Frequently Asked Questions
Following are some frequently asked questions about our products along with their answers.
If your questions are not answered here please contact
us and we shall respond to your request as soon as possible.
Is grid computing right for me?
Do you have more than one computer on your network? If yes, grid computing is relevant for you
because tasks that can be run in parallel will take fraction of the time compared to dedicated
Why can't I just write an SMP application with processor-binding?
Applications written for SMP (symmetric multi-processing systems) attempt to utilize multiple
CPUs simultaneously on a single node. Grid computing using our
products takes the next step of
not only utilizing multiple CPUs on a single node but also CPUs on separate
nodes. Therefore, your application is able to scale as your business grows.
How is this Platform different from other similar solutions?
Currently there are a couple of major vendors that advertise grid computing. The Oracle 10g
is a database management system that advertises itself as grid based database. The 10g database
architecture is actually quite limited because its underlying technology is RAC, which relies on
database storage to be accessible by all nodes directly. Therefore, Oracle itself limits the
maximum number of servers that can be part of a grid to be 100.
Sun Microsystems also has a product called N1 Grid that allows distributed computing but it only
tackles the CPU usage problem and still relies on an external database such as Oracle for data
Unlike above mentioned software, Velokri Platform provides a truly scalable secure computing
engine with built-in data management. Data management and or storage in above mentioned systems
happen at a central location (requiring special hardware in some cases) whereas in our platform data
itself is distributed across nodes. Using the concept of domains and application level routing our
platform can scale in multiple dimensions forming unique topologies to suit your business.
What is the difference between topology and simple pools some other vendors provide?
Pools are traditionally a set of nodes in a network. A set is
simply a collection without any structure built into it. A topology
uses such a set and gives it structure using some sort of a metric. In our platform, the application
level topology is
defined by channels of communication between nodes. Routing of requests happen through the
application itself and follow the specific topology defined.
Just as in simple point-set topology multiple topologies may be defined on a
set, the same applies to a set of nodes using communication channels.
Therefore, using a handful of nodes in our platform you can define application level
topologies of various types and hence giving your grid different structures.
All of the above terms are simply a means of defining communication channels between resources
(i.e., nodes) in a grid at the application level irrespective of how IP routes are defined in
a given network, unlike unstructured traditional pools.
Do you have a data sheet for your product?
Yes, you may download it here
What types of rules can you configure the dispatching server to follow?
The dispatching servers (i.e., the server that takes your set of tasks and dispatches them to
worker processes on separate nodes) can be configured the following ways:
- Fast In this mode the dispatcher
picks the fastest available OPS (Operations Server)
with the correct capability to dispatch a given atomic task/sub-task.
- All When configured in this mode the dispatcher sends
the task to all OPSes that have this capability in the grid. This configuration may
be used if you want to mirror rows for a database/table or if you want to execute a
particular task on all participating nodes.
- Round-robin In this mode the dispatcher cycles through
all OPSes in the grid that are capable of executing the given task one by one .
What operating systems are currently supported?
Our products are currently supported for Linux 2.6 (x86 and amd64) and Solaris 10 (sparc). Other
platform ports such as Windows, HP-UX, AIX etc may be availabe depending on customer needs.
Is it fast? What programming language is this written in?
The framework software itself is written in the fastest portable language (i.e., C) available. All
system calls are made using the POSIX framework and therefore our software may easily be ported to
other platforms. We believe in literate programming
and therefore use the CWEB
dialect of C.
Other software components that are provided as tools or external API are written in Perl, Java and
Copyright 2006-2019, Velokri, Inc.