Storage Peer Incite: Notes from Wikibon’s November 20, 2007 Research Meeting
This week Wikibon presents The EMC vision thing. Recently EMC has been criticized for its seeming lack of overall vision for the future of the storage industry. At its mid-November analyst event in Boston, it substantially refuted that charge by presenting a vision for the future backed up in many cases by specific plans.
One of the most interesting things EMC talked about was its estimate that more than 70% of data today is created by increasingly mobile individuals. Much of this data only exists on a single device, and a hard drive or device failure can result in the loss of what sometimes has considerable value. EMC proposes to extend storage management and data backup to these users via what might be called "storage-as-a-service." It will be an online system offering companies and individuals the ability to move data off their individual systems where it can be better managed, more accessible to business and personal partners and, most importantly more secure. Of course EMC is not the first with this idea; several data backup services are available online today for individuals to store their personal data. But this is the first time a large storage vendor has discussed entering this market, using the Software-as-a-Service (SaaS) model that has become increasingly prominant in recent months. It also promises to bridge the gap between personal, individual and company data, by providing corporate level data backup and management services to remote workers who are often highly mobile, and whose systems are often difficult for corporate IT to manage adequately because they are almost never inside the corporate firewalls and are not available for backup and other management activities at predictable times and logical locations. G Berton Latamore
During the week of 12 Nov. EMC invited the analyst community into its cloister to discuss its perspectives on direction, goals and tactics for trying to bring significant new innovation to the way users deploy and manage information infrastructure. EMC suggests this is a $60B opportunity vendors that will have significant benefits for customers. It expects these benefits to take four forms:
- Improved technologies for storing data, including better price points and higher availability;
- New ways to leverage information and the resources necessary to sustain that information, including better resource management and virtualization technologies;
- New approaches to optimizing investments in information infrastructure including advances in intelligent information management and enterprise content management; and,
- New types of services and capabilities for better protecting information at the archive security level.
Overall EMC's story came across as quite credible and suggests that the company has taken concerns regarding its overall vision to heart. While much work remains in defining what information infrastructure is, what it means and how users will adopt it, nonetheless a few intermediate term issues, where users should expect tension in information management activities, were raised by the discussions. The three most obvious are consumerization, energy efficiency and virtualization. The last two are relatively well understood, and it is interesting that EMC pushed consumerization as hard as it did.
Specifically, EMC says 75% of data is created by individuals and today stored locally, but over time will be managed centrally. EMC plans to bring services to individuals and enterprises supporting this reality with centralized Storage-as-a-Service solutions over the Internet for backing up and protecting that data. Whether EMC truly delivers on this promise in the context of the broader Software-as-a-Service (SaaS) environment remains to be seen, but it is interesting that the buzz around SaaS, which ebbs and flows every few years, has found its way into the storage domain with EMC as a major proponent.
In energy efficiency EMC promised spin down on all arrays eventually. But it did not provide any time frames or information on the kinds of algorithms it would use to ensure minimal impact on application performance. Given that spinning up a disk can take 20 seconds, applications will need to be aware of any spin down services on some level.
In the virtualization area, EMC is pushing both overall virtualization through VMware and file virtualization, based on the theory that SAN virtualization is better handled in the network. It introduced a new term – virtual provisioning – which is its spin on what thin provisioning ultimately will look like, and it expects to make this available in its products in 1Q08.
While each of these initiatives will involve significant change in how storage management and administration is envisioned, in the next few years we expect a greater rather than lesser dynamism in the marketplace as large numbers of sources of storage innovation appear, making the real test in innovation how companies like EMC sustain complex adoption by their users.
With that emphasis, we suggest four near term negotiation (4Q07-1Q08) approaches:
- Virtual (thin) provisioning will have significant bearing in capacity requirement in 1Q08. Users should include that in 4Q07 negotiations to get pricing that reflects this reality.
- Treat negotiations as a project to ensure that they cover the wide array of storage issues including application, capacity, and performance, so the right arrangement of capacity can be acquired.
- Continue to explore other sources of innovation. While we are encouraged by EMC's vision, that picture will be filled in over the next few years, and often in response to invention generated by startups and other sources in the marketplace.
- Explore new approaches to better leveraging overall software costs through tactics like in-box tiering where appropriate.
Action item: EMC is demonstrating an important commitment to storage invention by spending $1.5B a year on storage related R&D and $2B a year on storage related acquisitions. Customers can also be encouraged that EMC's leadership at a sales and service level will provide increasing facility for streamlining the adoption of the inventions generated by those investments. However, we remain convinced that storage will be a multi-source game for the next few years and consequently all sources should be explored, none left out, and customers should recognize that for the foreseeable future the responsibility for turning storage investment into business capabilities remains with users themselves.
In evaluating EMC's portfolio one of the key components of the strategy is 'optimizing' the information infrastructure. By 'optimize', EMC probably means allocating information infrastructure (resources and capital) as efficiently as possible. While EMC's vision is good and the domains of focus seem reasonable and leading, without clearer metering and accounting of costs (both capex and opex) it will remain unclear to what degree EMC infrastructure facilitates or impedes business capability.
Specifically, while customers can subjectively assess EMC's infrastructure value proposition, optimizing that infrastructure without clear cost transparency to the business remains illusory. While EMC is not alone in lacking these capabilities, and the ROI to EMC may not be compelling (because chargebacks are broken), the fact remains that without measurements at the business level, optimization cannot truly occur.
Action item: Metering and cost transparency are fundamental to initiatives like software as a service (SaaS) and Google-like storage architectures. Customers should observe EMC's capabilities and the development of its disciplines around storage accounting as a barometer of maturity in these emerging areas. To the extent these capabilities are integrated across solution offerings it will add real value; to the extent they remain disparate, customers, not EMC will bear the costs of optimizing infrastructure.
As storage growth continues, CFO’s increasingly point to the capital costs of storage as a problem. Three responses are possible. First, suppress future storage capacity requirements; given application and technology trends, this is not a workable option in almost any organization. Second, rapidly adopt storage-related technologies that better leverage storage resources: this is a fundamental goal of storage virtualization, but will put major suppliers of storage that rely on GB growth on the defensive (e.g., EMC). Third, attempt to substitute operating costs for capital costs by applying new “storage as a service” approaches to satisfying capacity requirements: this transfers unproven “software as a service” (SaaS) delivery models to the storage industry. EMC is the first major storage supplier to tout STaaS (storage as a service) for mainstream user accounts. In some domains (e.g., backup/restore of mobile user data) STaaS is a viable -- even attractive -- option: EMC claims that 55-60% of enterprise data is individually created, centrally storage, an excellent target for newly-acquired mozy.com. However, STaaS faces significant market development hurdles in other storage domains. Fundamentally, the basic problems of SaaS (no real cost advantages, real service limitations, customization constraints) haven’t changed, and are likely to be even more pronounced in STaaS.
Action item: STaaS (storage as a service) should not be considered as an alternative to other tactics for gaining control over capital expenditures related to storage.
At the EMC analyst day, the company signaled very clearly the order in which it will deliver advanced functionality to its storage arrays would be file level virtualization, thin provisioning and then block-level virtualization. It's executives talked very little about block-level virtualization. Wikibon members have indicated that the current top storage project priorities are:
- Storage support for server virtualization projects (Block & File),
- Block-based storage virtualization,
- Block-based thin provisioning,
- File-based storage consolidation.
On the same day, EMC showed a strong commitment to providing the services and file-based virtualization for VMware. If an organization's servers are being virtualized on VMware and are supported by file-based storage, EMC should be strongly considered as a partner for this project and also for SAN VMware projects where virtualization is not a requirement.
For file-based storage consolidation, EMC has good range of products including Rainfinity as a file-based virtualization offering. EMC and EMC services should be on any RFP list.
For EMC customers with significant investments in EMC products, one strategy is to leverage the future benefits of thin provisioning into the Q4 negotiations, and ensure a commitment to early delivery. A small project on 3PAR or Hitachi should also be considered so as to jump-start the planning and application testing required.
EMC customers looking to establish a heterogeneous virtualized storage environment to tackle reducing the costs of commissioning and decommissioning arrays and optimizing storage software management costs, can consider Invista, but there are almost certainly stronger offerings from IBM, Hitachi and others.
Action item: Users should create a list of storage projects with the potential payback and effort required for each project. Where EMC have well established technology and services, it should be included as a potential strategic partner in any RFP. For other projects, the choice of partner should be driven by user priorities and the willingness of the partner to help drive storage costs down.
The traditional approach to content management relies on the metadata from objects to rationalize, connect and organize the historical information of an organization. Some files (e.g., email archives) are rich in metadata, but most (e.g., Excel spreadsheets) are poor and require extensive and expensive classification to be successfully integrated into an overall corporate archive. This approach necessitates auto-classification by as yet unproven technologies.
An alternative approach is to use a consumer-based Google-like software infrastructure to ingest the data and search on just the content. Search technologies exist from Google, Microsoft, Yahoo and others. Infrastructure technologies that will exploit large-scale commodity storage and servers have been developed by Google, Amazon and others. For example, the Google File System (GFS) assumes that failures will occur in hardware, software and locations and manages security and data integrity across a distributed network without requiring backups.
EMC rightly concludes that its traditional storage and storage software will be an order of magnitude too expensive to compete for this alternative approach. At the EMC analyst conference, Joe Tucci pre-announced Hulk & Maui, based on commodity storage and servers. The market for this type of system will be service providers (such as Mozy, just bought by EMC) as well as large organizations that want to put in their own storage service infrastructures.
Action item: While EMC should be applauded for conceiving Hulk and Maui, users should not expect to see either the product or services based on this product to be mature any time soon. Any radically different approach will need significant time to test the integrity and reliability of the architecture and create operationally tested software.
EMC claims to be the only major storage company to build its own products 'from A to Z' (i.e. not relying on R&D/product development partners like LSI or OEM partners like NetApp or Hitachi). The benefit of this approach to customers is a rigorous availability and data protection discipline that includes 5 9's availability and strong data protection software on all product lines. Despite this fundamental approach, EMC's broad portfolio of hardware and software generally relies on different management interfaces and processes. A notable example is the DMX and CX product lines and their respective management suites, ControlCenter and Navisphere.
EMC's virtualization strategy is another example of how the company has acquired in sets of point solutions that largely remain separate. VMware, Rainfinity, Invista and other products in the portfolio are distinct approaches to solving different problems. What EMC is doing is a fine job of making sure these products work, and user indications are consistent that with EMC, they almost always do work and when they don't EMC acts expeditiously to fix problems.
Security with RSA and storage-as-a-service are two initiatives worth watching. With RSA, EMC is driving security capabilities across all products to include risk assessment, access control, ensuring securCustomers should not expect EMC to provide true 'out of the box' storage infrastructure across the board anytime soon. Rather customers should expect to build integrated storage capabilities from multiple piece parts. In the near-to-mid-term, EMC customers will continue to bear the costs of maintaining excellent skills in storage planning, design, integration and administration. e infrastructure, protecting data and assuring compliance. While no specific timeframe has been articulated, this is one area where EMC could potentially provide a capability that is truly integrated and auditable across the product set. To the degree that EMC can make information security more effective, easier to prove and affordable --with automated compliance-- it will drive good ROI for customers.
Storage-as-a-service is a new area and to the degree EMC applies a SaaS methodology to existing product lines it's hard to see how integration will be any better than it is today, although it could perhaps be less cumbersome to manage.
Action item: Customers should not expect EMC to provide true 'out of the box' storage infrastructure across the board anytime soon. Rather customers should expect to build integrated storage capabilities from multiple piece parts. In the near-to-mid-term, EMC customers will continue to bear the costs of maintaining excellent skills in storage planning, design, integration and administration.