What is our primary use case?
I am currently using the latest version. But before that, before I jumped into the version, I used the initial version of HANA, as well. This initial version of HANA was not that great, it had a lot of bugs. But the latest version is very good. It's excellent.
I'm afraid that HANA is not a relational database, it's a column-level database just like Sybase IQ. Sybase is also an activity product, an SAP product. SAP bought Sybase in May 2010. So normal Sybase is RDBMS. Sybase has one more variant called Sybase IQ. That is not RDBMS, that is a column-level database. Normal Sybase is a whole-level database. That's a column-level database. So SAP HANA is based on this column-level architecture.
One more thing. The success of HANA primarily depends on the RAM and the storage. HANA became a success because the cost of the solar devices has fallen down substantially. I don't know about British Pounds, but in Indian Rupees, earlier in 2007, 2008, when I was working for Microsoft, one terabyte of a SAN device, used to cost around 22.5 LAK. I would say I would have had a 100,000. I think that's the nature. So one SAN device was costing 22 LAKs. The same SAN device, in 2013 and 2014, was costing around three LAKs. So the SAN device cost reduced by more than 200%.
Also, in parallel, the RAM cost also decreased, and the technology and the fastness of RAM increased. This impacted the primary condition for RDB and RDBMSs like Oracle, Sybase, SQL Server, and the like, that they need to support the foreign key relationship, where I have a few tables. For example, if I have five to six tables, suppose the first table is employment information. The second table is employee career details or his project, something like that.
Now, instead of populating the tables with the same information, the primary condition of RDBMS was to have a foreign key relationship between these two tables and reduce the redundancy. That was a primary condition, but in HANA, thanks to the cheap storage and high-speed RAM, I may not even bother to do a redundancy of data. I can combine all the tables and make a huge table. And as an entire table, whatever its size, I can pin the table in the RAM so that my access of information is not from the hard disk, but is directly from the memory, which is much, much, much faster. That is the beauty of HANA.
What needs improvement?
I'm still researching the features of HANA. In terms of memory, data access and data pitching, HANA has scored a victory, no doubt about that. But when I compare the non HANA architecture with SAP, ERP, the SAP ERP comes in two levels. SAP ECC, which is a non HANA based product, and SAP S/4HANA, which is a HANA-based product. If I compare these two, there are almost around 5,000 to 6,000 tables, which were merged together in HANA, making it a robust architecture.
In earlier SAP we used to have fragmented, small-scale architecture. HANA is a robust architecture where one table itself is a behemoth quantity of many, many columns and a lot of redundant data. So my interest in HANA would be how SAP is catering to the demand of reducing the redundancy of data, and at the same time pinning the entire critical tables into the memory so that access to the data is faster. I am researching those factors.
For how long have I used the solution?
I have roughly five years of experience in SAP HANA, because I started working on SAP ECC, on logistics and other components. After that, HANA became famous only in the years 2013 and 2014. Then I started pursuing HANA very, very actively. Right now, my journey is continuing and after five to six years I have a good amount of knowledge and experience on HANA.
What do I think about the stability of the solution?
Buyer's Guide
SAP HANA
October 2024
Learn what your peers think about SAP HANA. Get advice and tips from experienced pros sharing their opinions. Updated: October 2024.
816,406 professionals have used our research since 2012.
What do I think about the scalability of the solution?
SAP HANA is vertically and horizontally scalable.
Our banking system uses HANA primarily for our financial transactions. There are our SAP financials running on HANA. This HANA SAP was on the Oracle database. We have migrated it. It's very, very complex and took almost one year for us to prepare the plan and migrate to HANA finance. There are around 700 to 800 users using the database and they're not facing any problem. It's fantastic.
How are customer service and support?
I would say I'm satisfied with technical support, buy it can be improved also. Improved in terms of data warehousing, because HANA was introduced for data warehousing and because SAP wanted to catch the OLTP market. Now they have introduced many things to attract the OLTP customers, especially in banking and telecom sectors. That's okay. You have to keep your business interests also. HANA's architecture is the foundation of the language of data, warehousing, and design. For any project or product, if it's based on data warehousing, I would say HANA is the language for that because what data warehousing wants is a data warehousing database.
Primarily, it's not an OLDP, it's OLAP, online analytics processing. And where the data is not changed, the data doesn't change as frequently as a OLTB database. For that kind of environment, I think HANA needs a lot of improvement in terms of making it more columnar. It has to incorporate up level design a little bit harder, as well.
You know MySQL database? Not Microsoft, MySQL. Microsoft is not SQL. M-Y-S-Q-L, has been bought by Oracle. Oracle bought MySQL, it acquired the MySQL company. If you look into the database, by default, MySQL engine is InnoDB. InnoDB is the default engine on MySQL. But, MySQL also gives you the flexibility of choosing your own engine. I don't want to know InnoDB, I have a huge Microsoft Excel file with around 10,000 rows, but I don't want to use InnoDB because I have to pay for that. To save those costs, at the time of starting MySQL engine, I can choose my type of data. Instead of InnoDB, I can choose Excel also. SAP HANA should give that kind of flexibility to its customers, making it more reachable to small SMEs, small and medium enterprises.
Now it is simple, because thanks to the cloud approach, it is giving a lot of flexibility to the customer, but if it wants to attack, hit the right target, acquiring the very, very small scale customer, who has around max 50 terabytes data or 100 terabyte data, a small scale company, small companies, that market should also be captured by SAP, not only the big companies. As the English saying goes, small things count. You can't ignore small things.
How was the initial setup?
The initial setup is pretty straightforward. The only thing was there were a lot of parameters which had to be taken into consideration and any parameter at installation will be paid. But one good thing about SAP HANA is even if you miss a single parameter, you cannot agree to it for the steps. The further steps will tell you that, "you have missed this step. You first complete it, then you can come here." That kind of interlinking is there. So yes, SAP installation is pretty straightforward, and very easy and smooth.
What other advice do I have?
I would recommend SAP HANA. No doubt I would definitely recommend it. But the thing is, if I adopt SAP HANA, my total cost of ownership in terms of having a functional consultant, as well as a HANA admin, would increase. I should first find a balance and analyze the data, "Do I really want to have HANA? What benefit will I have if I have HANA at my premises? And if I want to cut cost but also get the benefits of HANA, will the cloud option of HANA cater to my needs?" All those questions.
That is the company analysis I should do: what do they do differently? But many companies will be driven by the business needs, but at the same time some companies will also be driven by factors like the existing relationship with other vendors, like Oracle or SQL Server, and the kind of discounts they get when they buy that product. All those things will be there as driving factors. To answer your question, I would definitely recommend SAP HANA to anyone.
High availability and disaster recovery are very poor in HANA. High availability is measured on the barometer of RPO and RTO. RPO stands for recovery point objective, RTO stands for recovery time objective. The graph in which these two factors will be measured is from the five nines, the seven nines, or the three nines, that kind of factor. But it is a factor of my high availability. 99.9% of my database is available or 99.99999999%, giving a chance of 0.0001% for some kind of availability failure is because of natural disaster or some kind of electrical failures or something like that. So those are the factors you have to see for high availability.
My SAP HANA, technically, can withstand those calamities and recover itself from that disaster. That is called high availability. That high availability is there, but it is very, very, very minimal. If you're talking about high availability of HANA in actual high availability markets compared to Oracle and other RDBMS, HANA is a small child. If you remember when Microsoft SQL Server came into the RDBS market back in the year 1997, when they introduced the SQL 97, then they introduced the SQL 2000, SQL 2005. At that time, they introduced the high availability called Windows Cluster log shipping, mirroring the application.
At that time, in 2007 and 2008, Oracle introduced RAC, Real Application Clusters. Compared to the features of real application clusters, the Microsoft product was a small child. And Microsoft took that as a challenge and they improved and they improved. And in 2012 they introduced something called Always On. Always On is an improved version of high availability in SQL Server. HANA has to do that kind of stuff. HANA's high availability is immature.
On a scale of one to ten, I would rate SAP HANA an eight.
Which deployment model are you using for this solution?
Hybrid Cloud
Disclosure: I am a real user, and this review is based on my own experience and opinions.