Azure服务总线实体吞吐量 [英] Azure Service Bus Entity Throughput

问题描述

根据本文: https://azure.microsoft.com/en ...，服务总线每个队列/主题每秒最多可处理2000条消息.本文: https://azure.microsoft.com/en ...说这意味着分区队列或主题的整体吞吐量不再受单个消息代理或消息传递存储的性能的限制."如果我们创建分区队列/主题，我相信它会在内部创建16个分区.我的问题是:分区队列/主题的吞吐量是否线性增加到16 x 2000 = 32,000(大约)?还是保留2000 msg/sec的吞吐量.您能指出分区队列/主题的任何吞吐量基准吗?我们目前正在分析需要非常高吞吐量主题的方案.关于这个问题的指导将非常有帮助.

As per this article: https://azure.microsoft.com/en..., Service Bus can process upto 2000 messages per second per queue/topic. This article: https://azure.microsoft.com/en... says 'This means that the overall throughput of a partitioned queue or topic is no longer limited by the performance of a single message broker or messaging store.' If we create a partitioned queue/topic, I believe it creates 16 partitions internally. My question is: Does the throughput of a partitioned queue/topic rise linearly to 16 x 2000 = 32,000 (approximate) ? Or does the throughput of 2000 msg/sec remain. Can you please point me to any throughput benchmark for partitioned queue/topic? We are currently analyzing a scenario that needs a very high throughput topic. Guidance on this question will be really helpful.

此问题也已发布在Azure网站上，网址为: https://azure.microsoft.com/zh-CN/documentation/articles/service-bus-performance-improvements/

This question has been posted on the Azure site also, at: https://azure.microsoft.com/en-us/documentation/articles/service-bus-performance-improvements/

推荐答案

性能会因多种因素而有所不同.例如，许多高级功能(例如重复检测，会话，事务等)将增加开销，从而降低吞吐量.如果您想自己进行基准测试，请尝试运行以下性能示例: https://github.com/Azure-Samples/service-bus-dotnet-messaging-performance

Performance will vary based on a number of things. For example, many advanced features (like duplicate detection, sessions, transactions, etc,) will add overhead therefore decreasing throughput. In case you want to benchmark this yourself, try running this performance sample: https://github.com/Azure-Samples/service-bus-dotnet-messaging-performance

此外，由于基本/标准是共享的租户模型，因此您今天所获得的性能可能与明天所获得的性能有所不同.此外，您可能看不到线性比例尺，因为某些节点可能比其他节点更忙.

Also, because basic/standard are shared tenant models the performance you get today, may be different from the performance you get tomorrow. In addition you may not see linear scale as some nodes can be more busy than others.

对于真正的可预测性，保费是您最好的选择.

For true predictability, premium is your best bet.

这篇关于Azure服务总线实体吞吐量的文章就介绍到这了，希望我们推荐的答案对大家有所帮助，也希望大家多多支持IT屋！