February 28, 2008   Sign In |  About ebizQ |  Contact Us |  Join ebizQ Gold Club
Gian Trotta
First Look
Join ebizQ producers Gian Trotta and Krissi Danielson for interviews with the innovators, movers and shakers behind emerging enterprise software solutions.Have a solution that qualifies? E-mail Gian at gtrotta(at)ebizq.net

« Skyway Brings J2EE, AJAX and Adobe Dreamweaver to SOAs | Main | Cassatt: BEA Co-Founder Brings Virtualization's Advantages to SOAs »

April 30, 2007
IBM Expert: Maximize Your CPUs With Data Grid Architecture
Listen to the entire 8:42 podcast Download file


Resources and Agenda

Read a full text transcript

Get a Free Trial of ObjectGrid of WebSphere Extended Deployment

Read Billy Newport's /dev/websphere blog

Agenda:

1. The case for data grids
     a. workloads up; performance down

2. Solutions
     a. make the data co-resident with the application
     b. scale out the application and grid technology

3. How parallel processing works

4. Middleware’s role
     a. "Do all the hard stuff so customers can just work on their business logic"
      b. provide fault tolerance
      c. handling the data
      d. avoid manual setup of memory databases on multiple machines

5. Middleware's pluses:

     a. smater middleware configures grids on runtime doesn’t require more admins and lowers TCO.

6. Case studies:

       a. Telcos: lots of service records and tight customer response times
       b. Financials algorithmic trading and decimalized data have architecturs bursting at the seams

7. Data grid solutions:
       a. lets you add more boxes to grid rather than buy bigger boxes

      b. speeds apps and avoids redesigns

Have a question? Billy Newport will respond to comments posted below.

Billy Newport will regularly respond to any comments posted below.

Billy Newport, IBM's senior technical staff member and chief architect for WebSphere XD ObjectGrid, feels that conventional applications built in multiple tiers, such as data tiers and application tiers, have served the IT industry well.

However, the growth of data-intensive have both raised the implementation challenges and heightened the application benefits of data grid architecture.

“In some kinds of environments with the kind of work loads, and the way they have been increasing -- those application architectures have kind of been showing their age,” Newport said.

One solution includes transitioning to data grid-type architectures where “instead of pulling the application data from the database, the data is co-resident with the application,” Newport observed. 

Other Ways To Increased Performance

Data grid based systems can offer huge performance increases because it isn’t necessary to pull a hundred gigs of data to the client for  processing, then send the results back. “The data grid approach would have the same address base as that hundred gigs of data, but in a grid,” Newport explained. “That would allow multiple computers (hundreds or even thousands) to process smaller amounts simultaneously at memory speed. So response time is much improved.”

In the Middleware

Usually the role of middleware is to do all the hard stuff so customers can just work on their business logic, and middleware is a very important part of any data grade architecture.

“The Middleware handles things like fault tolerance, data replication, where the data gets placed on the grid, and other technical details,” Newport said.

“We can see how data grid-type applications will be able to process terabytes of data on thousands of machines in parallel. That gives the kind of real-time response times and real-time scaling that customers are demanding today,” Newport added.

Challenges Still Exist

Customers are getting tired of dealing with middleware that requires them to tell it how to do everything. They are saying that they want a much smarter middleware that can configure the data grid at runtime. And this middleware needs to figure how it's going to place the data on that grid, and how it's going to repair faults.

Large-Scale Applications Can Benefit Most

As an example, Newport points to the telco space where the requirements include extremely high availability and millisecond response times. “So when a customer wants to send a text message, it’s necessary to find out if they have text messaging enabled on their plan and to do that extremely quickly, with very high availability. And a data grid-type architecture to store what services are enabled for what customer is much more scalable a big box.”

If grids get larger, like 10,000 servers, that kind of architecture is absolutely essential because a leading requirement from customers is to lower the TCO. And, adds Newport, “They want to be able to have the same number of administrators with 1,000 servers, as with 2,000, 5,000 or 10,000. They need to keep the cost down. And that means you've got to automate as much as possible, to make a data grid economical to deploy.”

Industry-Specific Approaches

The telecom industry is one space with mixed architectures. It is now moving to voice-over IP. The voice sessions work well in a data grid architecture, where you route the calls to where the data is and we can maintain back-up copies for fail-over. The systems can be scaled up by adding more and more boxes. “Another important aspect for Telco is service provisioning. Where you want to know what given services a customer has when he makes a call. And you want to be able to look up the services very quickly. And you want to add and remove those services very quickly,” Newport explained.

Conventional databases don't really scale when you start to have a large wireless network of 10, 50, 100 million customers all making calls. What is needed is a kind of in-memory data base architecture that scales simply by adding more boxes. “So that all that information can be stored and deliver the kind of response times at the millisecond level that people demand when they're making voice calls,” says Newport.

The idea here is that as the telco company grows it can just add more boxes to the grid. The data grid middleware, like ObjectGrid from IBM will scale out the data across the new boxes. “So as you add more boxes, you're adding more CPUs, more network, and more memory. So that it's just going to become faster and faster and faster more servers are added,” Newport pointed out.

He concluded by noting that more information is available on the IBM extended deployment website. A trial version of ObjectGrid is available there for customers to download and to try out these features.

“The product addresses a lot of the problems in the middleware, and we're starting to see a lot of customers pick it up now, and move from the conventional architectures to these new data grid-type architectures,” Newport said.

Posted by gtrotta in |Digg This|Add to del.icio.us

Trackback Pings

TrackBack URL for this entry:
http://www.ebizq.net/mt/mt-tb.cgi/1740

Comments Post a comment




Remember Me?

(you may use HTML tags for style)

We ask that you type your code (displayed below) in the text box.This code is an image that cannot be read by a machine. It prevents automated programs from submitting comments.


Code:



Most Recent ebizQ Blog Entries
ADVERTISEMENT
Subscribe
Gian Trotta 's Articles
This Work
Accountability:The opinions expressed in this blog are solely representative of the blog's author, and not of ebizQ

Blogosphere

Marketing Solutions | Feedback | About ebizQ | Unsubscribe | Privacy Policy | Site Map