Home Smart 100 2011 DeepCloud MPP database (SMART 100)

DeepCloud MPP database (SMART 100)

0

Show your support for this innovation. Tweet it. Like it on Facebook. Leave a comment.

The following SMART 100 profile and the information it contains is a duplication of content submitted by the applicant during the entry process. As a function of entry, applicants were required to declare that all details are factually correct, do not infringe on another’s intellectual property and are not unlawful, threatening, defamatory, invasive of privacy, obscene, or otherwise objectionable. Some profiles have been edited for reasons of space and clarity. More about the SMART 100.

DeepCloud MPP database

This innovation initially came to life when…

After working with companies that needed to store massive and growing data collections on computers that could no longer match their growth in I/O capacity each year, we realised that eventually all these companies must resort to parallel processing of their large data collections.

They simply have no other direction to go in.

Our approach was to bring together the best available technologies to make the world’s fastest MPP system that permits simple expansion beyond the size of the largest systems in operation in the world today.

WHAT & HOW

The purpose of this innovation is to…

This system allows companies with large growing data collections to quickly analyse all of their data together and easily plan and calculate the cost of their future infrastructure requirements.

Unlike conventional systems, adding additional data and nodes makes all database queries run faster.

It does this by…

By implementing Ingres Vectorwise (the world’s fastest single node database) in parallel across the OpenMPI supercomputer, data transmission and propagation delays are significantly reduced. Technologies including remote DMA, OpenFabrics interconnects and virtualisation allow sustained growth, highly efficient communications and simple installation up to a theoretical maximum of 16,000 nodes, resulting in a blisteringly fast SQL engine that will process petabytes.

PURPOSE & BENEFITS

This innovation improves on what came before because…

Customers benefit from technical advances of commodity hardware rather than proprietary solutions of most of our competitors.

Additionally, the Vectorwise system gives us a 70x speed improvement over other solutions.

Furthermore, the use of the OpenMPI supercomputer adds extra facilities and more efficient performance. Combined with the OpenFabrics interconnecting mesh, we can scale further with reduced latency and greater throughput.

Its various benefits to the customer/end-user include…

The ability to linearly expand data warehousing power and capacity for a linear cost and to go beyond any current data warehouse limits.

Sequential DBMS fail about 30TB and require geometric increases in effort to expand at the high end.

COMPETITIVE LANDSCAPE

In the past, this problem was solved by…

The stagnation of computer I/O rates vs the exponential growth of machine-generated data is a relatively new phenomenon and has not been often solved other than by a handful of specialist companies in the last few years (see below). 

These companies have either used proprietary hardware or modified slow open-source DBMS to implement parallelism.

Its predecessors/competitors include…

Teradata sales of $1.9B in 2010
DataAllegro bought by Microsoft 2009 for $275M
Greenplum bought by EMC in mid 2010 for $450 M (est) Netezza bought by IBM in Dec 2010 for $1.7B
Vertica is currently being sold to HP

TARGET MARKET

It is made for…

Very large-scale analytics of massive volumes (petabytes) of event or relationship data via standard SQL based tools.

MPP systems are mostly used by Telcos for behaviour and fault analysis of Call Data Record (CDR) collections.

By banks for fraud detection.

By governments for crime/terrorism detection.

By stock exchanges for trade analysis.

By tax agencies for financial analysis.

Typically these systems sell for about $20K – $25K per TB of data with some sales being for systems with as much as four PetaBytes of data.

DISTRIBUTION STRATEGY

It is available for sale through…

The entry level edition is distributed via the internet, with upgrades and internet based support.

The larger installations will be sold and maintained via partnerships with consultancies in each country.

PaaS/SaaS systems will be available on a pay-per-use principle.

Our marketing strategy is to…

Achieve MySQL-like ubiquitousness via a low-cost entry-level product, establishing a vertical market to modularly upgrade to our premier product, the ultra-fast, Vectorwise-based Extreme Edition.

Additionally we will offer a pay-as-you-go backup/ETL/Data warehousing and Business Intelligence service, based in the cloud, which we expect will have an even higher potential over the longer term.

SHOW YOUR SUPPORT…

Voting for the Readers’ Choice Index has concluded, but you can still show your support!

  1. Tweet it: Top left of each page
  2. Trigger a Reaction: Facebook ‘Like’, etc.
  3. Leave a Comment: Anonymous comments excluded*

More about the READERS’ CHOICE INDEX.