By now most people in the SAP Technology space should at least be a little bit familiar with SAP’s Accelerated Application Deployment offering (AccAD for short). If not, I highly recommend you take the time to familiarize yourself with the product at SAP’s page here.
My aim in writing this blog is to add a little more “meat” to the argument that AccAD can be an excellent investment in time and effort for an organisation seeking to improve SAP’s performance over a Wide Area Network (WAN). The benefit is particularly evident when connecting to locations which are bandwidth constrained or have high latency.
Without further ado, I’d like to explore some results which we recently obtained in a global proof of concept (POC) which led to a subsequent Go Live for a client organisation.
Chapter One – The Problem
A single SAP Portal instance running in a central data centre is set to support users across the world, from Europe to Asia, Africa to the Americas.
- During initial Performance Testing a distinct relationship between Network Latency and Portal Response time is discovered.
- It seems that all those pesky HTTP round-trips conducted by the SAP Portal are having a terrible effect on the responsiveness of user functionality!
- Some log-on times measured from remote locations are taking minutes to complete, and this is unacceptable performance for the business.
The graph below shows the average response time to log-on to the Portal and BW Java systems as measured over a 3 hour test. This has been mapped against network latency at different global locations to establish the correlation between poor response time and network latency. The problem is starkly highlighted for the eastern European location, where not only is the average WAN latency greater than 350ms but the network is notoriously unreliable.
Chapter Two – The ACCAD Solution
Presented with the problem above, whereby the Portal and the BW Java reporting systems are going to perform slowly at remote locations, SAP recommended we have a look at AccAD and discover the benefits for ourselves.
At this point it is probably important to mention that we had already tuned our servers, code and applications so that they were performing acceptably in the data centre itself. This ensured that we could attribute most (if not all) of our performance problems to the systems between the data centre and remote locations, namely – the WAN.
The answer was to commence a Proof of Concept (POC) and test the AccAD solution “in the wild”. I’ve included a small diagram which simplistically outlines the AccAD architecture below:
Chapter Three – The Results
Following a number of false starts, set-up issues and tests, we finally had our results. So what did we find out about AccAD? Largely it was good news.
We discovered the following:
- The benefit of AccAD improved our portal and BW java responsiveness anything from 10% to 55% depending on network conditions, caching and latency. The greatest benefit was seen at locations with the highest latency and poorest network stability.
- Content that was not cached previously in Internet Explorer was improved by a minimum of 25% in all locations.
- Bandwidth usage to all locations was reduced by around 90%.
Most interestingly to me however, was the effect that ACCAD had on our Eastern European location and other locations with over 300ms network latency. The graph below shows the affect AccAD had on response times from the Portal over a 3 hour test at one remote location:
Essentially, what this meant for our users was the following:
- Improved average transaction response times by around 40%,
- Reduced maximum response times by around 55%.
- Reduced the spread of response times by over 65%, meaning that our users had a more consistent and reliable experience from the portal.
As you can see from the above testing, there is a strong argument to investigate the usage of AccAD for any organisation utilizing SAP web applications across a WAN. AccAD could well be that 40% improvement in responsiveness you’ve been seeking.
Consider using AccAD not just for response time improvements, but also to increase the consistency and reliability of your SAP applications in your remote offices.
Although there are obvious implementation costs (such as the need for new hardware at remote sites), it may be that AccAD could solve some performance problems that your organisation has been facing due to a slow or unreliable network.
I’d be interested to hear from other organisations that have considered or are considering using AccAD; your thoughts on the pros and cons and results that you’ve experienced.