application and storage service provider meltdown of the past. Gone is the willingness to build one-off infrastructure for every customer or to field name-brand gear when generic disk arrays will do just as well. This change in attitude might just make the difference between survival of a few of these providers, on the one hand, and a fairly rapid death knell for the entire cloud storage model, on the other.

I was also impressed by the statements of some vendors that cloud storage really is optimized for secondary storage, archive, and backups, where retrieval latencies (paralleling the World Wide Wait to which we all have become accustomed) play less of a role than they do in transaction-oriented primary storage apps. Not one vendor with whom I spoke seemed to be over-selling its capabilities.

Surprisingly, none of the vendors I chatted with was inclined to join cloud storage at the hip with x86 server virtualization. There were no claims that cloud storage would connect any more efficiently with VMware software than physical storage, or that cloud storage would improve the resource utilization inefficiency we already live with in both physical server-to-storage and virtual server-to-storage connection paradigms. That was a relief given the marketecture that inundates storage-savvy IT people from the pro-x86 virtualization crowd.

What Are Clouds, Anyway?
What I continued to find a bit disturbing was the lack of a refined definition of what exactly cloud storage (or cloud anything) is. Is it middleware? Is it simply a recasting of the old storage-on-demand services? Without a strong Web services standards hook, I fail to see what I can do with cloud storage that I can't do internally with storage virtualization software like DataCore's complemented with CA XOsoft for data replication across a WAN and Virtual Instruments for enhanced visibility and management of my Fibre Channel fabric.

I crossed swords with a few internal storage cloud vendors on this point. I would argue that true cloud storage platforms must be able to support multitenancy and SLAs, but not all vendors who adopt the cloud moniker do that.

Also, I was dismayed by an almost myopic focus either on cheap storage or on reduced risk among the vendors I interviewed and the RFI responses we received. The cheap storage play may be valid, but cost isn't the only component of business value; similarly, backing your mission-critical data to a service provider (or doing any replication at all for data protection) may lessen risk of a cataclysmic interruption of your business, but this alone isn't a comprehensive business case for technology.

Harvard Business Review has a penchant for triangles, using them to explain just about everything. In the case of business value, the triangle is bounded by cost containment, risk reduction, and top-line growth (an amalgam of improved user productivity, greater nimbleness, improved information support for decision making, and ability to adapt to changing markets). As many IT practitioners have discovered, you had better have a story to tell in each category or your initiative is unlikely to get funding. The same applies to constructing a compelling business value case for cloud storage: It must tell a persuasive story in all three categories, or DIY is probably not going to buy into it. Our interest in understanding the core business value case was incorporated into the RFI, but only one respondent addressed it.

Until a business value case is fully articulated, funding of cloud storage initiatives will be hard to come by. As the old saw goes, "No bucks, no Buck Rogers."

Jon Toigo is CEO of storage consultancy Toigo Partners International, founder and chairman of the Data Management Institute, and author of 13 books.

Evaluating Storage-As-A-Service Options
Cloud vendors respond to our request for information on the feasibility of storage services as an alternative to on-premises hardware.

Zetta Addresses Business' Storage Needs
Startup promises three-year TCO advantage of three to four times that of on-premises storage.

IBM Emphasizes Data Protection Services
Proposal lacks depth and confuses archiving with backup.

Experience Makes The Difference At Nirvanix
Comprehensive storage pitch reflects management's time working for application service providers.

Iron Mountain Moves Into Cloud Services
Its Virtual File Store emphasizes security, but its proposal lacks details.

Caringo Brings New Level Of Granularity To Data Storage
CAStor offers the advantages of EMC's Centera without the huge price tag.