Son of Monsta
Updating the blog managed to rise to the top of my To-Do list as I take advantage of the window in time that emerged today between my outbound email and voicemail to colleagues traveling home from SapphireNow Madrid and TechEd and the unlikelyhood of seeing too many responses before the weekend.
When such a window opens, (never as frequent as I would prefer), I especially run to it when it coincides with my having more news to share with you about projects in the SAP Co-Innovation Lab here in Palo Alto. We’ve got two or three spooling up in COIL now, giving me chance to bring the outside world up to speed about the projects and team expectations. For today I will focus one of the three and cover the others in subsequent posts.
Son of Monsta- My last blog outing provided a follow up on the long running “Monsta” Project. If you are still unaware of or for whatever reason missed this one, you can read the earlier blogs and hit links to all of the relevant content. I mention it briefly here, only because I now want to talk about ,… “Son of Monsta”!
One of my favorite aspects of the SAP Co-Innovation Lab is that so many of our successful products produce spillover effects, like the formation of exciting and important follow on projects, which is exactly what this is.
There have een other spillover effects from the Monsta project, like the core content of the Monsta whitepaper being applied to SAP Sybase ASE and SAP IQ marketing collateral and witnessing an increase in queries about the project from the field as account managers continue to discover customers interested in large scale BI 4 deployments. We now even have partners participating in the project generating content directed at their own channels, which are of course viewed by many with a high interest in business intelligence and data analytics. These spillover effects have a way of adding value that is almost never predicted or expected when a new COIL project is proposed, so this is something to keep in mind as a very desirable outcome produced by co-innovation projects.
So in Son of Monsta, we are building off of the former Monsta deployment (Supermicro twin blades, a 4u Seperserver, Red Hat RHEL 6, and SAP Sybase ASE and we will continue using F5 for our load balancing) where we again intend to run up to 10,000 concurrent users against BI 4 but now with SAP HANA as the underlying reporting database versus the use of IQ in Monsta. One additional change is that we are also deploying certified SAP HANA hardware from IBM.
The testing is still very straightforward. We will adjust our previously used test scripts given that our data set is now larger and different than before and this time around we will be entering into a BOE environment as part of the test. For each user logged onto the system, there is an open and refresh of Web Intelligence reports pulling 2,500, 5,000, 10,000 and 25,000 rows, open. Next there is simulated navigation to an Explorer work space. Then, log off. These actions repeat until the end of the test.
The team intends o gather performance metrics and statistics from the SAP HANA and BI4 environment to understand their behaviors under a 10,000 concurrent user load. The team will further look to analyze the performance observed and then apply and document possible best practices and optimization techniques.
We’ve also articulated a secondary and important goal to determine overhead of SSL encryption over the JDBC connection in a high-volume/high-load environment. While SAP HANA JDBC connections can be ncrypted with SSL Connections without SSL encryption communicate in clear text, ecurity policies will likely drive a need for encrypting the client and server data connections to SAP HANA. While recognizing the value and critical usefulness of SSL as an effective and efficient encryption method built for speed; it can come with serious overhead. Historically, smaller deployments were not necessarily prone to serious performance degradation given the overhead but as Jay Thoden Van Velzen suggests, “once multiplied by 10,000 concurrent users this may have impact on performance, CPU load and memory consumption, as well as throughput”. If you want , you can join in an online discussion hosted by Jay
Once the project team reaches its goal of 10,000 concurrent users without SSL encryption, we will apply SSL on the JDBC connection and re-run the test. In general, we are nearly set to execute as soon as we complete preparations with Soasta to establish our test harness and finalize our BOE configuration. Hopefully we will see first results right after theThanksgiving break.
Motivation, Interest and Need-
What is driving the interest in this project? To begin, we continue to witness a centralization trend in BI systems and for the need to support larger volumes of users comprised of an increasingly diverse audience. Usage of analytics is fast becoming both deeper and wider where SAP HANA will only serve to accelerate this trend. The data scientists, data analysts as traditional users are still out there but now the consumers of business intelligence data also include executives, product developers, and account managers, as well as financial personnel, marketing managers and even customer service representatives. While certainly anecdotal, at both SapphireNow and TechEd in instances where COIL presented its project results, we had customers at each event telling us about their efforts to deploy BI 4 to even more than 10,000 concurrent users. The increase per-user use of business analytics results in a real need for the results we can glean from projects designed to optimize these large scale deployments and continue to embue customer confidence.
If this project interests you, we will of course continue to provide updates. If you are aware of other projects out there similar to this that might benefit from learning more about this particular effort, please share! As always, comments welcome!