#optimisations

Realme 7 Pro Gets September Security Patch, Optimisations With Latest Update

Realme 7 Pro has started receiving its first software update less than a week after its launch in the country. Although the update is not listed on Realme’s site at the time of writing, Gadgets 360 spotted the update on our Realme 7 Pro review unit. The update comes with version RMX2170_11_A.11 and along with several optimisations, brings the latest September security patch. The phone is yet to…

This version brings a few user-invisible — but extremely useful nonetheless — changes:

integration of the cacheops django module, to avoid the database hammerings introcuded by the processor infrastructure. I prefer a lot hitting Redis than PostgreSQL to (re-(re-))load processors and processing chains at every article.

addition of interface button to trigger selective or global cache clearing; this helps making static-files and templates upgrades immediately visible to the user, or fix Javascript issues when non-cached templates fragments are desynchronized with too-agressively cached JS fragments (or the other way around).

resolution of some small bugs like logging strings arguments or corner-case typos (yeah I know, coverage-driven tests could have helped find these earlier ; help welcome)…

The first point still needs love and thinking. We could probably benefit from a dedicated task fetching and parsing chunks of articles at once in a batched manner. Chains and Processors could remain in memory, avoiding repeated and useless hits to any database (even Redis will not be touched thanks to the Django internal instance cache).

This would speed things up at worst. It would eventually lower the resources needed by celery, at the cost of less parallelization possible.

But as the workers are currently all on the same machine, this would be certainly bring a benefit, be it small.

Some satisfying level of parallelization could still be achieved by lowering the chunks of articles to fetch — in the constance configuration for example. Letting workers handle more jobs before beiing dropped from memory.

But this last point needs particular caution, because of external parser's memory leaks that greatly affect the machine. This issue is the only one that made me drop workers after a fixed amount of processing.

In case our new processors can be made to leak less (or not at all!), we could raise the number of jobs a worker can handle in its life, and thus again benefit from the Django internal cache.

One single optimization, multiple levels of benefits. I love this! But hunting in our current (very efficient) Cython-based parser is a very difficult task, and also very error prone.

Il existe encore quelques sites Internet qui présentent des textes ou liens cachés qui ne sont là uniquement pour les moteurs de recherche. Pour les découvrir la majorité du temps, il suffit de désactiver le JavaScript de son navigateur pour découvrir des lignes de textes de folie. Des sur-optimisations assez amusantes pour la plupart sachant que les effets sont négatifs ou même neutres certaines fois tellement c'est mal fait. Nous ne vous présenterons pas d'exemple ici, ce n'est pas cool.

Cela arrive aussi avec le netlinking que réalisent certains, des listes de liens cachés sur des sites Internet pour afficher par exemple des partenariats. Vous aurez deviné que c'est aussi de la sur-optimisation et que cela n'est pas recommandé par les principaux moteurs de recherche et qu'il est facile d'être pénalisé pour ce type de techniques qui sont d'ailleurs inutiles.