Big data is hot, NoSQL versus "NewSQL" is a debate that’s rapidly ramping up, and enterprises are trying to determine why they should bother investing in the "private cloud" with a constantly expanding supply of public cloud computing. These will all be topics that come into the spotlight this year in the Cloud Track at Interop 2011 in Las Vegas, May 8-12.
In addition to the main Cloud Track on May 10 and 11, there will be an Enterprise Summit cloud workshop that will focus on public clouds on May 8 and on private clouds on May 9. Stuart Charlton, head of infrastructure operations of the Canadian Pacific Railroad, will speak about his firm's private cloud at 10 a.m. May 9, and Christian Reilly, manager of global systems engineering at Bechtel, will speak at 11:45 a.m.
A Cloud Track panel on May 10 will feature a set of leading cloud service providers moderated by Randy Bias, co-founder and CTO of CloudScaling, a cloud computing consulting firm.
Bias is a sometimes controversial thinker and blogger on the subject of cloud. He created a stir in an April 13 blog headlined, "Salesforce Shows Us Enterprise Clouds Will Fail," which more than one reader mistook as a negative barb at Salesforce.
Instead, he was asserting that some enterprise clouds or "private" clouds were likely to fail because they don't implement Salesforce-like features. They avoid multi-tenancy and produce more siloed applications that don't connect easily to anything else. They miss the economies of scale and flexibility of commodity cloud computing, such as Amazon's EC2. It's a theme he's aired before.
In the past, Bias also has urged others to be outspoken on panels that he has chaired. His Interop panelists are: Zane Adam, general manager of Microsoft's Azure Cloud; Robert Collazo, senior systems engineer at Rackspace; Christopher Gesell, chief strategist--cloud services at Verizon Business; and Tom Mornini, CTO and co-founder of Engine Yard.
All four are major cloud service suppliers. Azure spokesmen in February reported that the Microsoft platform-as-a-service cloud had attracted 31,000 customers in its first year of operations. Rackspace is often named as the second-largest supplier of infrastructure as a service after Amazon's EC2 Web services unit. Verizon announced last year it was offering cloud infrastructure services through its worldwide data centers, then in January, acquired the experienced cloud supplier Terremark. Engine Yard is a Ruby application services specialist in San Francisco using cloud infrastructure in Amazon's EC2.
Alistair Croll, founder of Bitcurrent and chair of the Interop Cloud Track, will make opening remarks on "How To Think Cloud" at 10:15 a.m. on May 9. He will focus on the areas where cloud computing turns many key IT assumptions on their heads. For example, his session summary says, "Machines are free. Data doesn't always need to be right. We assume everything fails."
"Data doesn't always need to be right" refers to the possibility with NoSQL systems, often used in social networking settings, to give slightly different answers to the same query from two parties. The queries might be launched in a game setting or other environment where the system is gathering unstructured information from what may be hundreds of thousands of users at a time. Because the system emphasizes speed of data collection and speed of response over the precision, it sometimes delivers answers before the most recently collected data has been added to the database.
Cloud systems also frequently assume the underlying hardware may fail, and the software includes the ability to recover business logic and data and renew its operation. Croll said via email that he will elaborate more in his talk on how cloud systems are a departure from traditional IT.
Recent research from Croll's Bitcurrent, published under the Creative Commons license (from data collected by research firm Cedexis) sheds light on the experience of cloud users on nine cloud platforms. Its March 3 report, "Cloud Performance from the End User Perspective," concluded that the average response time to complete an HTTP request was 426.4 milliseconds. Cedexis collects data to help users of content delivery networks see how effectively their content is being delivered.
Matt Aslett, an analyst at the 451 Group, recently coined the term "NewSQL" to cover the SQL-based systems that are being built to be both high performance and scalable to handle extra-large data sets. The name is a play on the trendy NoSQL designation for non-SQL systems that scale out over a server cluster to rapidly handle big data sets. The SQL in NewSQL is the same as old SQL; it's the vendors behind the systems who are new.
Liran Zelkha, founder of ScaleBase, a NewSQL system, will address the issue of "Handling Big Data on Public Clouds" at 11:30 a.m. on May 10. His talk will cover using either NoSQL systems that scale out over many servers or NewSQL systems.
His firm offers a scalable database load-balancing system that implements transparent sharding on a database. It works with existing SQL systems such as MySQL or Amazon's Relational Database System. Sharding is the practice of partitioning or subdividing a database system into a set of independent units, distributed on separate servers, each using its own data.
Charles Babcock is an editor-at-large for InformationWeek.