Designing Highload-Block Structure in 1C-Bitrix

Designing Highload-Block Structure in 1C-Bitrix

Imagine: a catalog with 50,000 products and 200 characteristics — each new parameter adds millions of rows to b_iblock_element_property. The database starts to slow down, indexes bloat, the facet index takes hours to rebuild. How do you offload the system without losing performance? We see the solution in competent highload-block design. It's not just "create a table" — it's an engineering task that determines the speed of the entire catalog.

An HL-block is a generator of custom tables in MySQL/PostgreSQL on top of D7 ORM. Technically, it's a record in b_highload_block, a set of fields in b_user_field, and an automatically created table hl_{XML_ID}. Nothing magical — just a way to create a table with typed fields through the Bitrix interface without writing SQL. But exactly how this structure is designed determines whether the HL-block will work as a fast reference or become a bottleneck.

When HL-block, and when infoblock or custom table?

An HL-block replaces an infoblock where sections, SEO fields (META_TITLE, META_KEYWORDS), preview/detail images, and built-in publication mechanisms are not needed. These are reference books: brands, countries, tags, units of measurement, characteristics for facets.

An HL-block loses to a custom D7 table (DataManager) when:

• Composite indexes are needed (HL supports only indexes on individual fields via UF_*)
• Foreign keys and cascade operations are needed
• The table schema changes frequently and requires migrations

In these cases, it is better to create a class inheriting \Bitrix\Main\ORM\Data\DataManager and manage the table through it.

How to choose between HL-block and custom table?

Conclusion: use HL-block if composite indexes and relationships are not required. Otherwise — DataManager.

HL-block field types and their features

HL-blocks use the user field system (UF_*). Available types:

• string / string_formatted — VARCHAR. For names, titles.
• integer — INT. For numeric identifiers, sorting.
• double — DECIMAL/FLOAT. For prices, coefficients.
• boolean — TINYINT(1). Activity flags.
• file — stores file ID from b_file. For images in reference books.
• enumeration — binds to b_user_field_enum. For fixed statuses within an HL record.
• datetime / date — DATETIME / DATE.
• iblock_element / iblock_section — binds to an infoblock element or section. Use with caution: creates an implicit link between HL and infoblock.

The problem with the iblock_element field type in an HL-block: when an infoblock element is deleted, the HL record is not automatically updated — you need an event handler for OnAfterIBlockElementDelete.

How to properly index HL tables?

By default, an HL-block creates a table only with a PRIMARY KEY on the ID field. All other fields are without indexes. If the HL-block is used as a reference for the catalog facet index, additional indexes are usually not needed — the facet works with b_iblock_{ID}_index, not directly with the HL table.

But if the HL-block is used to store operational data (action history, order log, bonus program records) — indexes on filter fields are critical. They are added manually via SQL migration:

ALTER TABLE hl_loyalty_history ADD INDEX idx_user_id (UF_USER_ID);
ALTER TABLE hl_loyalty_history ADD INDEX idx_date (UF_DATE);

Bitrix does not provide a UI for managing HL-table indexes — only direct SQL or a migration script.

Case from our practice: HL-block for a B2B catalog

An electronics distributor. The catalog has 2,200 unique product characteristics (technical parameters). Initially — infoblock properties of type "String", b_iblock_element_property contained 8 million rows.

Solution: move reference characteristics to HL-blocks by domain:

• hl_tech_connectivity — connection interfaces (USB-C, HDMI, etc.)
• hl_tech_resolution — screen and matrix resolutions
• hl_tech_standard — standards (Wi-Fi 6, Bluetooth 5.2, etc.)

Each HL-block: fields UF_NAME (VARCHAR 255), UF_XML_ID (VARCHAR 50, unique), UF_ACTIVE (boolean), UF_SORT (integer). Indexes on UF_XML_ID — added manually for fast search during 1C import.

Result: b_iblock_element_property shrank from 8 million to 1.2 million rows (only numeric properties remained). The facet index rebuilds in 8 minutes instead of 45.

Managing HL-block data via D7

Working with an HL-block in code — through \Bitrix\Highloadblock\HighloadBlockTable and a dynamically created class:

$hlblock = \Bitrix\Highloadblock\HighloadBlockTable::getById($id)->fetch();
$entity = \Bitrix\Highloadblock\HighloadBlockTable::compileEntity($hlblock);
$dataClass = $entity->getDataClass();
$rows = $dataClass::getList(['filter' => ['UF_ACTIVE' => true]]);

This is a standard pattern — it is important to establish it in the project code style so that HL-block accesses do not spread across templates in the form of SQL queries.

What is included in the HL-block design work

Estimated time for a full cycle of 10-15 HL-blocks — from 2 to 5 working days. The cost is calculated individually depending on the complexity of the subject area and the need for migration. To get a consultation and evaluation of your project — just write to us. We are a certified 1C-Bitrix partner with 10+ years of experience and 50+ completed performance optimization projects. We guarantee transparency at every stage.

Useful resources:

• More about HL-blocks: 1C-Bitrix documentation
• ORM standards: Bitrix D7 ORM

Project Architecture Design on 1C-Bitrix: Avoiding Common Mistakes

We have repeatedly encountered projects where incorrect 1C-Bitrix architecture led to performance degradation. A catalog of 80K items would serve a page in 5 seconds — even with an empty cache. The architecture determines performance and support costs. Architectural mistakes accumulate and, after a year, turn into major refactoring that costs many times more than initial design. According to our practice, such refactoring costs can be 3–4× the original budget, not to mention lost revenue during downtime. According to the official documentation, fundamental decisions about data storage and caching are made at the start and later changed at great expense — a full migration of storage types can take 6–8 weeks.

Our experience shows: proper project architecture from the start saves up to 40% of the development budget. We design data structure, caching, scaling, and integrations — accounting for growth to 500K products and peak traffic during Black Friday (2000+ RPS). Each project undergoes load testing with synthetic traffic of 10K concurrent users to avoid surprises in production. Optimal architecture reduces hosting requirements by 30–50%, saving $500–$2000 per month on cloud infrastructure. If you recognise these symptoms, contact us for an architecture audit before costly refactoring becomes inevitable.

How to Choose Storage Type for 1C-Bitrix?

This is the first and most expensive architectural decision. Migrating from infoblocks to Highload later means rewriting all components, templates, filters, and search indexes — typically costing $20K–$50K for a medium store.

Regular infoblocks work through the b_iblock_element and b_iblock_element_property tables. Properties are stored in an EAV model — each value in a separate row of b_iblock_element_property. With 50 properties and 100K elements, you get 5 million rows in one table. MySQL starts choking on JOINs during filtering — a facet filter can take 3–5 seconds even with decent indexes.

Infoblocks are good for:

• Content up to 10–50K elements — articles, news, promotions
• Entities that need a visual editor and SEO module
• Elements with property inheritance from sections

Highload blocks are flat tables. One entity — one table with columns. No EAV. Filtering on indexed columns works an order of magnitude faster. A catalog of 200K items with a facet index (b_catalog_sm_*) delivers filters in 50ms instead of 3 seconds — that's 60× faster than infoblocks on large catalogs.

Highload blocks are required for:

• Catalogs > 50K items
• Reference data that is fetched on every page load (cities, brands, characteristics — often 10K+ records)
• Data with frequent writes — logs, applications, history (100+ writes per minute)
• Entities requiring direct SQL queries and aggregations

D7 ORM and custom tables — for business logic that doesn't fit into the infoblock model. Many-to-many relationships, computed fields, custom aggregations. Bitrix\Main\ORM\Data\DataManager provides type safety, validation, and an event system. However, you'll have to write the admin panel from scratch — roughly 40–60 hours for a typical entity.

Scaling 1C-Bitrix Without Performance Loss

Horizontal scaling is a topic where 90% of projects fail. However, people think about it only when the site is already down.

The first step — move sessions from files to Redis. Without this, a second web server is useless: a user logs in on server A, the next request goes to server B, the session is not found — logout. In .settings.php:

'session' => ['value' => ['mode' => 'redis', 'host' => '127.0.0.1', 'port' => 6379]]

Next:

• nginx upstream or HAProxy distributes requests. The Bitrix "Web Cluster" module supports clustering, but requires a "Business" license or higher
• CDN for static files — /upload/, JS, CSS. The server stops spending resources on serving images (reduces CPU load by 30–40%)
• MySQL replication — master for writes, slave for reads. Bitrix supports up to 9 slave connections via .settings.php. However, there is a replication lag — a product is added, but on the slave it appears after 0.5–2 seconds. Use sticky reads for critical data

Vertical scaling is cheaper and faster initially:

• EXPLAIN every heavy query. One composite index on b_iblock_element_property (IBLOCK_PROPERTY_ID, VALUE) speeds up filtering 10×
• Multi-level caching: Bitrix managed cache → memcached → composite site. Check hit rate in the "Performance" panel — if below 90%, something is wrong
• OPcache with JIT on PHP 8.1+ — free 15–30% acceleration

Composite site mode can serve pages in 0.1s for anonymous users — we use it for 80% of traffic.

Offloading Heavy Processes from the Monolith

Bitrix is a monolith, and that's fine. Breaking it into microservices is madness. But offloading heavy processes is the right move.

Import/export is the most common pain. Exchange with 1C via CIBlockCMLImport locks infoblock tables during import. 100K items — that's 20–40 minutes when filtering on the site slows down. Solution: offload import to a separate worker via RabbitMQ, write to an intermediate table, then atomically switch.

• Search — Elasticsearch instead of the built-in search.title. Full-text and faceted search, autocomplete, typo correction. Load on MySQL is completely removed. We achieve <100ms for full-text search on 500K products.
• Notifications — push, SMS, email via queue. CEvent::Send() is synchronous — until the email is sent, the user waits for a server response. A queue (RabbitMQ or Redis list) reduces response time by 200–500ms.
• Report generation — PDF, Excel on large volumes (10K+ rows). Separate process, result — a download link.

API: REST, GraphQL, Webhooks

Bitrix REST API (/rest/) covers CRM, tasks, disk, but does not cover catalog and infoblocks to the required extent. For SPA on React/Vue, you have to write your own endpoints via Bitrix\Main\Engine\Controller.

• GraphQL — for mobile applications where traffic is expensive. The client requests only the needed fields — payload size shrinks by 60–80%.
• Webhooks — event model: new order → POST to external URL. No need to poll the API every 5 minutes.
• Versioning — /api/v1/, /api/v2/. Without this, API updates break all consumers at once.
• OpenAPI/Swagger — auto-generation of documentation. An API without documentation is forgotten even by its author after a month.

Main Sources of Technical Debt in Bitrix

Technical debt in Bitrix is specific. Three main sources:

1. Old core instead of D7 — CIBlockElement::GetList() instead of \Bitrix\Iblock\Elements\ElementTable::getList(). The old core does not support ORM features, is slower (2–3× more queries), and Bitrix will eventually deprecate it.
2. Direct SQL in component templates — $DB->Query("SELECT...") directly in template.php. Move to service classes, replace with ORM.
3. Business logic in result_modifier.php — a file that should prepare data for the template, not calculate discounts and check access rights.

Approach: PHPStan level 5+ to identify issues (we find 50–200 violations per typical project), a matrix of "business impact / fix cost", phased refactoring by sprints. Not everything at once — but the trend must be downward.

Avoiding Costly Refactoring

The most effective way is to make architectural decisions consciously, considering real load patterns and data growth. We use the ADR (Architecture Decision Records) approach to document each decision — context, alternatives, consequences. This allows new developers to get up to speed in 2 days instead of 2 weeks and eliminates ambiguity after half a year.

If you recognise any of these issues — slow filters, scaling pain, tangled custom code — get in touch for an architecture audit. We'll identify technical debt and propose a migration plan.

Documentation: ADR Instead of Word Files

• ADR — Architecture Decision Records. A short file: context, decision, consequences. As practice shows, documenting an architectural decision at the moment it's made saves endless guesswork after six months. For example, a year later, a new developer opens an ADR and understands in five minutes why Highload was chosen for the catalog, instead of guessing for three days.
• Diagrams — servers, data flows, integration points. PlantUML or Mermaid, stored in the repository next to the code.
• ER diagrams — infoblocks, properties, relationships. Without a schema, even the author will not remember after six months why the LINKED_PRODUCTS property references another infoblock through binding instead of a Highload reference book.
• Runbook — deployment, rollback, scaling, actions during a crash. Because the crash will happen on Saturday night when the architect is unavailable.

How We Design Architecture

1. Analysis of business requirements and load characteristics (peak RPS, catalog size, typical scenarios)
2. Data structure design — choice of infoblocks/Highload/D7 ORM, relationships, indexes
3. Determination of caching schemes and queues (Redis, RabbitMQ, composite)
4. Prototyping and load testing on real data (200K records, 30+ properties)
5. Documentation — ADR, ER diagrams, runbook, API specifications
6. Project review — internal and with the client

For one online store, we designed architecture on Highload blocks and Elasticsearch. Product filtering down to 50ms, time to first byte 0.3s. Hosting cost savings: 45% per month.

Scope of Work

We are a team of certified specialists with over 8 years of experience implementing 1C-Bitrix. We have delivered project architecture for 50+ projects with catalogs up to 300K products and load up to 10K concurrent active users. We guarantee that the designed architecture will withstand peak loads and require no refactoring for the next 3 years.

Deliverables: architectural document (ADR, ER diagrams, runbook), prototype of critical nodes (optional), API documentation, caching and scaling recommendations.

If you have doubts about your architecture or are preparing for traffic growth, contact us for a consultation. We will audit the current structure and propose an optimal strategy. Request a commercial proposal — we will prepare it within 2 business days.