Techniques for Achieving High Concurrency: Optimistic Lock, Read‑Write Splitting, CDN, Caching, and Load Balancing

Server performance, database performance, network latency, and even programming language choices affect the maximum concurrent requests a system can handle; the common solution is to split responsibilities as finely as possible.

Optimistic Lock – a classic database optimization that uses a version field to prevent dirty reads without exclusive locks. Example table definition and Java entity annotations:

CREATE TABLE tbl (
    id VARCHAR(32),
    optimistic_lock NUMERIC(12) -- version field
);

@Version
@Column(name = "optimistic_lock", columnDefinition = "INTEGER")
private long optimisticLock;

Read/Write Splitting – separating master (write) and slave (read) databases, often configured with two data sources in Spring:

<!-- Define data source, using a custom implementation -->
<bean id="dataSource" class="cn.itcast.usermanage.spring.DynamicDataSource">
    <!-- Set multiple data sources -->
    <property name="targetDataSources">
        <map key-type="java.lang.String">
            <!-- Keys must match those used in code -->
            <entry key="master" value-ref="masterDataSource"/>
            <entry key="slave" value-ref="slave01DataSource"/>
        </map>
    </property>
    <!-- Default to the write (master) data source -->
    <property name="defaultTargetDataSource" ref="masterDataSource"/>
</bean>

DAO methods then choose the appropriate source for reads or writes, and master/slave servers synchronize to keep data consistent.

Resource Separation & CDN Acceleration – static assets such as images, JavaScript, CSS, or ZIP files are stored on separate servers or CDNs, reducing load on the main application server and bringing content closer to users.

Dynamic‑to‑Static Conversion – for idempotent requests, pre‑compute results (e.g., daily batch jobs that calculate each user’s monthly transaction total) and serve the cached static data directly, eliminating the calculation step at request time.

Cache – another form of dynamic‑to‑static, typically a key‑value store in memory. Example using MemCachedClient:

MemCachedClient mc = new MemCachedClient();
String key = "cacheKey1";
Object value = SomeClass.getObject();
mc.set(key, value);

Server Mirroring & Load Balancing – deploying multiple server instances (mirrors) across regions or ISPs, then distributing traffic based on load. Example Tomcat connector configuration for two instances:

<Connector port="11009" protocol="AJP/1.3" redirectPort="8443" />
<Engine name="Catalina" defaultHost="localhost" jvmRoute="tomcat1">
    <Cluster className="org.apache.catalina.ha.tcp.SimpleTcpCluster"/>
</Engine>

<Connector port="12009" protocol="AJP/1.3" redirectPort="8443" />
<Engine name="Catalina" defaultHost="localhost" jvmRoute="tomcat2">
    <Cluster className="org.apache.catalina.ha.tcp.SimpleTcpCluster"/>
</Engine>

Additional considerations include database sharding, proper synchronization, and basic SQL optimizations such as avoiding full‑table scans.

Source: https://www.v2ex.com/t/364701