An article to understand Alibaba Cloud's load balancing performance-guaranteed instance

What is a load balancing performance guaranteed instance?

Guaranteed performance instances provide guaranteed performance indicators (performance SLAs). In contrast, the performance-sharing instance does not guarantee the performance indicators of the instance, and the resources are shared by all instances.

Before Alibaba Cloud Load Balancer launched performance-guaranteed instances, all the instances purchased by users were performance-sharing instances. On the console, you can view the type of purchased instance.

Move the mouse to the green icon of the guaranteed performance instance to view the specific performance indicators, as shown in the following figure.

instance type

The three key indicators of performance-guaranteed instances are as follows:

  • Maximum number of connections - Max Connection

    The maximum number of connections defines the maximum number of connections that a load balancing instance can carry. When the number of connections on the instance exceeds the maximum number of connections defined by the specification, the new connection request will be discarded.

  • New connections per second - Connection Per Second (CPS)

    New connections per second defines the rate at which new connections are made. When the rate of new connections exceeds the number of new connections per second defined by the specification, the new connection request will be discarded.

  • Queries Per Second - Query Per Second (QPS)

    The number of requests per second is a concept unique to Layer 7 monitoring, which refers to the number of HTTP/HTTPS queries (requests) that can be completed per second. When the request rate exceeds the number of queries per second defined by the specification, the new connection request will be dropped.

Alibaba Cloud load balancing performance-guaranteed instances provide the following six instance specifications.

Specification 1 Minimal I (slb.s1.small) 5000 3000 1000
Size 2 Standard I (slb.s2.small) 50000 5000 5000
Specification 3 Standard II (slb.s2.medium) 100000 10000 10000
Size 4 High-order type I (slb.s3.small) 200000 20000 20000
Size 5 Advanced Type II (slb.s3.medium) 500000 50000 30000
Size 6 Super I (slb.s3.large) 1000000 100000 50000

How to choose a performance-guaranteed instance?

Users need to select performance specifications based on specific services. The overall principles are as follows:

  • If it is a four-layer monitoring, the focus is on the number of concurrent connections for long connections, then the maximum number of (concurrent) connections should be used as a key indicator for reference. According to different business scenarios, you need to estimate the maximum number of concurrent connections that a load balancing instance needs to carry, and select the corresponding specifications.

  • 如果是七层监听,关注的重点是QPS的性能,QPS决定了一个七层应用系统的吞吐量。同样,也需要根据经验对QPS进行预估。在初步选定一个规格后,在业务压测和实测过程中对规格进行微调。

  • 结合与性能保障型实例一起推出的其它关键监控指标,查看实际业务流量的走势、峰值情况,对性能规格进行更加精确的选取。更多详细信息,参考监控数据

    并发连接数监控示例

    number of concurrent connections

    新建连接数监控示例

    Number of new connections

    QPS监控示例

    QPS

性能保障型实例的变配操作限制

用户可在控制台对性能保障型实例进行变配,如下图所示。

Variation 1

Variation 2

按量付费的性能保障型实例的规格可以升配也可以降配,包年包月的性能保障型实例目前暂时只允许升配,不允许降配。

因此,建议先使用按量付费的实例进行业务测试,确认好规格后再购买所需规格的包年包月实例。

另外,变更性能保障型实例规格时,如果同时变更计费方式(按流量计费或按带宽计费),则规格变更需要到次日零点才能生效。如果仅仅是对实例规格进行变更,变更立即生效。建议在变更规格时,尽量不要变更计费方式。

注意:由于历史存量原因,部分实例可能存在于较老的集群。此部分实例在变配到性能保障型实例时,因为需要将实例迁移,因此可能出现10-30秒的业务中断,因此建议在业务低谷期进行此类变配,或通过GSLB来做实例间的负载均衡后,再进行变配。

warning

性能保障型实例售卖开放计划

阿里云负载均衡从5月份开始在美东1地域小范围公测性能保障型实例,目前已经开放性能保障型实例地域和后续计划如下:

  • 美东1(弗吉尼亚):已上线
  • 华北2(北京):已上线
  • 华东1(杭州):已上线
  • 华东2(上海):已上线
  • 华北1(青岛):已上线
  • 华北3(张家口)及海外各地域:后续陆续上限开放,敬请关注

如何处理原有的性能共享型实例?

原有的性能共享型实例将继续保持性能共享型实例,不收取规格费。也可以通过变配,将性能共享型实例升级成性能保障型实例。变更成性能保障型后,后续在保障型实例正式收费时,该实例将收取规格费。

change

为何有时性能保障型实例看起来达不到规格中的性能指标上限?

短木板原理。

性能保障型实例并不保障三个指标(包含带宽指标)同时达到指定规格的指标上限。即规格中那个指标先达到峰值,就以哪个指标开始限速。同样,如果购买了按带宽付费的实例,当实例带宽达到峰值上限时,也可能会出现因为带宽限速而导致某些指标达不到规格上限的情况。

比如某用户选择高阶型I(slb.s3.small)实例,当实例的QPS已经达到20000,但并发连接数确远未达到20万,那么该实例最大连接数可能永远都不会达到规格上限,因为新建的连接请求会因为QPS达到上限而被丢弃。

注意:如果使用了高防产品导致业务流量的源IP变的非常少,有可能产生HTTPS业务流量负载不均衡的情况,此类情况下建议工单联系售后处理。

为什么有时候低配的性能保障型实例比共享型实例的性能还差一些?

性能共享型实例在资源上是共享的,在业务高峰期性能指标没有保障;而性能保障型实例可以保障任何时刻的性能。

何时开放API来创建、变配性能保障型实例?

性能保障型实例暂时还未开放API来进行创建和变配。开放时间请关注后续官网公告、邮件及短信通知。

还可以购买性能共享型实例吗?

当前继续开放性能共享型实例的售卖,后续性能共享型实例下线计划会通过官网公告、邮件等方式通知。

性能保障型实例如何收费?

性能保障型实例根据不同规格收取规格费,如下表所示。阿里云负载均衡提供了一种免费的规格,可满足以前绝大部分共享型实例用户的需求。

性能保障型实例何时开始收费?

当前性能保障型实例处于公测期,性能规格免费。后续收费计划将通过官网公告、短信、邮件方式进行通知。

Original link: https://yq.aliyun.com/articles/203367?spm=5176.100244.teamhomeleft.48.udfZ2M

Guess you like

Origin http://43.154.161.224:23101/article/api/json?id=326479109&siteId=291194637