Friday, December 21, 2012

HP, Clouds and Big Data Again


Along with not being a lawyer, I am not an IT professional, and I'm proud to make both statements.  Since my original post on the cloud and big data as it affects HP, I've done quite a bit more research into these issues.  If I can borrow a lyric from Judy Collins' "Both Sides Now," I would say, "I've looked at clouds from both sides now."

I stand by my first, somewhat intuitive conclusion, and that is that "The Cloud" is largely hype, something that IT gurus enjoy as they wrap straightforward concepts in a cocoon of mystery.  At the same time, I've come to a different conclusion about first generation cloud players, namely that many companies will have their hands in delivering cloud solutions.  While the industry giants like IBM, Oracle, Cisco, VM Ware, HP, Amazon and Cray will all have roles to play in different segments, innovative startups will add their secret sauces until the industry inevitably consolidates, once the first generation cloud implementations are mature.

Although "cloud" is the current buzzword, the underlying approach and technologies have been developed over years and decades.  With the explosion of corporate data centers has come a massive sprawl of physical servers, which has flattened out at some 32 million servers, according to a 2011 white paper prepared by IDC, sponsored by HP.

As virtualization of servers really started to take off in the 2005 period, today the installed base of logical servers stands at more than 80 million units, according to IDC.  CTOs benefited from falling unit prices of servers, as well as from the flattening out of demand in units, as virtualization of machines accelerated.  Energy costs of data center management have been relatively constant, helped by more efficient servers, virtualization, stable utility rates, and energy conservation measures within the data centers.

The one cost component that has rampaged out of control is "Management and Administration," according to the IDC report.  The industry wide spend is some $50 billion.  As corporations saw their IT infrastructures start to sprawl, they became choked by an explosion of virtual machine images, the consequent overprovisioning of storage and data network facilities, and even a physical thicket of cabling.

Within the management and administrative expenditures, CIOs/CTOs and data center managers have had to focus on very micro issues like tracking down the causes of individual CPU failures within jerry-rigged racks of servers.  For this problem, HP has developed HP Operations Orchestration software that allows managers to design work flows and automate the processes for monitoring system performance, availability, down times, and the ability to quickly document and identify root causes.

According to the IDC report,
"Many large organizations have serious sprawl and incompatibility issues created by years of meeting their immediate needs by using a project-by-project approach." 
 So, HP and others looking to establish dominant positions in converged infrastructure management and cloud computing have developed offerings that sometimes combine optimized servers, storage and network appliances, along with management software and consulting.

The more a reader looks through the myriad offerings from HP, the more obvious it is that the knowledge, experience, and technical as well as business acumen of the consultants will be the differentiating point for the customer when choosing between, say HP and IBM.

The move to a fully automated converged IT infrastructure will reduce annual IT costs to provide a unit of workload throughput by several orders of magnitude, according to the IDC report.  That's why Meg Whitman talked about the unprecedented opportunity in the "cloud."

Remember, though that some parts of cloud computing, such as public clouds are, and likely will continue to be dominated by Amazon and perhaps other players like it.  This is a commodity business.

Large corporations may choose to develop "private clouds," or hybrid models where some lower-value data are stored on public clouds while mission-critical data remain in private clouds.  HP certainly has the ingredients and a positive legacy of relationships with the big customers to suggest that it can succeed.

It will, however, be competing with IBM, Accenture, Oracle, VMware and others.  It will take time and expense to upgrade and expand the consulting and customer interface talent pools.

A part of the offering for the converged infrastructure solution should be the ability to handle and extract value from "big data," which was one of the main reasons for acquiring Autonomy.  Unfortunately, only about $2.4 billion of the $11.2 billion of assets from the Autonomy purchase remain on the books after the last write-down.

Thinking back to the Analyst Day demo of Autonomy's IDOL engine, it really was a relatively primitive application, in which real time data from various social media channels were monitored, and counts were made by keywords.  The most frequently occurring keywords then scrolled across a dashboard in different sizes and colors of type.  This might be an application that a marketing officer might use, but it hardly seems mission critical.  If HP wants to play in big data, it remains to be seen if the Autonomy acquisition will be enough to provide the offerings HP needs for its customers.

Another area of hype is in the concept of "big data" itself.  The handling of extremely large, complex data streams made up of words, numbers, and symbols, along with real analytical engines beyond counting, is not something that can be done by many players in the business today.  These gigantic data sets are typified by university or private genetic research operations, or global meteorological networks, or private energy development and production companies. These data sets cannot efficiently be moved routinely back and forth among cloud servers.  Cray, Inc.  is very active with these kind of customers, and they lead with customized data appliances and analytical tools.

HP seems to be trying to extract whatever value it can out of Autonomy, but it may need to partner, or perhaps acquire, smaller, more specialized players who develop unique capabilities in  the field of big data.  The whole notion of HP acquiring a company should induce afib in hearts of their shareholders.  Also, aside from a selfish need for liquidity, it's hard to believe an innovative entrepreneur would want to see her company absorbed within a dysfunctional HP.

Cray, by contrast, seems to be effectively implementing a partnership model for many of its offerings.

2013 should be a "proof of concept" year for HP in the whole area of "cloud computing."





No comments: