ඉතින් උබට වෙන්න ඕනි දේ පැහැදිලිව LLM එකට කියපන් එතකොට solution එක දෙනවා.redis use karnne in memory architecture ochchara request handle karanne dunnoth kelevena server ekata. redis khomath use karanna ba. RabbitMQ vala disk enable karanna puluwan storage valata but websocket vage simple communication ekak handle karata kafka tharam resilience na 1000000 request handle karanna concurrently. anith eka oka push based, so backpressure handle karnna hari amarui.
Idempotency
RabbitMQ and Redis dekama EOS support karanne na, let's comes to Kafka. Producer side ekin EOS enable karala Idempotency achieve karanna puluwan. othanath kelvaenna puluwan mama passe explain karannam. consumer side ekin idempotency achieve karanna nam, one partion and concurrency level 1 ekata yanna venava. ehem neththnam Redis or DB use karala track karanna venava, Redis is in memory and not suitable for this, when it comes to DB it is fucking expensive. but achieve karanna puluwan but the problem is.
"it is not guaranteed Orders"
mama prashna awwema mata ona multiple request execute karnna order based with multiple partions.
passe explain karnnam kiwwe. verdaila hari load balancer ekin ekama request ekak service dekaketa send karaoth EOS enable karath thibbath instance dekak two different producer vidiya salakanna and message sequence number also differnt. so data duplicate venava, payment or banking app nam loku payment ekak nam kelvenava.
ආ ඉදා පුළුවන් නන් කරපං වගේ ඉන්නේ නැතුව උගෙන් support එක ගනින්.
