SAP HANA – we did it in 4 days, and lived to tell the tale


SAP Teched is fast approaching, and I have an ASUG session on HANA where I am presenting.  Due to a variety of reasons – I could not get my presentation and demo completed earlier. And then the idea dawned on me – why not crowd source this within the larger organization ? Having worked on a number of global projects in IBM – there was not a doubt in my mind that I could  find a team of super smart people from across the globe who will work with me to accomplish the goal.  We also enlisted the help of SAP, which they gladly provided.

Goal was simple – Stand up a whole infrastructure from scratch – hardware and software, and do a meaningful scenario of loading data using data services, putting it in HANA, and getting it to show up in multiple BI 4.0 reporting tools. We explicitly did not want to take any short cuts – no programming workarounds – enhancements or mods, no use of existing stable systems in the landscape, etc. And if we could not complete it in less than a week – we decided to stop. And for the information of my friends in enterprise collaboration – we made utmost use of lotus suite. Lotus connections, sametime, lotus notes and good ol’ telephone.

We did not want any “managers” in the team – we only wanted hands-on techies, irrespective of the official titles.  As it turned out – we had a core team comprised of folks from India, Europe and USA.  Plus we had an extended team in IBM and SAP who we could call on if we needed help. Some of the core team guys were on vacation, some had to do fasting through the day and only eat at night – and those who were working, all had one or more active projects that needed their attention.  Essentially – this had to be done on the side. This team had skills in basis, Linux, BI, ETL, SQL and some java.  I think on an average – we all had more than 10 years experience in SAP and/or BO  technologies.

The first big challenge was getting the Linux OS loaded on the box. It took almost a day to figure it out, but then finished quickly.  We could not have done this without the active help of  the IBM HANA/BWA  expert  Henning Sackewitz – thanks guys. HANA installation was next – and it was relatively smooth. We did the Rev 13 of HANA, and we could finish this in a few hours.  Although my basis colleagues are routinely used to seeing .SAR files and SAPCAR utility – I was a bit annoyed using them. Eventually I got the right version installed on my laptop – and studio/client got installed very quickly. However, not every one had the same luck. Few others had to fight the studio install before they were up and running.  Bottom line – my question for SAP – Why do I need to do command line options to install a windows application in 2011?  Is it that hard to package a visual installer that I can unzip and install?

Onwards to data services. Seriously a huge pain in the neck to get this one up and running. The post installation steps needed to get it working are not explained in a straight forward manner. We spent a lot of time hunting for the right .jar files and .dll files for the front end. Even as I type this, designer does not work on my PC. Thankfully – it did work for few others, so we could move ahead.

Data services integration to HANA was also quite painful. If the source is ECC- it is quite smooth. But once we  used some flat file sources – it gave us a tonne of false starts. Just setting up the connections with HANA took us a while. So my next question to SAP – why can’t I connect two of your systems in a more straightforward way?  For example : why would I need to read through your documentation to understand that the only port that works  is 30015? if you have a consistent logic to derive this anyway according to the documentation, why can’t the software figure this out without me wasting time?  SAP should consider spending some serious design thinking time on making sure the integration between data services, hana and BI 4.0 is seamless as advertised.

BI 4.0 was a known beast since we had experience playing with it before – and hence the additional grief was kind of minimal.  The connections to HANA are not exactly straightforward – but not too difficult either. We could expose universes and Explorer to HANA relatively quickly.

HANA studio is eclipse based, and hence had a familiar feel to it. However, when SAP talks about design thinking – this would not be what they are referring to as a top example. Context menu is not always very useful – and getting help on some features are just not possible. Quick launch is a nice touch – except there are some links which do not give useful responses. For example : Configure import server – you put in the data, hit an ok and then you wait. Guess what – data is saved, and you can see it if you go in a second time. But system won’t tell u what happened after you hit the OK button.  Another example: If you right click a table – would you not expect to see an option to see the data flow on how it is getting populated? or even an option to create data provisioning for it? well – no dice.  If you find me at Teched, I can give a few more of these misses.

And what is with the error messages? If you try to display data – and if you don’t have the right previleges, all you get is a java exception that shows up in top right of the screen. It is useless – and we have to trace to find the issue.  Took us a bit to figure out all the right access requirements. I did not get a good feeling that I can implement HANA security for a productive instance. Need to work on this some more .

Other than to test some inserts – we did not use SQL script at all. We stuck to the visual tools, and all its limitations. Calc views would work a lot better with SQL, but our intent was to do it in the most straightforward manner possible. Visual modelling is ok – but has room for plenty of improvement on consistency.  For example – I could get automatic mapping for unions, but not for joins.

In any case – by yesterday evening, we over came all the hurdles – and could get data to flow from source to hana, and get it to show up in BI 4.0 reports.  That is quite an achievement – and I am incredibly proud to be part of the team.

Special thanks to my fellow gang members – Gagan, Tomas, Manish, Ayub, Sudhir ,Suresh and Abhishek.  You guys rocked it. It is amazing that a small group of passionate techies could accomplish so much, with limited time and resources.  We speak different languages, work in different time zones, belong to different organizations and hold various formal titles – but we made it work.
Without Tag Robertson doing the needful in getting us the shiny new hana box, none of this would have worked in the first place – thanks Tag !.  And many thanks to Bill, Elaine, Birdie and entire hosting team for the fast turnaround in getting the box wired and ready for us.  The ever helpful Xavier was always came through for us and we owe him one.  And I specifically want to thank our BI Partner Craig Rich and IBM’s SAP CTO Chuck Kichler for their unwavering support.

And from SAP – Patrice, Balaji, Mani and other colleagues – many thanks for the continued support.   There are many more from SAP and IBM who have contributed to our success – you know who you are, and I am very grateful to all of you.

Although I criticized it a fair bit – I also admire how well HANA has turned out to be by Rev 13.  SAP’s HANA team should be genuinely proud in getting a 1.0 product this far ahead. There is some work to be done – but that does not take away any of the credit that is due to the team that develops hana at SAP.

Now – lets try to put HANA into the cloud over the next weekend :)

About these ads

25 thoughts on “SAP HANA – we did it in 4 days, and lived to tell the tale

  1. Vijay Great Blog , brings back all the fun memories from the weekend. But there is nothing like a good night sleep after knowing that you have been a part of something awesome :))))

    @ Gagan, we probably would only get some sleep after TECHED :)))

  2. You are right that it would be more easy if all would be automated and without issues but then we would not have chance to learn HANA to the level we did. :-D
    From my experience if automated things are breking down then you need this extra knowledge to get it fixed and usually there are very few experts around that are really aware what is happening when some automation wizzard is running.
    I am not saying that we should not automate but rather that I appreaciate every opportunity when things are broken because then I can learn. (As you said here many things were broken so I was happy we could learn a lot.) :-)

  3. Hi Vijay,

    Great blog.
    I am sure you and the team would have enjoyed the whole experience, :)

    Thanks for you time in writing this down. I read it twice and enjoyed very much.

    regards,
    Sobin

  4. It is fun indeed as I have done similar exercise in my stable. My favorite HANA error so far was during data loads: “[ROLLBACK] – column store error: [0] No error “. And you were lucky to get JDBC error when you missed authorizations. I was struggling to get value help in another dialog, where there no error displayed at all and at the end happened to be authorization issue as well. Another strange thing – I was not able to find from the studio the schema owner (to request permissions) until going deep into metadata tables. Share your pain and joy. Cheers. -Vitaliy

  5. Excellent post and I guess now you understand my frustration with HANA when you implement quite complex use cases where you have to use SQLScript and load high volume of data with DS4.0.

    I actually hardly remember some of the issues anymore that you describe as they pale compared to what you experience when you have to hit a certain performance metric end to end and find out last minute that the recommended front-end tool just can’t do it or you have to use logical partitioning – back to the drawing board.

    Rev13 is definitely a step up from previous revisions but we still experience long activation times that HANA development team is looking into.

    There is just so little experience out there (including within SAP) that everything right now is trial and error – thanks god we have clients that understand we are working on a version 1 product.

    Looks like we have a lot to discuss with SAP at the upcoming TechEd!

  6. Learn to do a HANA “hello world” – attend my Teched session BI105…? Can i have web ex link to attend the session

    • Great post. Eagerly waiting for your session on HANA at TechED. Do you borrow time from parallel universe, by any chance :-)

  7. Great post Vijay! I also have some questions. First: which scenario you have install? And second: in many SAP guides about HANA you’ll find inaccuracies. And the last one at this moment: did you use Sybase? So I also have some problems with this installation (like everyone ;-) )
    Thx for reply

  8. Pingback: SAP rolls out new in-memory applications for HANA

  9. A beginner learning about bi , bo and latest Hana , excellent blog and enjoyed reading the entire process of it being set-up .
    Kind regards,
    Koushik

  10. Hi Vijay,
    I have recently started working with the HANA team. It is interesting to see how it all started. Thanks for sharing!

  11. Pingback: HANA related « dailylearnbw

  12. Hi Vijay, Kudos…Here I am reading the LSS first HANA install experience, good to know the enthusiasm of the core team :)

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s