Why a 128 MB Buffer Pool Slows Down Bitrix?
MySQL spends a whole second on a query that should run in 10 milliseconds. SHOW STATUS LIKE 'Innodb_buffer_pool_reads' returns tens of thousands of physical reads per minute — data is constantly read from disk. The cause is almost always the same: innodb_buffer_pool_size is set to the default 128 MB, while the working data set of a Bitrix site has long exceeded a gigabyte.
We have repeatedly encountered situations where owners of online stores on 1C-Bitrix complain about slowdowns under peak load, and the root of the problem is precisely the buffer pool. For example, recently on a project with 50,000 products, the hit_ratio was 87% — every seventh read went to disk. After tuning the pool to 4 GB, hit_ratio rose to 99.8%, average query time dropped from 350 to 20 ms. Over 5 years, we have tuned InnoDB on hundreds of projects and guarantee at least a 2-3x speedup in queries after proper configuration.
What is the InnoDB Buffer Pool and Why Its Size Is Critical
The buffer pool is the RAM allocated by InnoDB for caching data and index pages. If the working data set fits in the buffer pool, MySQL reads from RAM. If not, each miss costs several milliseconds of disk I/O. More about the mechanism can be read on Wikipedia.
A typical medium-sized Bitrix site:
-
b_iblock_element_property— 500 MB – 3 GB -
b_iblock_element— 50–300 MB -
b_catalog_price,b_catalog_product— 100–500 MB - indexes for these tables — about the same
Total working set — 1–8 GB. The default 128 MB covers less than 10%.
Diagnosis and Calculation of Buffer Pool Size
How to Diagnose Insufficiency?
Run the queries:
SELECT (1 - (Innodb_buffer_pool_reads / Innodb_buffer_pool_read_requests)) * 100 AS hit_ratio FROM ( SELECT VARIABLE_VALUE AS Innodb_buffer_pool_reads FROM information_schema.GLOBAL_STATUS WHERE VARIABLE_NAME = 'Innodb_buffer_pool_reads' ) r, ( SELECT VARIABLE_VALUE AS Innodb_buffer_pool_read_requests FROM information_schema.GLOBAL_STATUS WHERE VARIABLE_NAME = 'Innodb_buffer_pool_read_requests' ) rr; SHOW STATUS LIKE 'Innodb_buffer_pool_pages_flushed'; SHOW STATUS LIKE 'Innodb_buffer_pool_wait_free'; If hit_ratio is below 99% — the buffer pool is definitely too small. If Innodb_buffer_pool_wait_free is non-zero — critical shortage: write operations are waiting for free pages.
How to Calculate the Optimal Size?
Find the actual size of data and indexes:
SELECT ROUND(SUM(data_length + index_length) / 1024 / 1024, 0) AS total_mb, ROUND(SUM(data_length) / 1024 / 1024, 0) AS data_mb, ROUND(SUM(index_length) / 1024 / 1024, 0) AS index_mb FROM information_schema.TABLES WHERE table_schema = 'your_database'; Rule: buffer pool = 70–80% of the working data set size, but not more than 70–80% of total server RAM. On a server with 8 GB RAM and a 4 GB database — set 4–5 GB.
Configuring Buffer Pool in my.cnf and Online Change
[mysqld] innodb_buffer_pool_size = 4G innodb_buffer_pool_instances = 4 innodb_buffer_pool_chunk_size = 128M innodb_buffer_pool_instances — number of independent pools. Reduces thread contention for mutexes. Rule: 1 instance per 1 GB of pool, max 64. For a 4 GB pool — 4 instances.
innodb_buffer_pool_chunk_size must be a divisor of the total: innodb_buffer_pool_size = chunk_size * instances * N. In the example: 4G = 128M * 4 * 8 — correct.
MySQL 5.7.5+ supports online buffer pool resize:
SET GLOBAL innodb_buffer_pool_size = 4294967296; -- 4 GB in bytes The resize process is asynchronous. Monitor progress:
SHOW STATUS LIKE 'Innodb_buffer_pool_resize_status'; While resizing, there may be a temporary performance drop. Change during a low-load window.
Which InnoDB Parameters Are Critical for Bitrix?
Besides buffer pool, it's important to set:
innodb_log_file_size = 512M innodb_log_buffer_size = 64M innodb_flush_method = O_DIRECT innodb_io_capacity = 2000 innodb_io_capacity_max = 4000 innodb_flush_method = O_DIRECT is critical on Linux: without it, data is cached twice — in the InnoDB buffer pool and in the kernel page cache. This wastes RAM.
Monitoring and Results
Monitoring After Changes
After 30–60 minutes under load:
SELECT ROUND(Innodb_buffer_pool_bytes_data / 1024 / 1024 / 1024, 2) AS cached_gb FROM (SELECT VARIABLE_VALUE AS Innodb_buffer_pool_bytes_data FROM information_schema.GLOBAL_STATUS WHERE VARIABLE_NAME = 'Innodb_buffer_pool_bytes_data') t; SELECT ROUND( (SELECT VARIABLE_VALUE FROM information_schema.GLOBAL_STATUS WHERE VARIABLE_NAME = 'Innodb_buffer_pool_pages_data') / (SELECT VARIABLE_VALUE FROM information_schema.GLOBAL_STATUS WHERE VARIABLE_NAME = 'Innodb_buffer_pool_pages_total') * 100, 1 ) AS pool_used_pct; If pool_used_pct is consistently 95–100% — the pool is full; consider increasing. If 60–70% — current size has headroom.
Performance Comparison Before and After Tuning
| Parameter | Before (128 MB) | After (4 GB) |
|---|---|---|
| Hit ratio | 85–92% | 99.5%+ |
| Avg query time | 450 ms | 25 ms |
| Innodb_buffer_pool_reads/sec | 15,000 | 50 |
| Disk load (iowait) | 30% | 2% |
| Project Type | Recommended RAM | Recommended Buffer Pool |
|---|---|---|
| Store up to 10,000 products | 4 GB | 2 GB |
| Store up to 50,000 products | 8 GB | 5 GB |
| Store up to 200,000 products | 16 GB | 10 GB |
Typical Mistakes in Configuration
- Setting size larger than available RAM (causes swap).
- Ignoring
innodb_flush_method = O_DIRECTon Linux. - Changing size online without monitoring progress.
- Choosing values not multiples of
chunk_size(rounding error).
What's Included in a Turnkey Buffer Pool Tuning
- Diagnosis of current state: hit_ratio, data size, configuration.
- Calculation of optimal size considering RAM and load.
- Parameter changes (buffer pool, log, flush method) with online application.
- Post-change monitoring and report with recommendations.
- Consultation on other bottlenecks (indexes, Bitrix caching).
- Guarantee on result: hit_ratio > 99% after pool warm-up.
Contact us for a free preliminary diagnosis. Order buffer pool tuning and get database acceleration today. We'll help select the configuration for your project and ensure stable performance under peak load.