But in the last few months my blog production has declined significantly.  Part of that was due to my changed role to a consumer of information instead of a producer.  And part of that decline originated with a change of of heart in terms of what I want to do.  Because of that change of heart I last week resigned from EMC.  I will be spending many months pursuing some personal interests.

At this point I have nothing more to contribute to this blog.  I may start professionally writing again after my sabbatical.  But I am just as likely to write on a new blog as here.  But I will leave this content online indefinitely should any of the articles prove useful.

Thank you to VMware for letting me contribute a drop to the incredible wave of virtualization that washed over the industry.  Thank you to EMC for bringing me to Asia and teaching me business, both that of a vendor and the customers we work so hard for.  And thank you to each and every one of you for being a reader.

The IT World article caught a very interesting definition of availability Oracle used. Oracle pledges 99.5% availability but allows itself to remove some types of outages it its availability calculations. For instance, from Oracle’s document, the following events do not count as downtime with respect to their availability pledge:

“Unavailability of management, auxiliary or administration services, including administration tools, reporting services, utilities, or other services supporting core transaction processing”

I am not quite sure how to interpret this. It seems that any number of Oracle’s internal management tools can go offline and and have some undefined impact on customer’s applications. Presumably these failures would only interrupt management activities. In that case it makes sense not to count the time as unavailable. However, the business impact of these management problems on the customer is not clear.

“Outages occurring as a result of any actions or omissions taken by Oracle at the request or direction of Customer”

This seems to be a big one. If the customer requests help from Oracle, any outages as a result of that request are not measured against their 99.5% availability pledge. I do not understand why they would not be.

“Outages due to system administration, commands, or file transfers performed by Customer users or representatives”

Surely Oracle cannot be held responsible for a customer bringing down his own applications. But is it possible for a file transfer to do so? I am not sure what that last phrase is implying.

Any time a service provider documents their SLA the nerdy side of me enjoys read it. The service provider market is evolving rapidly. And while some of Oracle’s statements may be vague I believe they have been more bold than Amazon in documenting their intentions. As these documents become more and more clear it will be easier for customers to choose among providers.

One of the fun things of being a vSpecialist is participating is the rich discussions we have internally on every part of our business.  We have a smart team with incredible knowledge of virtualization and cloud.  When any new concept gets raised in our internal discussion groups, it is refined in the crucible of debate.

This SDDC idea is exactly the type of thing we like to talk about.  Its big. Its transformative. It is cutting edge. We are internally asking ourselves and answering a lot of questions on the SDDC. For instance:

• What exactly does SDDC mean?
• How is it different from the EMC/VMware cloud vision?
• What is the difference between cloud and SDDC?

I wanted to add my voice on the SDDC story to EMC’s internal discussions.  I did this through a narrated presentation that covers the SDDC in nine minutes.  After dozens of emails and a half a dozen phone discussions, it looks like this message is solidifying.  I submit it here to you for your consideration.

First, two snapshots of consolidation ratios from 2011 and 2013 expectations.

There are more, highly-consolidated servers, obviously.  But there are also more sparsely-consolidated servers.  I suspect this to be the result VMware’s efforts to convince customers to virtualize large mission critical applications.  Big Oracle databases, for instance, benefit from the improved operational model enabled by virtualization.  But they cannot be consolidated like tier-2 applications. This means more single-VM servers running big applications.

Here is one more chart with interesting, albeit confusing, information.

The interesting part is the incredible increase in desktop virtualization in the past two years.  12% more of InformationWeek’s respondents will embark on VDI journeys in 2013 when compared to 2011.  Network virtualization should increase substantially, possibly reflecting the excitement around Nicira and VXLAN.  But the confusing thing is the decrease is respondents that say they will virtualize servers and storage.

I cannot understand how this is possible.  Except that maybe during response some people thought they could only select one option.  That might have shifted results from more established virtualization techniques (server, storage) to newer ones (desktop, network).

I have only shared a fraction of the interesting stuff in this document.  It is over 30 pages long and contains 28 figures.  I highly recommend you check it out.

Before I get to that conclusion, let me share Dan’s graph on the changing price of infrastructure.

Compute and capacity are getting cheap more quickly than transport.

Compute power and storage capacity have exploded in the recent decades.  Network bandwidth has increased, but not nearly as quickly.  These differing progressions mean that bandwidth optimization is elevating its priority in datacenter design.

If this trend continues, data locality, (as in, “are my applications as close as possible to the source of their data”), will dominate the next generation of multi-datacenter design.  Applications should first be placed near their data’s source.  The cost of infrastructure and data storage become less relevant.

Today customers flock to EC2 for a perceive decrease in cost of infrastructure.  (Personally I suspect off-premise application hosting is not as cheap as people believe.  Most cost assessments ignore data transfer rates, the cost of outages, the risk of data loss, etc.)  But as compute and storage costs continue to drop faster than bandwidth, people will no longer say “off-premise is cheaper”.  They will say, “closer to the information source is cheaper”.

This means service providers can make money by selling (and reselling and reselling) access to common data.  They can further differentiate with unique data management and manipulation services such as analytics.  This is a much richer business proposition than the existing, Amazon-inspired “race to the bottom” of costs and margins.

By finding customers that was to use the same data, service providers will orient themselves to industry verticals.  For instance, a social media service provider would ingest all of Twitter.  It could then allow many customers to run analytics against the entire Twitter feed.  Customers are paying for the analytics and reduced WAN access.  Because service providers are amortizing their WAN costs across all their customers they can charge a lower price for access to the Twitter than an individual customer could get.  With access to this common data, SPs could provide unique, high-value, high-margin services that customers might not be able to build themselves.

Conceptually, this changes the hybrid cloud map.  In the old map, three companies analyzing Twitter trends would independently pull feeds.  They would process and store the data internally and farm out some small portion of their infrastructure to their hybrid cloud service provider.

In a model where bandwidth costs dominate, a service provider would pull the Twitter feed.  The provider’s customers would use in-house services for analysis and return the high value results to the enterprise.

And this model would work across multiple industry verticals, not just social media.  There are common components to oil and gas, finance, retail, and many others verticals.

Bandwidth usage is considerably less in the second model.  The cloud provider has unlimited options for industry-specific service creation.  And with that differentiation, service providers can grow a business with high value to customers and large returns to their investors.

The world is waiting for service providers to create these industry vertical “malls”.  When that happens I think we will see the real promise of the hybrid cloud ecosystem.

The three figures below were generated by EMC data scientists using data from a couple of sources. In one case Twitter provided EMC access to the Twitter “firehose” for nine days: we collected every tweet in the world. We stored in a Greenplum database one billion tweets. Tableau Software contributed their expertise to turn the data into these interactive charts that you can explore.

First, an interactive graphic correlating Major League Baseball’s teams’ wins versus tweets. Boundless enthusiasm for the Yankees buoy their twitter count no matter how the team fares. An incredible season by the Nationals generates less interest than a lackluster season by the Red Sox. Also check the statistical outliers for the Astros and the Cardinals. Drill down into those results and see events and mistakes that skewed the data.

The next Tableau visual shows Singaporeans tweeting about traffic. Singaporean taxi driver complaints about traffic are endless. But I think Singapore traffic is amazing. You can only complain about local traffic if you have never seen Bangkok, Beijing, or Los Angeles.

The third graphic shows results of radiation measurements since last year’s Fukushima disaster. Concerned about fallout, armies of concerned citizens performed measurements at massive scale. You can see from this chart the red zones near the accident and the safe areas around the country.

All fifteen of our interactive graphics are available online, courtesy of Tableau Software. The visitors to our Singapore event got a kick out of the graphics and I think you will, too. If you are an aspiring data scientist you should consider visualizing your hobbyist data with Tableau. It is completely free for personal use!

These are two cases of big data in action: the collection and analysis of vast quantities of information. These topics are addressed in Rick Smolan’s video summary of The Human Face of Big Data (HFOBD) project, which is ongoing this year.  Earlier this week his company, Against All Odds Productions, released a smartphone app to collect data from 10,000,000 people (hopefully). With that large anonymous contribution of “data exhaust”, this project may learn some interesting things.  I downloaded the app on Friday to participate.

The app was minimally functional for me on Friday.  It installed and I was able to answer its questions.  But the Facebook login did not work. I elected to allow it to track my movements throughout the day.  Unfortunately the location services on my Samsung Galaxy Note draw significant power.  My phone’s battery was dead by the late afternoon.  I could not figure out how to turn off the app’s use of location services so I uninstalled it that afternoon.

Today my girlfriend and I installed the app and went through the questions together over a beer.  With instantaneous feedback from other users, I saw a few trends worth sharing.

First, a brief demographic observation.  About 40% each of married and single people.

I once read that a generation ago only 10% of couples had a single child.  Apparently that percentage has doubled since then. As a product of that generation, I fall into that 10%.  The HOFBD app suggests most respondents are also children of my mother’s generation.

Life has been good to me, as it has to 83% of the app’s respondents. I would be fascinated to sit with a sample of the 17% that feel life has been unfair to them. Surely some people start in bad situations. Does 17% reflect this? Or is it biased by pessimists?

38% of respondents would rescue electronics from their burning homes. (Anyone could have predicted this among the Android and iPhone users running this app.) 10% would rescue money, which I keep in banks instead of under my mattress. And 2% of respondents would grab their gun. Should there be a fire in your neighborhood keep the guys with guns away from those carrying their life savings!

Recall that I took this survey twice, the second time sitting next to my girlfriend. I have to admit that the second time I answered differently the question, “what is the one thing you would change about your partner?” And I made a couple other minor modifications, too. That’s why I was surprised to see my data doppelgänger was me!

Despite its bugs, this app provided a fun diversion for a few minutes this afternoon. I hope you all will download it and participate in the project. For more on the project see Rick Smolan’s video below.

Rick Smolan – The Human Face of Big Data from Steven Dean on Vimeo.

For the first meeting Ben has asked VMware’s Iwan Rahabok and me to present on Disaster Recovery, Downtime Avoidance, and High Availability.  Iwan is a VMware SE in Singapore that I have worked with for many years now.  We are both looking forward to leading an interactive discussion with everyone that attends.

If you are in town this Wednesday night please come by the Singapore VMware office on Level 6 of Suntec City Tower Four at 17:00.  This will be a great opportunity to learn more about current VMware solutions and future directions.  It is also a great chance to network with other professionals in VMware community.  Let your friends and colleagues know this event is on.

I have for years been correcting this misunderstanding among customers. Despite Oracle’s clear statement of support we often see conflicting messages from Oracle’s sales teams. I was pleased to see a video from Oracle’s Richard Garstagen at VMworld 2012 on the unambiguous support of VMware environments:

Happily, Richard even mentions the effectiveness of host affinity in limiting the license costs. He also addresses the fear that reproduction on physical is a serious concern for virtualized Oracle databases.

I first read about this on License Consulting. Thanks for your article!

I believe that mixed DR/DA/HA environments are so complex that very few people in the world fully understand their complexities. Scott Lowe might be the only person I have worked with that appears to have a complete grasp of the situation. I know I do not. So, you might find it strange that I was called upon to lead this talk.

Well, I learned a lot in the talk.  Both from the research prior to its first delivery and from the hallway conversations that followed my sessions.  I documented, learned and created a few key concepts for this talk. Here they are.

Downtime avoidance and disaster avoidance are two different things. EMC uses VPLEX to deliver the key characteristic of both downtime and disaster avoidance: zero downtime migration over distance. The difference between them is defined by the customer’s expectation of a disaster’s radius and the distance between the datacenters. When two datacenters are far enough apart that they do not both fall within the radius of disaster, it is possible to implement a disaster avoidance solution between them. If they both fall within that radius, only downtime avoidance is possible.

Some customers enjoy VPLEX between two relatively close datacenters. They can then migrate workloads to avoid planned outages. But if those datacenters both fall within the same flood plane, they do not have a disaster avoidance solution. Despite this difference of function, there is no technical difference between disaster and downtime avoidance. So, they can both be called “DA”.

The combination of DR, DA, multi-site HA can produce unexpected effects.  Most of the weirdness comes from using old versions of vSphere.  Prior to vSphere 5 HA was controlled by five, arbitrarily placed primary nodes.  When they found themselves at the same site, that site’s failure would disrupt HA.  Prior to vSphere 4.1 there was no way to recommend virtual machine placement at a site using host affinity.  I tell customers these days that multi-site HA should not be attempted without vSphere 5.

Combining DR, DA, and HA provides different values and challenges. I argue that “DR only” is an “illegal combination”. Any customer using vSphere should have HA enabled.

HA is not DR.  This simple fact is not universally understood.  HA provides a simple, automated restart mechanism to account for relatively small failures.  DR requires documented, tested, potentially complex plans to recover an entire site’s loss.  I now tell customers that HA is a technology.  DR is a process.

Joint DR/DA solutions are available to EMC’s customers using a combination of VMware Site Recovery Manager, EMC VPLEX, and any number of layer two virtualization/flattening technologies.  EMC released with GeoSynchrony 5.1 earlier this year a RecoverPoint splitter to allow replication from a VPLEX deployment to a remote site.  With VMware SRM, this means that customers can enjoy DA and HA between the two VPLEX sites and DR to the third, protected site.

In truth SRM is not a prerequisite for a joint DR/DA solution.  In Singapore we have a customer that deployed VPLEX to achieve cross-island DA.  The customer uses “traditional” methods of scripting to simplify and automate the disaster recovery process.  For environments like this script creation is pretty straightforward.  But script maintenance can be costly and risky.  I prefer to see customers using SRM for their DR plans.

Stretching L2 networks across distant sites is far too complicated for me (and many IT professionals) to handle without help.  The design of complex networks remains an area of IT that I am woefully inadequate.  But in my experience many of the professionals I engage are not strong in this area, either.  If you are considering a DA or cross-site HA scenario, I recommend you buddy-up with a network expert.  And make sure you understand the demands your configuration will place on the cross-site link.  Often dedicated links between distance sites are costly and limited.  If you allow DRS or HA to place virtual machines at sites distant from their users or intranet ingress points, you may find an artificial and avoidable bottleneck caused by the limited link.