Was this helpful?
System Sizing
Installation Options
You have three options to install and run Vector:
1. Use a dedicated server (and dedicated storage) and the base Operating System to run Vector.
2. Create a virtual machine and run Vector inside the virtual environment.
3. Use resources in the Cloud to run Vector.
This chapter focuses on Option 1. Most of the discussion is relevant for all options, although a virtual environment is not as ideal as a dedicated hardware environment, especially as it relates to storage throughput. When using Vector in the Cloud, you may not be able to influence the physical storage throughput. If most or all your data will fit in memory in a virtual environment or in the Cloud, then you may not need to be concerned about this.
The maximum possible data processing throughput Vector can achieve depends on several factors including the data types involved in the query, the data, and the query itself, including the type of computation. Vector has been observed processing data at incredible throughputs reaching over 1.5 GB/s per CPU core.
In all cases you will need to justify the budget for the configuration you choose. For example, you can trade faster CPUs for less memory or more memory for slower storage. Given the affordability of servers as of the publication of this document, you should consider what you can save over the lifetime of your application versus the benefits users can achieve by serving them well. Faster CPUs that use more power generally perform faster, so choose higher wattage CPUs over less power-hungry ones. Access to memory is faster than access to disk, so for most applications, you want to spend your budget on more memory before you think about faster storage options.
Last modified date: 04/26/2023