Sizing Considerations

This section explains how to determine the disk space and RAM that will be required for your Exasol database.

Factors that impact sizing

There are several factors that determine the storage disk space and RAM that will be required for an Exasol system. The major factors to consider are:

The expected volume of your raw data

The volume of your uncompressed (raw) data has the largest impact on the estimation of the required storage disk space. The larger the data volume, the more storage space is required. Table data in an Exasol database is automatically compressed to optimize disk space usage.

Performance

The amount of RAM that is allocated for the database (DB RAM) will have a direct impact on performance. When estimating how much DB RAM you need, you can use the rule-of-thumb of 10% of the volume of your uncompressed data. For a more precise value you can calculate the amount of active data, or use the actual values in Exasol system tables.

Cluster redundancy

The planned redundancy for the cluster affects the required storage space. With redundancy 2 – meaning that the same data is stored in two segments – the required storage space is effectively doubled.

Number of reserve nodes

Reserve nodes are used in case of a node failure. A reserve node must have the same hardware configuration as the active nodes in a cluster. For more information, see Fail Safety (On-Prem)

Backup strategy

Certain backup strategies require more storage space. If you plan on storing backups in the cluster (local backups) with redundancy 2, then this must be taken into account when estimating disk size. For more information, see Backup and Restore.

Business continuity strategy

The architecture of your business continuity solution influences sizing. For example, if you plan on having a reserve cluster in a separate data center, you will need to double your hardware costs. For more information, see Business Continuity.

Operating system reserved RAM

We recommend that you reserve 10% RAM for the operating system on each node. This must be considered when calculating the amount of RAM to allocate for the database (DB RAM).

Database disk space

The total required disk space for an Exasol system is the sum of the required database disk space and the required backup disk space. The factors that you should consider when calculating the database disk space are:

Compressed data volume

Table data in an Exasol database is automatically compressed. A typical compression rate is ~2.5, which means that if the raw data volume is 2500 GiB, the estimated compressed data volume is ~1000 GiB.

Index volume

Indexes are automatically created and maintained by the database and require disk space. The index size depends on the chosen data model and queries, and can range from 2% to over 100% of the compressed data. A typical Exasol system will have an index volume of about 15% of compressed data.

Statistical and auditing data volume

Statistical data volume is small. However, if you switch auditing on in the system, the required disk space increases because each login and each query is stored in the corresponding auditing tables.

Temporary data and fragmentation

When intermediate results do not fit into the database RAM they are swapped out to a temporary volume. We recommend that you reserve extra headroom for temporary DB RAM.

The persistent volume can become fragmented to some degree, which increases the amount of disk space taken up by the data. We recommend that you reserve additional disk space to avoid running out of space due to fragmentation.

As a rule of thumb, add 60% of the compressed data volume (without redundancy) as combined additional headroom for temporary data and fragmentation.

Database disk space calculation

To calculate the required disk space for the database, use the following equation:

Compressed data volume + Index volume + Statistical and auditing data volume 
* Redundancy 
+ Temporary data and fragmentation headroom
= Required database disk space
Example

In this example we have the following input parameters:

  • Raw data volume: 2500 GiB
  • Redundancy: 2

Using the above equation and rule-of-thumb guidelines, the calculation will be as follows:

Parameter Value Explanation
Compressed data 1000 GiB 2500 GiB of raw data / 2.5 (average compression factor)
Index volume 150 GiB 15% of compressed data
Statistical and auditing data 50 GiB 5% of compressed data
Total data volume (net) 1200 GiB Compressed data + index volume + statistical and auditing data
Total data volume with redundancy 2400 GiB Total data volume (net) × 2
Headroom for temporary data and fragmentation 720 GiB 60% of compressed data without redundancy
     
Required database disk space 3200 GiB Total data volume with redundancy + headroom for temp data and fragmentation

Backup disk space

To calculate the required disk space for backups, use one of the following equations depending on your backup configuration:

Backups are stored outside of the cluster (remote backups)

(Full backup size * (number of full backups + 1)) 
+ (Incremental backup size * number of incremental backups)
= Required backup disk space

Backups are stored in the cluster (local backups)

((Full backup size * (number of full backups + 1)) 
+ (Incremental backup size * number of incremental backups)) * 2
= Required backup disk space

Creating a new backup does not remove the old backup. Because of this, there must be headroom for an extra backup during backup creation in addition to the total number of stored backups (number of full backups + 1).

Example

In this example we have the following input parameters:

  • Total net data volume: 1200 GiB
  • Number of full backups: 2
  • Number of incremental backups: 3
  • Maximum incremental backup size: 100% of full backup
  • Cluster-internal backup: Yes
  • Backup redundancy: 2

The backup space calculation (including headroom for an extra backup during backup creation) will then be as follows:

Parameter Value Explanation
Total data volume (net) 1200 GiB Compressed data + index volume + statistical and auditing data
Full backup data size 2400 GiB Total data volume (net) × 2
Incremental backup data size 3600 GiB Total data volume (net) × 3
Required backup disk space without redundancy 6000 GiB Full backup data size + Incremental backup data size
Required backup disk space with redundancy 12000 GiB

Required backup disk space without redundancy × 2

Because different versions of an object can be accessed my multiple queries run by different users, the backup can be larger than the physical layout of the objects themselves. We recommend that you include additional space to allow for this in the archive volume.

Database RAM (DB RAM)

An Exasol database typically performs well with database RAM of 10% of the raw (uncompressed) data volume. However, several other factors also affect the required DB RAM:

Index volume

Indexes are automatically created and maintained by the database. The index size depends on the chosen data model and queries, and can range from 2% to over 100% of the compressed data. A typical Exasol system will have an index volume of about 15% of compressed data.

Higher index volumes can negatively impact system performance and require more DB RAM.

If you have a running system, you can get the actual size of the indexes from the AUXILIARY_SIZE_* columns in the EXA_DB_SIZE_* statistical system tables. For more information, see Statistical System Tables.

Temporary data

When intermediate results do not fit into the available DB RAM they are swapped out to a temporary volume, which causes significant performance deterioration. Because of this, we recommend that you reserve extra headroom for temporary data (TEMP DB RAM).

User defined functions (UDFs)

When processing large amounts of data using UDFs, the RAM required for those UDFs must be available on every node. The UDFs are executed in parallel, which means that there can be as many instances of a UDF per node as there are cores. Therefore, you have to consider the total amount of RAM that the UDF instances need for processing the queries.

For example, if a query uses 500 MiB per UDF instance on a 72 core machine in an 8-node cluster, this requires an additional 282 GiB of DB RAM.

You can also specify how many UDF instances are created within the UDF. For more information, see UDF Instance Limiting.

Database RAM calculation

To calculate the required DB RAM you can use the following equation:

MAX( COMPRESSED DATA VOLUME * ESTIMATED DB RAM PCT, INDEX VOLUME * INDEX SCALE FACTOR ) 
+ COMPRESSED DATA VOLUME * TEMP DB RAM HEADROOM PCT 
+ MAX UDF RAM * NUMBER OF CORES * NUMBER OF NODES 
= Required DB RAM
Parameter Explanation
COMPRESSED DATA VOLUME

The compressed size of your raw data.

Table data in an Exasol database is automatically compressed. A typical compression rate is ~2.5, which means that if the raw data volume is 2500 GiB, the estimated compressed data volume is ~1000 GiB.

ESTIMATED DB RAM PCT An estimated percentage of the compressed data volume that will be needed for DB RAM. A typical value is 20%.
INDEX VOLUME

The estimated size of indexes in the database.

A typical Exasol system will have an index volume of about 15% of compressed data.
INDEX SCALE FACTOR

A scale factor to allow for headroom for index volume.

To ensure that index maintenance does not impact system performance, extra DB RAM should be considered for processing indexes. The index scale factor is multiplied by the estimated index volume and represents the minimum DB RAM required. The index factor is especially important if you expect to have a high index volume.
TEMP DB RAM HEADROOM PCT

Additional DB RAM for processing temporary data, as a percentage of the compressed data volume.
MAX UDF RAM The total amount of RAM that will be needed for your UDF instances.
NUMBER OF CORES The number of cores in each node.
NUMBER OF NODES The number of nodes in the cluster.

If you have a running system, you can use the RECOMMENDED_DB_RAM_SIZE_* columns of the EXA_DB_SIZE_* statistical system tables to get a recommended DB RAM size. For more information, see Statistical System Tables.

Example:

In this example we have the following input parameters:

Raw (uncompressed) data volume 2500 GiB
DB RAM size as % of compressed data volume 20%
Index scale factor 1.3
Headroom for temporary DB RAM 5%
Estimated UDF RAM requirement None

The resulting equation would then be:

MAX(1000 GiB * 20%, 150 GiB * 1.3) + 1000 GiB * 5% = 250 GiB

Explanation:

  • Required DB RAM based on the compressed data volume will be our starting point, since this is larger than the estimated index volume multiplied with the index scale factor (200 GiB vs. 150 GiB).

  • Additional headroom required for TEMP DB RAM is calculated to be 50 GiB (5% of 1000 GiB).

  • The total required DB RAM will thus be 200 GiB + 50 GiB = 250 GiB.