Long term database experience has supported OLTP workloads and delivers reliable cross platform migrationsSAP is not putting money into modernizing SAP Adaptive Server Enterprise. One of the things I discovered on the last project I was on was that they did not incorporate the Intel new instruction set in SAP Adaptive Server Enterprise. Intel has augmented its instruction set referred to as new instructions. They did that to make conversion easier. When you migrate SAP Adaptive Server Enterprise cross-platform, you go through a process where it converts the character set. If you are going from AIX to Linux or from Solaris to Linux, Linux is referred to as Little Endian, while AIX or Solaris are considered Big Endian. This is determined by how the product stores data. The word size of these processors is 32 bits long. If you start numbering from the little end, it is referred to as Little Endian. If you start numbering from the big end, it is called Big Endian. To migrate a SAP Adaptive Server Enterprise database from a Big Endian setup like AIX or Solaris to migrate to a Big Endian setup on an Intel, the operating system determines whether it is Little Endian or Big Endian. When you migrate from Big Endian to Little Endian, the database has to go through a character set conversion, and some of these databases are quite large with gigabytes and gigabytes of data. They have to do a character set conversion to the existing database before they do anything else. The worst part is that you have to rebuild all the indexes when you do that. When you switch endianness of the database, you have to rebuild all the indexes. It will automatically do that for system tables, but for actual user databases, you have to rebuild all your indexes, and it takes a long time. SAP Adaptive Server Enterprise is a relational database and is the predecessor of Microsoft SQL Server. All that functionality that Microsoft SQL Server had came from essentially SAP Adaptive Server Enterprise. The problem with SAP Adaptive Server Enterprise these days is it is not expanding its place in the marketplace or expanding its position in the marketplace. A lot of companies have migrated away from SAP Adaptive Server Enterprise. It works fairly well, but the problem is SAP Adaptive Server Enterprise was architected to be an OLTP engine and is now doing things for larger databases that were not in its original intended purpose. The endianness of the RDBMS is a major impediment to continuing to use SAP Adaptive Server Enterprise. You have a multi-gigabyte database, and it will go through a conversion process in a single-threaded fashion, and then you have to rebuild the indexes. Rebuilding the indexes is lengthy and time-consuming. The part of the conversion process that is concerned with conversion of the character set is single-threaded. You may have eight cores on your machine or virtual machine and only one can be used in the conversion process. There is another problem with the whole thing in that it will sometimes not operate properly. Under certain workloads, SAP Adaptive Server Enterprise will become overwhelmed. When you convert it, it does not operate properly in all circumstances. The root cause of that is that SAP in its desire to save money and desire to orphan the product has not recompiled or redeveloped the product to take advantage of the Intel new instruction set. Other relational databases such as Oracle and Microsoft SQL Server have the same issue to deal with, but with those platforms, they are taking advantage of the new instruction set. There are some additional Intel instruction sets or instructions in their Intel instruction set. With SAP Adaptive Server Enterprise, they did not bother to incorporate support for the new instruction set instructions. In certain circumstances, the database does not operate properly. It is unable to do what it needs to do. If you do your research and go on the internet and see what happens with Oracle or Microsoft SQL Server, what comes back is that it takes 4% longer to perform a lot of the instructions. When you are using the new instruction set, it adds 4% to the runtime of the database.
