password
username
Sponsored by CakeMail, an email marketing software.
Newsletter preview


IBM developerWorks Web services/XML tip newsletter
----------------------------------------

IBM's resource for developers
http://www.ibm.com/vrm/newsletter_10802_2998_65790_email_DYN_1IN/yjq136238161

----------------------------------------
17 March 2008, Vol. 8, Issue 3
----------------------------------------

"SOA SERVICES IN A GRID AND NETCENTRIC WORLD"
http://www.ibm.com/vrm/newsletter_10802_2998_65791_email_DYN_1IN/yjq136238161

Level: Intermediate

Judith Myerson (jmyerson@bellatlantic.net)
Systems engineer and architect


Hello, tip readers.

Get to know grid types, grid computing, and Global Information Grid
(GIG). This article focuses on issues related to harnessing unused
resources for computer power that's too intensive for a stand-alone
machine. Explore examples of solutions, such as monitoring change in
grid scale, grid coupling switch, and GIG and Service Oriented
Architecture (SOA) testing methodology.

For the full tip, read on below.

Until next time,
The Web services/XML tip team at developerWorks

Correspondence:
mailto:dwnews@us.ibm.com


_________________________________________________________________
INTRODUCTION

In the developerWorks series "Use SLAs in a Web services context," I
discuss securing multiple Web services with a service-level agreement
(SLA) guarantee. Another series, "Work with Web services in enterprise-
wide SOAs," covers consolidating your SOAs as a three-dimensional
integration hug to improve speed and reliability, defense in depth for
multiple SOAs, and speeding up Web services applications with the XML-
binary Optimized Packaging (XOP). In the same series, I explored load-
balancing Web services, cultural considerations of SOA adoption in the
federal sector, and tight coupling Web services in the SOA. (See
resources at the bottom of this page for other articles by the author.)

Each of these shows a trend toward the optimization of resources to
execute Web services in multiple SOAs. Transitioning SOA services to a
grid and netcentric style is a way of harnessing and sharing unused
resources from computers in the grid.

By moving Web services that connect the applications and systems to the
grid, you can extend the resource capacity of multiple computers in
collaboration with one another in parallel. This represents a paradigm
shift from static use of a stand-alone machine's resources at one
location to dynamic sharing of resources of multiple machines in
parallel at any location.

In this article, you look at what grid types are, what grid computing
is about, and what the aims of GIG are. Find out what's missing from
the concept and structure of grid computing, and get suggestions for
solving problems.


_________________________________________________________________
GRID TYPES (SERVICES)

Grid computing from a service perspective depends on what type of grid
you want to use: dedicated, nondedicated, or distributed. Let's break
these down and learn more about them.


Dedicated grids

A dedicated grid consists of dedicated hardware and computing resources
used for your grid. The dedicated grid provides the most control and
flexibility over your grid architecture, as you can select all aspects
of the operation. It's the most flexible form of grid computing, giving
you the freedom to choose the topology and networking hardware you want
to use and that are most appropriate for the situation.


Nondedicated grids

Nondedicated grids use the resources and environment of an existing
computing infrastructure. For example, a grid that uses the computing
resources of your company when the desktop or server computers would
normally be largely idle is a nondedicated grid. You have less control
over the environment and network structure, because you can't change
the core structure used to support the machines when they're not used
in a grid. You're more likely to rely on the existing network and
infrastructure and have little or no control over the networking
decisions.


Distributed grids

A distributed grid is made up of resources of machines that can be
located anywhere, internal or external, distributed over a WAN or the
Internet. You have virtually no control over the network structure, but
you do have the ability and control to ensure that the distributed
components can communicate with each other effectively. The management
concerns in this case are aimed more at providing access, security
(including firewalls and authentication), and backup solutions for
providing connectivity in the event of a failure.


_________________________________________________________________
GRID COMPUTING RECAP

IBM was an early advocate and contributor to commercial grid computing
to create a single-system image from the virtualization of distributed
computing and data resources, such as processing, network bandwidth,
and storage capacity. Simultaneously, the resources of many computers
in a network are applied to a single problem that requires a lot of
computer processing cycles or access to large amounts of data.

Grid computing is a way to solve problems that require an enormous
amount of computing power. You can think of it as distributed and
large-scale cluster computing and as a form of network-distributed
parallel processing. It can be confined to the network of computer
workstations within a corporation across geographical boundaries, or
it can be a public collaboration (for example, peer-to-peer computing).

The resources of thousands and thousands of computers can be
cooperatively managed through collaboration toward a common objective.
Because the load of resources can be balanced in the grid with unused
resources, grid computing is like an extreme form of load balancing.

Grid computing requires the use of software that can divide and farm
out pieces of a program as one large system image to several thousand
computers. One concern is that if one piece of the software on a
workstation fails, other pieces of the software on other workstations
may fail. This is only if the single piece doesn't have a failover
piece on another workstation and relies on other pieces of software to
accomplish one or more grid computing task. Another concern is low
latency that can result from inadequate utilization of unused resources
in workstations.


_________________________________________________________________
GLOBAL INFORMATION GRID

Grid computing fits with the U.S. Department of Defense's (DoD) GIG
vision given the heterogeneity of the DoD's systems. There are three
ways of using grid computing as they pertain to GIG:

* Computational grid: A grid focusing on computationally intensive
operations
* Data grid: A data computing system that deals with data -- the
controlled sharing and management of large amounts of distributed
data
* Equipment grid: Where the surrounding grid is used to control the
equipment remotely and to analyze the data produced

The U.S. Department of Defense defines the GIG Enterprise Services under
the data grid type. This represents a paradigm shift from a system-
netcentric network to a data-centric network.


Real-time decisions

GIG evolved in response to the need for an environment in which users
can access, share, collect, process, store, disseminate, and manage
information on demand from any location in the grid.

The GIG aims to achieve information superiority in a network-centric
environment by enabling various systems and messaging-based Web
services to interoperate with each other in parallel in solving
problems too intensive for any stand-alone machines. GIG users can post
and retrieve information and make real-time decisions rather than
relying on historical information from multiple automated information
systems applications.


Low latency

Intensive solutions require very high throughput with low latency, like
that offered by IBM WebSphere MQ Low Latency Messaging. These solutions
address the explosion of data volumes across financial markets in high-
velocity trading and analytic environments.

Designed for one-to-many multicast messaging, low-latency messaging
software can deliver approximately 1 million 120-byte messages per
second on Ethernet, close to 3 million 120-byte messages per second on
InfiniBand, and more than 8 million smaller messages per second, all on
common x86 servers. Testing has also measured very low latency of 30
microseconds for 120-byte messages delivered at 10,000 messages per
second on InfiniBand or 61 microseconds on Ethernet.


_________________________________________________________________
WHAT'S MISSING FROM THE GRID

GIG lends itself to an SOA on a grid carrying information on demand.
This means grid computing now relies on an open set of standards and
protocols, including key SOA standards for Web services.

When these Web services are moved to the grid, these standards aren't
adequate in resolving resource and performance problems at the grid
level. We need something more encompassing than WS-Resource Transfer in
different areas of the grid environment; in terms of what's required,
we need to think about using it as a method for storing and recovering
general information about grid-to-grid monitoring and management as well
as security.

The problem is that Web services, normally loosely coupled, run whether
the resource is scarce or not. We need to find ways to ensure that the
resources on multiple workstations aren't wasted when they're in the
grid. To find them, consider what's missing from the grid, then offer
some solutions.


Monitoring change in grid scale

Volumes of resources can change from low to high and vice versa in the
background in a nongrid environment. The resource is either scarce or
isn't scarce while Web services are waiting to send or receive a
message. If the change in scale isn't adequately controlled at thousands
of workstations, it can have an impact on the one system image in the
grid, resulting in resource overloads.

One solution is to develop a grid monitor of how the unused resources of
each workstation are harnessed and shared by other workstations. If the
system finds that the unused resources on any workstation aren't
properly harnessed, it should send an alert to the grid and system
administrators so they can look up details in the logs for resolution.


Grid switch coupling

My article about tight coupling Web services in an SOA (see resources below)
discusses considering a Web service with a coupling switch mechanism at the
workstation level. This switch would flip to tight coupling from loose
coupling when the Web service receives an alert that its corresponding
resources have reached certain levels. When the Web service makes the
switch, certain WS standards must be switched (for example, WS-Context
for loose coupling to WS-Addressing for tight coupling).

At the grid level we can go further than that. There should be a Web
service at the grid level to send an alert to specified workstations to
switch from loose coupling to tight coupling of some Web services when
the resources at the grid level reach certain levels. If this is
reversed, there should be a Web service on a specified workstation that
can send an alert to the grid when the resources of other Web services
switched to tight coupling in the same machine have reached certain
levels.


_________________________________________________________________
GIG AND SOA TESTING METHODOLOGY

To ensure the capabilities of GIG and SOA support the needs of the
intended users, comprehensive testing is needed. The complexity of GIG
and SOA enterprise services requires that methodology become more in
depth and thorough. At the same time, to keep up with the rapid change
and short development life cycles expected from the system builders,
tests have to be ready to conduct in timescales, ranging from machine-
specific functional to grid enterprise.


_________________________________________________________________
CONCLUSION

You need a team of developers, testers, and system and grid
administrators to enable SOA services to be grid and netcentric. To
make the transition, you must plan ahead for the requirements and
responsibilities of developing, migrating, testing, and deploying SOA
services at the grid level. Resolving these issues makes your job of
transitioning SOA services to the grid a lot easier. You can use IBM
Rational ClearQuest, IBM Rational Tester for SOA Quality, IBM Rational
Functional Tester, and WebSphere MQ Low Latency to increase productivity
by reducing testing and defect tracking time at the grid level.


=================================================================
LINKS TO OTHER GOOD STUFF

::: Resources related to this tip :::
nla_internal_2661676.jpg Full text of this tip on the Web :::
http://www.ibm.com/vrm/newsletter_10802_2998_65791_email_DYN_1IN/yjq136238161

::: Read Judith Myerson's article "Tight-coupling Web services in the SOA" :::
http://www.ibm.com/vrm/newsletter_10802_2998_65791_email_DYN_11IN/yjq136238161

::: IBM developerWorks SOA and Web services zone :::
http://www.ibm.com/vrm/newsletter_10802_2998_65791_email_DYN_5IN/yjq136238161

::: IBM developerWorks XML zone :::
http://www.ibm.com/vrm/newsletter_10802_2998_65791_email_DYN_6IN/yjq136238161

::: Subscribe to the developerWorks weekly newsletter :::
http://www.ibm.com/vrm/newsletter_10802_2998_65791_email_DYN_7IN/yjq136238161

::: IBM's privacy policy :::
http://www.ibm.com/vrm/newsletter_10802_2998_65791_email_DYN_8IN/yjq136238161

::: IBM's copyright and trademark information :::
http://www.ibm.com/vrm/newsletter_10802_2998_65791_email_DYN_9IN/yjq136238161


=================================================================

THIS NEWSLETTER IS FOR INFORMATION ONLY. This newsletter should not be
interpreted to be a commitment on the part of IBM, and, after the
publication date, IBM cannot guarantee the accuracy of any information
presented. You may copy and distribute this newsletter, as long as:

1. All text is copied without modification and all pages are included.
2. All copies contain IBM's copyright notice and any other notices
provided therein.
3. This document is not distributed for profit.

IBM developerWorks
http://www.ibm.com/vrm/newsletter_10802_2998_65791_email_DYN_10IN/yjq136238161

----------------------------------------
About this newsletter

Manage your subscriptions:
http://www.ibm.com/vrm/newsletter_10802_2998_43352_email_DYN_1IN/yjq136238161

Subscribe:
http://www.ibm.com/vrm/newsletter_10802_2998_43352_email_DYN_4IN/yjq136238161

***:
http://www.ibm.com/vrm/newsletter_10802_2998_43352_email_DYN_3IN/yjq136238161

Contact editor:
mailto:dwnews@us.ibm.com

To ensure proper delivery please add enewsletter@us.ibm.com to your address book.

You received this e-mail because you are subscribed to IBM's developerWorks Web services/XML newsletter as: TAYLLORCRISS@GMAIL.COM

© International Business Machines Corporation 2007. All rights reserved.

IBM Corporation
Attn: Developer Communications, M/D 241
150 Kettletown Road
Southbury, CT USA 06488

Contact IBM:
http://www.ibm.com/vrm/newsletter_10802_2998_43352_email_DYN_2IN/yjq136238161