53c575db3f
To cope with writing large files out of order, the system needs some form of strategy to avoid writing them directly to disk, as this causes insane amounts of disk thrashing. By default, the data is just buffered in RAM. This works well on a large server, but smaller systems struggle. To help systems with small RAM process large amounts of data, the old RandomWriteFunnel is brought back if the system property 'system.conserve-memory' is set to true. RandomWriteFunnel is buffering the construction by creating a series of small files that pigeonhole the writes into rough neighborhoods, and then it goes over the files one by one to construct one area of the file at a time. This is relatively slow and uses more than twice the disk size. A new interface RandomFileAssembler is introduced as an abstraction for this operation. A third strategy, direct mmaps, is also introduced if the file is very small (less than 1 GB). In this domain, disk thrashing is unlikely since it will comfortably fit in RAM. |
||
|---|---|---|
| .. | ||
| api | ||
| common | ||
| features-convert | ||
| features-crawl | ||
| features-index | ||
| features-qs | ||
| features-search | ||
| libraries | ||
| process-models | ||
| processes | ||
| services-application | ||
| services-core | ||
| tools | ||
| readme.md | ||
Code
This is a pretty large and diverse project with many moving parts.
You'll find a short description in each module of what it does and how it relates to other modules. The modules each have names like "library" or "process" or "feature". These have specific meanings. See doc/module-taxonomy.md.
Overview
A map of the most important components and how they relate can be found below.
The core part of the search engine is the index service, which is responsible for storing and retrieving the document data. The index serive is partitioned, along with the executor service, which is responsible for executing processes. At least one instance of each service must be run, but more can be run alongside. Multiple partitions is desirable in production to distribute load across multiple physical drives, as well as reducing the impact of downtime.
Search queries are delegated via the query service, which is a proxy that fans out the query to all eligible index services. The control service is responsible for distributing commands to the executor service, and for monitoring the health of the system. It also offers a web interface for operating the system.
Services
- core services Most of these services are stateful, memory hungry, and doing heavy lifting.
- application services Mostly stateless gateways providing access to the core services.
-
- api - public API
-
- search - marginalia search application
- an internal API
Processes
Processes are batch jobs that deal with data retrieval, processing and loading. These are spawned and orchestrated by the executor service, which is controlled by the control service.
Tools
Features
Features are relatively stand-alone components that serve some part of the domain. They aren't domain-independent, but isolated.
Libraries and primitives
Libraries are stand-alone code that is independent of the domain logic.