Mobility

Explaining why MessageQueue in RocketMQ is split into writeQueue and readQueue. First, let me briefly introduce what a queue in RocketMQ is. Here is an architecture diagram of RocketMQ found online.As we know, when using RocketMQ, we produce or consume against a specific topic. MessageQueue is a part of a topic, similar to the concept of partitions in Kafka. A queue is the minimum unit for consumer consumption, meaning the number of consumers cannot exceed the number of queues. A very speci...

ClickHouse is an open-source database for real-time data analytics developed by Yandex, initially used in multiple data analytics projects internally at Yandex. To introduce ClickHouse, we first need to introduce Yandex. ClickHouse’s emergence is closely related to Yandex’s business needs. Yandex is Russia’s largest search engine, with many data analytics projects. The largest of these by data volume is Yandex.Metrica, a website analytics service similar to Baidu Analytics, with data volume s...

Bitwise operations are a widely used computation method in computer science. When used appropriately, they can greatly improve computational efficiency. Today, I’d like to introduce a clever application of bitwise operations: converting between uppercase and lowercase letters. To tackle this problem, we can first examine the ASCII code characteristics of uppercase and lowercase letters. As we can see, the ASCII code values of each pair of uppercase and lowercase letters differ by 32. In bi...

In a previous article, I introduced the basic syntax for using ActiveMQ on the client side. It is worth noting the consumer side: 12345678910111213141516171819202122232425262728293031323334353637383940414243444546474849 public void consume() throws JMSException { ActiveMQConnectionFactory cf = new ActiveMQConnectionFactory(); cf.setBrokerURL("tcp://localhost:61616"); Connection connection = cf.createConnection(); connection.start(); Sessio...

Recently, I participated in optimizing an older billing system and learned some common techniques for balance deduction under high concurrency. I also tried some approaches, so I’m summarizing my findings here. Problem DescriptionFor a billing system, concurrency issues are actually divided into two categories. The first is high application concurrency—essentially a large number of users and high access volume. This type of problem is no different from general high concurrency problems and ...

Problem IntroductionI encountered an interesting problem today while scanning code with FindBugs — it’s about the ternary operator. Let me record it here. It’s about code like this: 12boolean b = true; Long a = b ? 0l : Long.valueOf(2); FindBugs gave the warning: “Boxed value is unboxed and then immediately reboxed” — meaning a boxed object is unboxed and then immediately reboxed. This problem is actually quite common. I didn’t pay much attention to it at first, and just habitually change...

OverviewI’ve been looking for an apartment recently, and I wanted to find a location that’s relatively convenient to get to several places. Checking the map manually was quite troublesome, so I decided to build a small tool for planning Beijing subway routes. This article briefly introduces the implementation process. The current functionality is relatively simple — the main method takes an input starting station and lists the routes with the minimum number of stops from all other stations to...

In a previous article Understanding Redis Cluster Principles Through Hands-on Practice, we briefly introduced the design principles of Redis Cluster. Data in Redis Cluster is distributed across 16384 slots according to certain rules, and these slots are mapped to different nodes based on the configuration. We know that after the cluster is running stably, data can still be transferred at the slot level. However, I’ve never been entirely sure about the specific transfer process, including clus...

OverviewAsyncLoopThread is a custom loop task execution utility in JStorm. The implementation is not complex, and on its own may not warrant a dedicated article. However, it is so widely used in JStorm — for features such as supervisor/nimbus heartbeat, fetching new topologies, updating worker status, and many more — that it’s worth introducing here, as it will also help with reading other parts of the codebase. UsageAsyncLoopThread can actually be extracted and used in your own projec...

IntroductionData persistence is a concern that many systems deal with, especially for systems like Redis and ActiveMQ where data is primarily stored in memory. Since data is stored in memory, there is a risk of data loss when the system crashes. The solution to this problem is to write data to disk through some mechanism — that is, persistence. ActiveMQ provides three persistence methods, based on JDBC, KahaDB, and LevelDB respectively. Currently, the officially recommended method is KahaDB-b...