Versions Compared

Key

  • This line was added.
  • This line was removed.
  • Formatting was changed.

Table of Contents

...

This chapter describes things to consider when designing a DB or system using ALTIBASE. It is recommended to consider the following items and refer to the design because not only the impact on the design stage but also on the development stage and the operation stage is significant.

...

There are two replication methods: Lazy The lazy method and the Eager method. The Lazy method allows the delay of data transmission, and on the other hand, the Eager method does not allow it. Therefore, the Lazy method is adopted for tasks that allow data transmission delay, and the Eager method is set and operated appropriately for each task if data that does not necessarily need to be synchronized.

 

Comparison

Lazy Method

Eager Method

Performance

90% performance level of a single server

Performance degradation compared to the Lazy method

SynchrnoziationSynchronization

This is a structure that checks only the success of transmission of the changed transaction log to another server, so it is a high performance, but synchronization may be delated.

To prevent data inconsistency on another server, performance is rapidly delayed because it is also committed locally on another server, but the synchronization is not delayed.

 

...

Altibase replication is a form sending changed log of a transaction occurring on the local server. This means that even if one million statements are changed with one SQL statement, the same is the structure in which a million changed transaction logs are transmitted, not the same for as one SQL statement transmitted to a remote server. 

...

In addition to HA, it is necessary to establish a backup policy for data file damage caused by physical disk failure and failure caused by transaction log file corruption. The Altibase backup method provides both full and incremental backup methods using the archive mode.

...

In the case of the numeric type provided by ALTIBASE, a type such as Numeric (9) can be sufficiently expressed as an integer type, but occupies 8 bytes of space. It occupies 4 bytes per record more than integer type. In terms of processing performance, numeric internally incurs the cost of converting back to native type, which is somewhat slower than the integer type.

It is recommended to the followings following when selecting a column type.

Consideration

Description

Numeric Type

1.Consider to minimize  Consider minimizing the conversion cost of Native ßà Non-Native

(Native type [Integer, Double, BigInt] is more advantageous in terms of performance)

2. Double type is selected when the precision of the real type is not required.
3. If Sum/Avg is used, select Double/BigInt type

Date Type

If the date-related operation is important, select the Date type. If it is limited to search and comparison operations, select the Char/Varchar type. (Date type is fixed at 8 bytes)

Join

Consider that the type conversion does not occur if the column is a Join

...

For the ALTIBASE partition table (as of version 6.3.1), partitioned local index and non-partitioned global index are supported, and partitioned global index is not supported. Therefore, it is recommended that the partition table uses historical data for the purpose of storage for each month and purports of business because a significant performance decrease occurs when querying in the  the form that excludes the partition key in the condition clause that queries the entire partition table.

...