如何用Firebase或SNS向150万用户发送含用户名的自定义推送通知？

你当前使用messaging.send()单条逐个发送的方式，150万次独立HTTP请求必然导致耗时极长——每次请求都要建立连接、处理响应，效率极低。以下是基于Firebase和AWS SNS的优化方案：

一、Firebase Cloud Messaging (FCM) 优化方案 #

1. 改用批量发送API #

FCM提供send_multicast()接口，一次最多可发送500条消息，大幅减少HTTP请求次数。每条消息可携带独立的用户个性化数据：

from firebase_admin import messaging

# 假设user_list包含所有用户的{username, device_token}数据
messages = []
for user in user_list:
    update_body = {
        "username": user["username"],
        "content": "hope you are doing good...."
    }
    message = messaging.Message(
        data=update_body,
        token=user["device_token"],
    )
    messages.append(message)
    
    # 每攒够500条发送一次
    if len(messages) == 500:
        response = messaging.send_multicast(messaging.MulticastMessage(messages=messages))
        print(f"批次发送：成功{response.success_count}条，失败{response.failure_count}条")
        messages = []

# 发送剩余不足500条的消息
if messages:
    response = messaging.send_multicast(messaging.MulticastMessage(messages=messages))
    print(f"剩余发送：成功{response.success_count}条，失败{response.failure_count}条")

2. 客户端渲染个性化内容 #

把通用模板和用户名作为数据字段发送，由客户端负责拼接成完整通知正文，减少服务器端的字符串拼接开销：

// Android客户端示例
override fun onMessageReceived(remoteMessage: RemoteMessage) {
    val username = remoteMessage.data["username"] ?: "there"
    val content = remoteMessage.data["content"] ?: "hope you're doing well"
    val notificationBody = "Hey $username, $content"
    // 调用本地通知展示逻辑
    showCustomNotification(notificationBody)
}

3. 异步并发处理 #

用线程池并行发送批量请求，进一步提升发送效率：

from concurrent.futures import ThreadPoolExecutor
from firebase_admin import messaging

def send_batch(msg_batch):
    response = messaging.send_multicast(messaging.MulticastMessage(messages=msg_batch))
    return response.success_count, response.failure_count

# 拆分用户列表为500条一组的批次
batches = [user_list[i:i+500] for i in range(0, len(user_list), 500)]
message_batches = []
for batch in batches:
    msg_batch = []
    for user in batch:
        msg = messaging.Message(
            data={"username": user["username"], "content": "hope you are doing good...."},
            token=user["device_token"]
        )
        msg_batch.append(msg)
    message_batches.append(msg_batch)

# 用10个线程并行发送
with ThreadPoolExecutor(max_workers=10) as executor:
    results = executor.map(send_batch, message_batches)

# 统计总结果
total_success, total_failure = 0, 0
for success, failure in results:
    total_success += success
    total_failure += failure
print(f"总发送结果：成功{total_success}条，失败{total_failure}条")

二、AWS SNS 推送方案 #

1. 注册设备端点并关联用户属性 #

将用户设备令牌注册为SNS端点，同时绑定username属性，方便后续模板调用：

import boto3

sns = boto3.client('sns', region_name='your-region')

# 批量注册端点（需提前创建SNS平台应用）
for user in user_list:
    endpoint_arn = sns.create_platform_endpoint(
        PlatformApplicationArn='your-platform-app-arn',
        Token=user["device_token"],
        Attributes={"username": user["username"]}
    )["EndpointArn"]
    # 将endpoint_arn与用户ID关联存入数据库
    save_user_endpoint(user["id"], endpoint_arn)

2. 使用模板批量发送 #

提前在SNS创建消息模板（内容示例：Hey {{username}}, hope you are doing good....），然后批量发送到端点：

# 从数据库获取所有用户的endpoint_arn
endpoint_arns = [user["endpoint_arn"] for user in user_list]
template_arn = 'your-sns-template-arn'

# 每100条为一组批量发送
for i in range(0, len(endpoint_arns), 100):
    batch = endpoint_arns[i:i+100]
    request_entries = [
        {
            'Id': str(idx),
            'TargetArn': arn,
            'TemplateArn': template_arn
        } for idx, arn in enumerate(batch)
    ]
    response = sns.publish_batch(PublishBatchRequestEntries=request_entries)
    print(f"批次发送：成功{len(response['Successful'])}条，失败{len(response['Failed'])}条")

三、方案对比 #

• FCM：适合已有Firebase生态的项目，批量API简单易用，免费额度足够支撑大流量，客户端个性化处理灵活。
• AWS SNS：适合AWS生态项目，端点管理和模板功能更成熟，支持多平台推送，按发送量计费。

内容的提问来源于stack exchange，提问作者santosh kumar panda