Part One: Adaptive Computing and Grid Computing at SAP: Are they the same or different?
In order to delve into this topic we must first spend some time understanding what each of these is. I found myself getting the two confused this year and maybe you are wondering the same thing, probably because they are both fairly new topics in the SAP arena, at least compared to our overall product selection. So let’s dive a little deeper into each topic. This will be a three part series: In this section we will talk about adaptive computing. In sections two and three we will discuss grid computing and compare adaptive computing to grid computing noting the similarities and differences.
Adaptive Computing
Adaptive Computing is a capability released in SAP NetWeaver ’04. This feature is an optional implementation strategy that customers can use to help make their landscape more flexible, more efficient to operate, and less costly to maintain. Let’s start with a little history on why SAP released this new technology. As we rolled out more and more SAP solutions, the feedback from our customers was clearly that their IT Landscapes were becoming harder and harder to manage and they needed our help to find solutions to make it easier and more affordable. We therefore, developed adaptive computing to help solve these concerns. Adaptive computing can be used anywhere in the landscape, it can be used in development and QA, sandbox and training, or in production. SAP solutions such as mySAP Enterprise Resource Planning (ERP), mySAP Customer Relationship Management (CRM), or any component released on 4.6C or higher can run as solutions within our adaptive framework.
In order to create this adaptiveframework, customers must first work with their hardware and network partners to virtualize their IT Landscape. What does “virtualize” mean? Instead of dedicating a particular piece of hardware specifically to a permanent application, it is possible to organize the hardware, including CPU, Disk and Memory into pools that can be called upon to give load when and where it is needed. You can define one pool or several pools and these can be generically called into action. This virtualization occurs a little differently depending on which hardware partner is involved. Each hardware partner brings their own tools and environment to the table. You may hear terms for adaptive computing that our hardware partners use such as N1 for Sun, Adaptive Enterprise for HP and On Demand for IBM. Once the pool is organized it may physically represent 1 or 100 physical servers behind it, there is no limit to the size of the pool but it has one label.
So, once the customer has virtualized their landscape, is that all there is to it? Well, that is a great first step, but there is still more to do. Having the pools alone only creates the ability, now we have to add the “adaptive” link to the applications, and this is what SAP calls the Adaptive Computing Controller (ACC). This controller is an application that is written by SAP to manage the communication between the SAP applications and the virtualized IT landscape. This controller holds the definitions of the “pools” as well as the definitions of the “services” that need resources. The controller specifically refers to applications as services so that we can accurately represent definite pieces of the application that need to be assigned to a resource pool. For example, we can break an application down into application server services, central instance services, and database services. Each of these can be assigned to the same pool or different pools.
Now that we have virtualization coupled with the Adaptive Computing Controller written by SAP, what can we do to adapt? This is the fun part!
Once the services are defined to the controller and the pools are defined to the controller, it is possible to start, stop or relocate any service within minutes. These activities used to take days or weekends in a traditional environment. Now it can be done in just a few minutes. The activity can be triggered manually or by a third party tool.
So why would I want to relocate a service on the fly? Let’s give a few examples:
- Customer A runs both mySAP ERP and mySAP CRM. From 8am to 5pm the customer’s peak load is consistently in ERP but at 5pm, the mySAP ERP activity slows down. The mySAP CRM activity, on the other hand, picks up at 5pm because all the disconnected sales reps are uploading the orders they have collected throughout the day. Instead of having dedicated hardware for mySAP ERP for the daily peak, and dedicated hardware for mySAP CRM for the nightly peak, the customer can buy only enough hardware to cover the higher of the two peaks, and can use adaptive computing to determine which service is accessing what hardware at a particular time. Perhaps 5 application servers run in mySAP ERP during the day, 2 application servers in mySAP CRM. At 5pm, 2 mySAP ERP application server services can relocate to become mySAP CRM application server services. Now mySAP CRM has 4 application servers to handle the load.
- Customer B has a complicated landscape that includes several sandbox systems, several training systems, and several production copies. Instead of dedicating hardware to all of these environments, one pool of hardware could be used to “bring up” the environments that they need for a given timeframe. When they are not needed, they can be saved to disk and stored for the next time they are needed.
Hopefully you get the idea and can think of many other places that you can use adaptive computing.
What does it cost to put adaptive computing in place? There is no doubt that there is a cost to virtualize an IT environment, particularly for a customer that hasn’t started to use technologies such as storage area networks. But what we are seeing with the ramp-up customers is that there is a significant overall savings when you consider the entire process. Cost savings are realized not only from the cost savings on hardware, but from the cost savings that result from having a standard footprint and maintenance strategy. One customer is saying that they have an overall savings of 50%!
You might be wondering how your current technology complies with the adaptive framework. On the Service Marketplace you can find a quick link adaptive that provides more information about the subject. Included there you will find a link to the hardware partners who have passed the requirements of our compliance test. Hardware and network partners work with SAP to ensure that their tools are compliant with our Adaptive Computing Controller. The list is growing daily and you can see that there are already many options. I should also mention the Adaptive Computing Council which was organized by SAP and our technology partners. SAP and Partners meet on a regular basis to discuss requirements of the compliance test, latest technology trends and future plans for adaptive computing.
I hope that this has provided a good overview of adaptive computing and the advantages of its use. The Adaptive Computing Controller is delivered as part of SAP NetWeaver ’04. It is a piece of the technology platform that will get us to our long term goal of Enterprise Service Architecture.
In the next part of our three part series, I will give you more information on grid computing, and in the third and final part of the series I will compare/contrast grid computing to adaptive computing… See you then.
