|Next Page||Prev Page||Main Index||Home|
Normally technical pieces are decidedly understated, and hence,
defensible. In the case of MPbase, even this understatement
can look like a set of outrageous marketing claims. Because of
this, I will now lay out for the reader a quick set of the features
and functions without any attempt to understate them into a more
As wild as the following sounds, it is all provable. Everything
stated here is a FACT. Everything stated here has seen at least
one working example. So, here is Next Paradigm Systems'
presentation of MPbase's fact sheet, believe
it or not!
Initial load & index creation
MPbase has no separate mass-load function. All rows are
loaded into MPbase using the update function. The update
function is the most efficient possible load method. So no mass-load
function is required. Because mass-data-loads use the update function,
index creation is an automatic and integral part of the process.
Because of the nature of the load process, it is NOT physically
possible to see an inconsistent image of the database.
Because within MPbase it is not physically possible to
see an inconstant image of the data, backup/restore processing
can safely be done in segments at any time. Using the roll-forward
and roll-backward logs, such a piecemeal backup can be restored
as a fixed-point snapshot of any time between the starting and
ending point of the backup.
When MPbase is running on a cluster with mirroring enabled,
reorganizational unload/reloads can be done without taking the
database down. User queries and even updates can continue throughout
this process. The new copy of the database can be verified and
then can be accepted or rejected without the users ever knowing
of its existence. Once the new copy is accepted, MPbase
can switch to it without any disruption to the user queries.
Compression + Encryption = Speed
The manner in which MPbase stores the data is highly compressed
and encrypted. This is done to increase the data access speed.
A byproduct of this compression is the saving of disk space. Typical
compression rates for MPbase are on the order of 90% or
better for non-image ASCII data. Image data rates are only about
50% for the lossless compression of single-frame scanned images.
Loosely coupled parallel clustering
With MPbase the database image is decoupled from the hardware.
This allows an MPbase cluster to be fault tolerant at the
machine level. This means an MPbase cluster has NO single
point of failure. Any machine in the cluster may be shutdown at
any time without effecting the function of queries or updates.
This includes inflight transactions as well as any new transactions.
Big Binary Blobs
In MPbase, the concept of big binary blobs has no meaning. MPbase can be set up to handle any type of data blocks in an intelligent manner. Queries of image data can be made at the pixel level. This allows questions to be asked of the internal content of such an image. Questions such as, Are there any discontinuities at a level below the screen resolution? In other words, On this x-ray are there any breaks not visible on the screen?
Fully Predictable Scaling
With MPbase, the black magic of scaling goes away. If you
have MPbase in production and you suddenly find you need
100 times the capacity, no big deal. You can place one hardware
order with 100% confidence in the final performance numbers you
will get. If there is a surprise it will be on the high side.
You will get more than you thought you would.
Isolate Interesting Data
One of the aspects of "Multidimensional Data-Intelligent
Run Length Encoding" (MDIRLE) is its inherent ability to
isolate "interesting data." As MDIRLE uses a content
addressable memory (CAM) schema, data that does not conform is
isolated. This data is inevitably one of two types, data in error,
or interesting data. In either case, it is data you need to know
Quick System Development
What is IT? It is Information Technology. What is
a database? A technology that handles information.
If a database was powerful enough it would be the IT department.
MPbase is everything you need except the user front-end.
A new system equals a new front-end. Everything else can be handled
in MPbase. As the information-handling tasks can now reside
where they belong, the application becomes a very simple front-end
The Theory Of Relativity
Everything is relative to the viewpoint of the observer. The MPbase
view can support any possible view of the information in an MPbase.
This means a system requiring a relational view can "see"
a relational MPbase. A system needing an indexed file system
can "see" an indexed file system MPbase. A system
needing a multidimensional cube can "see" a multidimensional
cube MPbase. A system needing a hierarchical database can
"see" a hierarchical MPbase. All of the systems
can be using the same MPbase at the same time. The view
has nothing to do with the physical storage formats used in MPbase.
It is all relative.
|Next Page||Prev Page||Main Index||Home||© 1998-2004 NPSI|