Reading the real bottleneck before you touch anything
TYPO3 is a heavier application than most CMS platforms, and on CloudLinux-based shared hosting every account runs inside an LVE (Lightweight Virtual Environment) cage with hard ceilings on CPU, memory, I/O, process count, and entry processes. When a TYPO3 site stalls, returns 508 errors, or throws white screens under load, the cause is almost always one specific resource hitting its cap, not the whole server being slow. Guessing wastes time, so start by reading the numbers CloudLinux already records for you.
In cPanel Jupiter, open Metrics → Resource Usage (the "CPU and Concurrent Connection Usage" tool). The Faults column is the one that matters: a fault means a limit was reached and your account was throttled or a process was refused. On DirectAdmin Evolution, the same data lives under Advanced Features → Resource Usage (LVE). Look at which row accumulates faults: CPU, PMEM (physical memory), EP (entry processes), NPROC (number of processes), or IO/IOPS. A TYPO3 backend that freezes while an editor saves a large page while cron also runs is usually an EP or NPROC ceiling. A frontend that collapses only during traffic spikes is typically CPU or EP. A site that is slow all day with no traffic is frequently I/O or a database problem.
Pair those graphs with the per-request evidence. TYPO3 writes application errors to var/log/typo3_*.log inside your installation (older 9.x/10.x builds use typo3temp/var/log/), while PHP fatals and memory-exhaustion messages land in the error_log file that LiteSpeed drops into the directory where the fault occurred. Open File Manager, enable "Show Hidden Files," and read the most recent error_log in your document root and inside typo3/. Lines containing Allowed memory size of ... exhausted point at PHP memory, Maximum execution time points at CPU-time limits, and repeated mysqli connection errors point at database saturation. Only after you can name the faulting resource should you change a setting.
PHP memory, execution time, and worker pressure
The most common TYPO3 complaint on shared hosting is the backend dying during extension installs, scheduler runs, or large reference index updates with a memory-exhausted fatal. TYPO3 12 LTS realistically wants 256 MB to run the backend comfortably, and heavy tasks such as bin/typo3 referenceindex:update or large Extension Manager operations can push past 512 MB. You control this at the account level without touching any server file.
In cPanel use Select PHP Version → Options (the PHP Selector) and raise memory_limit, max_execution_time, and max_input_vars. DirectAdmin exposes the same values under Select PHP Version → Options when PHP Selector is enabled. If the UI does not expose a value, set it in a .user.ini file in your TYPO3 document root:
; .user.ini in the TYPO3 web root
memory_limit = 512M
max_execution_time = 240
max_input_vars = 1500
upload_max_filesize = 64M
post_max_size = 68M
opcache.enable = 1Two cautions. First, .user.ini changes are cached for the duration of user_ini.cache_ttl (300 seconds by default), so wait five minutes or touch the file again before retesting. Second, max_execution_time only governs PHP's own counter; the CloudLinux CPU-time limit still applies, so a 240-second value does not grant 240 seconds of raw CPU if your plan caps you lower. For long maintenance jobs, run them through the TYPO3 CLI as a cron command instead of the browser, because CLI processes are not bound by max_execution_time and a cron entry survives a closed browser tab.
"PHP workers" on LiteSpeed shared hosting map to your Entry Processes (EP) limit, usually between 20 and 50 concurrent PHP requests depending on plan. TYPO3 is especially prone to exhausting this because an uncached frontend request, a backend AJAX call, and an image-processing call can each hold a worker for seconds. The fix is almost entirely about caching rather than demanding more workers. Confirm the TYPO3 caching framework is backed by the database or, better, a local file backend, and never run production with [FE][disableNoCacheParameter] = false left open to abuse. Raising memory or worker-adjacent settings without fixing caching simply lets you hit the EP wall faster.
Caching, OPcache, and I/O relief that stays on shared hosting
TYPO3 rendering is template- and query-heavy, so the single biggest CPU and I/O win on shared hosting is making sure the frontend is actually served from cache. In the TYPO3 backend, confirm pages are cacheable (avoid config.no_cache = 1 in TypoScript except for deliberate dynamic pages) and that your cache tables are being used rather than rebuilt on every hit. Flush and warm caches from Admin Tools → Maintenance → Flush TYPO3 and PHP Cache, then watch whether CPU faults drop on the next traffic window.
Enable OPcache through the PHP Selector so compiled PHP stays in memory between requests; TYPO3's large class map benefits enormously. Add a sensible interned-strings buffer in .user.ini:
opcache.enable = 1
opcache.memory_consumption = 128
opcache.interned_strings_buffer = 16
opcache.max_accelerated_files = 20000
opcache.validate_timestamps = 1For the frontend, serve static assets and let LiteSpeed cache full pages where appropriate. A clean .htaccess in your web root removes a surprising amount of I/O and CPU by handing compression and expiry to the server instead of PHP:
<IfModule LiteSpeed>
CacheLookup on
</IfModule>
<IfModule mod_expires.c>
ExpiresActive On
ExpiresByType image/webp "access plus 1 month"
ExpiresByType text/css "access plus 1 week"
ExpiresByType application/javascript "access plus 1 week"
</IfModule>
<IfModule mod_deflate.c>
AddOutputFilterByType DEFLATE text/html text/css application/javascript
</IfModule>Inode and I/O exhaustion deserve attention because TYPO3 generates thousands of small files. Processed images in fileadmin/_processed_/ and typo3temp/assets/, plus log files in var/log/, quietly consume inodes. Check your inode count in cPanel under Statistics (File Usage) or DirectAdmin's usage panel. If you are near the inode cap, run Admin Tools → Maintenance → Remove Temporary Assets to clear processed images safely, and prune old logs via File Manager. A site that keeps regenerating the same thumbnails on every request is both an I/O and inode drain; offloading media delivery reduces that pressure, a topic covered in the TYPO3 media optimization guide in this category.
Database saturation and the honest VPS threshold
When CPU and memory look fine but pages are slow and the error_log shows connection timeouts, the bottleneck is usually the database. Open phpMyAdmin and inspect the largest tables. On TYPO3 the usual offenders are sys_log, sys_history, cache_* tables, and sys_refindex. A sys_log table with millions of rows slows the backend dashboard and inflates backups. You can safely truncate expired cache tables and trim the log from Admin Tools → Maintenance → Analyze Database Structure and the scheduler's "Table garbage collection" task. Schedule that task to run nightly so growth never gets out of hand. Add missing indexes only through TYPO3's own database analyzer rather than hand-editing in phpMyAdmin, so the schema stays consistent with extension expectations.
On shared hosting you cannot tune innodb_buffer_pool_size or any server-wide MySQL variable, and you should not try. What you can control is query volume: enable persistent caching, keep the reference index healthy with a scheduled referenceindex:update, and disable verbose SQL logging in production. If after all of this the Resource Usage graph still shows sustained CPU faults during normal traffic, or your database working set clearly exceeds what a shared instance can keep warm, you have reached a genuine ceiling.
Recommend a Cloud VPS when the evidence is concrete: steady EP or CPU faults across multiple days at normal load, a database that cannot be trimmed below the memory the shared instance allocates, background scheduler tasks that routinely exceed CPU-time limits, or inode usage that climbs faster than cleanup can recover. A VPS lets you own the PHP-FPM pool sizing, MySQL buffer pool, and OPcache memory outright. Until those signals are unambiguous, the caching, OPcache, inode, and database-trimming steps above typically recover most of the headroom a busy TYPO3 site needs while keeping you on a managed shared plan.