降低网络AI应用开发门槛，AIOps两大部署模式来助力

{"type":"doc","content":[{"type":"blockquote","content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"​​​​​​​​​​​​​​​​摘要：企业运维需求及挑战，来看看华为AIOps如何解决！","attrs":{}}]}],"attrs":{}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"本文分享自华为云社区","attrs":{}},{"type":"link","attrs":{"href":"https://bbs.huaweicloud.com/blogs/280543?utm_source=infoq&utm_medium=bbs-ex&utm_campaign=other&utm_content=content","title":"","type":null},"content":[{"type":"text","text":"《【云驻共创】AIOps？企业运维新力量！》","attrs":{}}]},{"type":"text","text":"，原文作者：启明。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"国际惯例，我们先介绍一下AIOps的概念：AIOps，即 ArtificialIntelligence for IT Operations，智能运维，将人工智能应用于运维领域，基于已有的运维数据（日志、监控信息、应用信息等），通过机器学习的方式来进一步解决自动化运维没办法解决的问题。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"Gartner预测，当前的IT应用程序会发生剧变，而且管理整个IT生态系统的方式也会改变。这些变化的关键是Gartner所称的AIOps平台。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"我们今天要讨论的，就是AIOps的需求挑战，以及我们通过怎么样的方式去应对这种挑战。","attrs":{}}]},{"type":"heading","attrs":{"align":null,"level":1},"content":[{"type":"text","text":"AIOps需求及挑战","attrs":{}}]},{"type":"heading","attrs":{"align":null,"level":2},"content":[{"type":"text","text":"（一）新技术、新挑战，呼唤高度智能的电信网络","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"近年来，以5G为代表的新技术在电信网络中得到了快速的应用。新技术的应用，给我们带来了很多的收益，比如大连接、低时延、高速率等等。5G的发展，让这些数据都至少有一个数量级的提升。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"image","attrs":{"src":"https://static001.geekbang.org/infoq/f7/f7506f204a92d2b763937064e350667d.jpeg","alt":null,"title":null,"style":[{"key":"width","value":"75%"},{"key":"bordertype","value":"none"}],"href":null,"fromPaste":true,"pastePass":true}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"​但是，数据量级的提升，伴随着的，是运维难度的增加，从而给运维带来了如下挑战：","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"strong","attrs":{}}],"text":"1. 网络复杂性：","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"strong","attrs":{}}],"text":"数据量级的增大，让网络变得更加复杂","attrs":{}},{"type":"text","text":"：新技术得到了快速应用，旧技术却没有同步退出，导致我们每引入一项新技术，都需要在原来的复杂度上做一个加法。而在某些场景式，甚至要去做乘法。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"比如，在无线领域，2G/3G/4G/5G，“四代同堂”；在核心网，PS/CS/MS物联网等等十域并存......如此高的网络复杂度势必会给运维带来相当大的挑战。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"strong","attrs":{}}],"text":"2. 2B新需求","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"运维的第二个挑战是To B的新场景，也就是企业应用。5G的应用推动了智能制造，网络也逐步融入到了企业的生产制造流程当中。在这种情况下，对网络可靠性的要求必然会提高，毕竟网络一旦出问题，生产流程就可能会受影响，甚至会中断，这样造成的损失将会非常大。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"strong","attrs":{}}],"text":"3. 成本压力","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"成本压力主要是由前面两个挑战传导而来。前两个挑战导致我们要么面临一个比较复杂的网络，要么就是有更高的要求。如果我们以传统的运维方式去应对的话，必然会导致成本的急剧上升。当然，成本的提高，还有一个因素就是能耗。毕竟，5G的能耗要远高于4G的能耗。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"针对上述这些挑战，我们要如何去应对呢？","attrs":{}},{"type":"text","marks":[{"type":"strong","attrs":{}}],"text":"AI技术是关键。","attrs":{}}]},{"type":"heading","attrs":{"align":null,"level":2},"content":[{"type":"text","text":"（二）AI是提升电信网络自动化和智能化的关键技术","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"在运维成本方面，有统计显示，90%的运维都需要人工去参与，而70%的成本就是人力成本。在这种情况下，一个很自然的想法就是能不能使用AI的技术来降低人的成本，来提高运维效率。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"比如刚才提到5G能耗问题，我们能否通过人工智能的技术来去降低能耗呢？从过往的实践经验来看，上述问题的答案是肯定的。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"接下来，我们通过三个例子来说明。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"image","attrs":{"src":"https://static001.geekbang.org/infoq/89/89603ebe1e5a9fc3433928e1228b17f9.jpeg","alt":null,"title":null,"style":[{"key":"width","value":"75%"},{"key":"bordertype","value":"none"}],"href":null,"fromPaste":true,"pastePass":true}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"strong","attrs":{}}],"text":"1. 基站节能","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"第一个例子是基站节能。基站的能耗是非常高的。在布网初期，基站用户较少，有时候基站常常是空开。针对这种情况，运营商的解决方案是对话务量做出一些预测。如果我们能精准预测话务量的话，那么，在话务量小的时候，我们就可以把一定量的载波关掉，从而达到节能的目的。据统计，在预测话务量的过程中，通过LSTM神经网络来做预测，可以实现节能10%以上。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"strong","attrs":{}}],"text":"2. 核心网KPI异常检测","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"第二个例子，是异常检测。在运营商的核心网部署KPI异常检测服务。原有的异常检测服务，是使用固定阈值进行告警通知。而AI技术，则更加智能、及时、准确地识别异常。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"strong","attrs":{}}],"text":"3. 故障识别及根因定位","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"通常网络上一旦发生故障，就会触发大量的告警，而系统同时又以高经纬维度进行运维派单。如果多个网员上报多个告警，那么就会出现这种重复派单。也就是说发生了一个故障，多网员上报告警，最后可能导致在多个域（无线域和传输域等）都去派单。","attrs":{}}]},{"type":"heading","attrs":{"align":null,"level":2},"content":[{"type":"text","text":"（三）开发AI应用仍然面临挑战：开发门槛高、周期长","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"从上面三个例子我们可以看出，AI相对来说，还是非常靠谱的。但是既然AI如此靠谱，为什么没有得到全面快速的应用呢？因为AI的开发还面临着不小的挑战，简单概括就是六个字：","attrs":{}},{"type":"text","marks":[{"type":"strong","attrs":{}}],"text":"门槛高，周期长","attrs":{}},{"type":"text","text":"。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"image","attrs":{"src":"https://static001.geekbang.org/infoq/4b/4b06784ec3ecbd5c15aa4f427ea4979f.jpeg","alt":null,"title":null,"style":[{"key":"width","value":"75%"},{"key":"bordertype","value":"none"}],"href":null,"fromPaste":true,"pastePass":true}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"​上图是Gartner的一份研究报告。它从四个维度分析了AI应用的主要障碍。其中最主要的3点：","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"bulletedlist","content":[{"type":"listitem","attrs":{"listStyle":null},"content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"人员技能","attrs":{}}]}]},{"type":"listitem","attrs":{"listStyle":null},"content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"理解增益与用途","attrs":{}}]}]},{"type":"listitem","attrs":{"listStyle":null},"content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"数据范围与质量","attrs":{}}]}]}],"attrs":{}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"这就回到我们说的六个字：门槛高，周期长。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"strong","attrs":{}}],"text":"1. 门槛高","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"此处说的“门槛高”，","attrs":{}},{"type":"text","marks":[{"type":"strong","attrs":{}}],"text":"第一点是指缺乏AI算法开发人员。","attrs":{}},{"type":"text","text":"一般的运维团队不会配置专门的AI算法开发人员，这样必然导致AI技能的缺失。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"但这不是最关键的，因为AI人员通过培训、培养、招聘等手段，都可以解决。 ","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"最关键的，","attrs":{}},{"type":"text","marks":[{"type":"strong","attrs":{}}],"text":"也就是我们说的第二点，算法与业务结合难。","attrs":{}},{"type":"text","text":"如果要想把一个应用做好，最好的是从业务出发，根据业务的实际情况选择合适的算法，这样才能把应用做好。但在实际操作过程中，首先，我们需要有一个业务专家对运维要有深刻的理解；其次，还需要有一个精通AI的算法专家。在这之后，需要他们有充足的时间和意愿坐下来深入的交流。在这里，时间和意愿都会成为阻碍。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"strong","attrs":{}}],"text":"第三点是数据。","attrs":{}},{"type":"text","text":"数据包含两个问题：工程问题和标注问题。即，开发一个AI应用实际上是相当大的工程量，因为首先需要接入海量的多模态的数据去完成模型的训练和推理，最后还要去完成结果的展示，包括去对接一些现有的系统。因此除了前面需要的运维专家和算法专家，还需要很多工程开发人员。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"strong","attrs":{}}],"text":"2. 周期长","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"开发门槛高，就决定了开发周期长，毕竟有这么高的门槛，如果不能很好的解决的话，那么周期必然会特别长。开发周期长会导致：","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"第一，","attrs":{}},{"type":"text","marks":[{"type":"strong","attrs":{}}],"text":"理解增益和用途","attrs":{}},{"type":"text","text":"。怎么理解呢？也就是说，如果我们长时间拿不到结果，那么企业决策人员就可能对AI能产生的效果会表示怀疑；","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"第二，","attrs":{}},{"type":"text","marks":[{"type":"strong","attrs":{}}],"text":"时间越长，大家对项目的期望就会越高。","attrs":{}},{"type":"text","text":"假设同样是做一个东西取得了同样的效果，比如说故障修复时长降低5%，两年做出来的和一个月做出来的，得到的评价可能就完全不一样。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"针对AIOps落地过程中遇到的挑战，华为推出的AIOps服务！现在我们一起来看看AIOps服务具体是什么，以及它是如何解决我们前面面临的挑战的。","attrs":{}}]},{"type":"heading","attrs":{"align":null,"level":1},"content":[{"type":"text","text":"华为AIOps服务","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"image","attrs":{"src":"https://static001.geekbang.org/infoq/ce/ce3457e038b2a995c3240fa42ef4d344.jpeg","alt":null,"title":null,"style":[{"key":"width","value":"75%"},{"key":"bordertype","value":"none"}],"href":null,"fromPaste":true,"pastePass":true}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"​上图是AIOps服务的整体框架。AIOps从下到上分成了四层：","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"strong","attrs":{}}],"text":"第一层：数据的采集和治理。","attrs":{}},{"type":"text","text":"数据采集治理，听上去容易，做起来难，为什么呢？因为要面对的数据类型多，接口和数据类型也不统一。光去适配这些数据，都有可能累的焦头烂额。相对来说，华为AIOps服务首先支持通用的接口，然后对一些常见的设备都已经预置完成，最后能达到自动对接，数据自动治理的一个水平。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"strong","attrs":{}}],"text":"第二层：AI原子能力。","attrs":{}},{"type":"text","text":"华为AIOps共有二十多个原子能力，覆盖检测、预测、识别、诊断四大场景。原子能力不仅仅是AI算法的一个实现。每一个原子能力都经过实际局点数据的检验，针对具体的运营场景做过优化。同时，每一个原子能力也都融入了华为以前的运维经验，某些原子能力甚至能做到不训练可以直接使用。 ","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"strong","attrs":{}}],"text":"第三层：编排能力。","attrs":{}},{"type":"text","text":"包括流程的编排和大屏的编排，还有RPA的编排。原子能力是AIOps智能运维的基础组件，流程编排操作简单灵活，只需从组件库中拖拽数据及AI运维能力进行组合，即可完成命令场景端到端的图形化编排，真正支撑合作伙伴拉低开发门槛，高效率的构建AI应用编排框架。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"strong","attrs":{}}],"text":"第四层：行业AI app。","attrs":{}},{"type":"text","text":"针对最典型的场景开箱即用。通过丰富的2D和3D可视化组件，如提供了超过30个图表控件，覆盖折线、拓扑、列表、柱形等样式，并提供多个地图控件、交互控件及媒体控件搭建。运维效果大屏时只需从组件库里拖拽出各类控件，按需组合自由布局、灵活配置应用的各种报表，辅助监控和分析，例如DIY微服务健康监控大厅，使其能够可视化，展示接口平均成功率、接口平均时延、接口失败率、接口调用次数等。同时提供KPI告警列表，为运营人员提供故障预警参考依据，拖拽所需控件号，对控件的样式，数据及交互进行个性化定制，使其满足展示要求。后端数据还可使用app组合流程里定义的各类中间数据。配置完成后即可一键预览和发布运维效果，大屏展示接口，平均成功率，接口平均时延，接口失败率，接口调用次数等，快速实现DIY可视化大屏。","attrs":{}}]},{"type":"heading","attrs":{"align":null,"level":2},"content":[{"type":"text","text":"（一）RPA助力AIOps对接现有运维系统","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"除了展示位，推理结果必须能够帮助进行故障的恢复。现阶段一般是对接现有的系统，比如工单系统（需要工单邮箱的人要去处理）、自动回复和问题单。如果通过人工去对接，费时费力并且容易出错。因此机器人流程自动化，也就是RPA服务，水到渠成。RPA服务可以完成数据的对接、搬运及工单的发放等等，减少人力投入，降低出错成本。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"image","attrs":{"src":"https://static001.geekbang.org/infoq/f6/f6be5dec2ddc60bc51ee5b8bec8fff9b.jpeg","alt":null,"title":null,"style":[{"key":"width","value":"75%"},{"key":"bordertype","value":"none"}],"href":null,"fromPaste":true,"pastePass":true}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"heading","attrs":{"align":null,"level":2},"content":[{"type":"text","text":"（二）10+开箱即用的App，支持快速部署","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"针对一些最典型的场景，华为云AIOps把编排能力都已经提前准备好，也即，","attrs":{}},{"type":"text","marks":[{"type":"strong","attrs":{}}],"text":"有十多种开箱即用的App","attrs":{}},{"type":"text","text":"，如园区网络、DC网络、IT应用、运营商网络等等场景全覆盖；","attrs":{}},{"type":"text","marks":[{"type":"strong","attrs":{}}],"text":"灵活部署","attrs":{}},{"type":"text","text":"，支持公有云、HCS部署、On Premise部署、及云地协同等；","attrs":{}},{"type":"text","marks":[{"type":"strong","attrs":{}}],"text":"开放生态，","attrs":{}},{"type":"text","text":"支持合作伙伴开发行业App，并将AI应用发布到AI市场，合作共赢，共建网络AI生态。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"下面我们以“KPI异常检测”App来演示一下如何使用一个开箱即用的App。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"第一步：导入网元列表；","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"第二步：配置性能、告警数据源；","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"第三步：数据源关联到App；","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"第四步：启动App；","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"第五步：查看大屏，分析故障。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"image","attrs":{"src":"https://static001.geekbang.org/infoq/f3/f37fb53ea35f848ea1fa758c7e360efb.jpeg","alt":null,"title":null,"style":[{"key":"width","value":"75%"},{"key":"bordertype","value":"none"}],"href":null,"fromPaste":true,"pastePass":true}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"heading","attrs":{"align":null,"level":1},"content":[{"type":"text","text":"AIOps使能园区网络智能运维","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"那么AIOps是如何解决园区中实际运维的呢？","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"heading","attrs":{"align":null,"level":2},"content":[{"type":"text","text":"（一）园区网络建维模式","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"image","attrs":{"src":"https://static001.geekbang.org/infoq/11/11f564be05e1d494afde4425b6dc0e4e.jpeg","alt":null,"title":null,"style":[{"key":"width","value":"75%"},{"key":"bordertype","value":"none"}],"href":null,"fromPaste":true,"pastePass":true}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"​上图为园区网络的两种建维模式：","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"strong","attrs":{}}],"text":"2B和2C共用大网的OMC：","attrs":{}},{"type":"text","text":"当前的主流模式。企业去租用运营商的无线设备及其他的一些设备。这种模式的问题在于，终端由企业维护，网络由运营商维护，那么出现问题的时候很难分清责任；另外一个问题是，运营商侧的运维能力和组织构筑大网2C的O域，难以支撑企业内网高SLA，强化客户诉求。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"strong","attrs":{}}],"text":"2B和2C分开OMC（EMS）：","attrs":{}},{"type":"text","text":"企业采购5G CPE、无线、核心网等全部设备进行维护，具备端到端的视图。从工信部发文、VDF、奥迪园区及企业SLA保障来看，","attrs":{}},{"type":"text","marks":[{"type":"strong","attrs":{}}],"text":"企业租用运营商频谱或专用频谱自建5G网络会逐步成为主流。","attrs":{}}]},{"type":"heading","attrs":{"align":null,"level":2},"content":[{"type":"text","text":"（二）业务场景和痛点分析：园区客户需要简单易用、多域融合的网络运维","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"strong","attrs":{}}],"text":" 1. 典型网络现状","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"image","attrs":{"src":"https://static001.geekbang.org/infoq/67/676d28c3285993a2df215aff2284d162.jpeg","alt":null,"title":null,"style":[{"key":"width","value":"75%"},{"key":"bordertype","value":"none"}],"href":null,"fromPaste":true,"pastePass":true}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"​上图是一个园区比较常见的一个视频检测的业务。我们可以看到，即便是一个最常见的业务，也大概十来个网元都会参与到其中，从5G的无线到传输到边缘计算，甚至是核心网，都会去参与其中。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"strong","attrs":{}}],"text":"2. 园区应用","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"image","attrs":{"src":"https://static001.geekbang.org/infoq/79/79f1cfbe4a41ee377a17b586b6ab31ae.jpeg","alt":null,"title":null,"style":[{"key":"width","value":"75%"},{"key":"bordertype","value":"none"}],"href":null,"fromPaste":true,"pastePass":true}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"​上图列出了园区里面常见的一些应用，包括边缘的AI检测、智能物流、室内定位等。所有的这些业务其实都和上一张图类似，即任何一个简单的业务都要涉及到多个域的参与。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"那么园区与运营商运维的差异是什么呢？主要有以下三点：","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"strong","attrs":{}}],"text":"用户：","attrs":{}},{"type":"text","text":"缺乏专业的通信知识，网络运维能力弱；","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"strong","attrs":{}}],"text":"网络：","attrs":{}},{"type":"text","text":"组网相对简单，但涉及多域、无线、传接、数通、IT等；","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"strong","attrs":{}}],"text":"SLA：","attrs":{}},{"type":"text","text":"生产系统网络端到端SLA合同要求高，7X24小时，99.99%。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"因此，客户如果是园区运维的话，有如下痛点：","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"strong","attrs":{}}],"text":"技能：","attrs":{}},{"type":"text","text":"5G 2B引入使得网络更加复杂，企业工程师缺乏相关技能，运维困难；","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"strong","attrs":{}}],"text":"工具：","attrs":{}},{"type":"text","text":"缺乏有效的运维工具，复杂网络问题定位需要跨域专家现场会诊，成本高，耗时长。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"总结来说","attrs":{}},{"type":"text","marks":[{"type":"strong","attrs":{}}],"text":"，园区网络跨域设备需要实现数据融合，支撑端到端分析及呈现，最终实现企业ICT基础设施的统一运维。而园区网络涉及网络设备多，边界模糊，需要有统一的跨域定界定位能力，加速生产网络问题定位。","attrs":{}}]},{"type":"heading","attrs":{"align":null,"level":2},"content":[{"type":"text","text":"（三）传统人工、工具化运维不能满足园区网络新需求，急需智能化转型","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"image","attrs":{"src":"https://static001.geekbang.org/infoq/66/66f07a0377c339464d7f906a879c2a55.jpeg","alt":null,"title":null,"style":[{"key":"width","value":"75%"},{"key":"bordertype","value":"none"}],"href":null,"fromPaste":true,"pastePass":true}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"​根据上图的数据，我们可以看到：","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"strong","attrs":{}}],"text":"被动式运维：","attrs":{}},{"type":"text","text":"75%的问题都是由用户发现而非主动检测，如果由用户发现，那么用户很可能就会投诉；","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"strong","attrs":{}}],"text":"自动化程度低：","attrs":{}},{"type":"text","text":"企业成本中70%的运营成本属于人力成本，成本激增；","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"strong","attrs":{}}],"text":"故障解决困难：","attrs":{}},{"type":"text","text":"90%故障的恢复时间是用来做问题定位的，真正的问题修复时间占比非常小。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":" ","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"这样看来，","attrs":{}},{"type":"text","marks":[{"type":"strong","attrs":{}}],"text":"无论是从效率还是效果这两方面去考虑，都有一个诉求就是引入人工智能去解决问题，使能网络运维的预测、分析、决策的自动化闭环","attrs":{}},{"type":"text","text":"。","attrs":{}}]},{"type":"heading","attrs":{"align":null,"level":2},"content":[{"type":"text","text":"（四）跨域故障定位算法流程","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"image","attrs":{"src":"https://static001.geekbang.org/infoq/d5/d56664c514ab9b74da9e1ec489bbb2a1.jpeg","alt":null,"title":null,"style":[{"key":"width","value":"75%"},{"key":"bordertype","value":"none"}],"href":null,"fromPaste":true,"pastePass":true}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"​上图是跨域故障定位的算法流程。整个流程如下：","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"strong","attrs":{}}],"text":"输入：","attrs":{}}]},{"type":"bulletedlist","content":[{"type":"listitem","attrs":{"listStyle":null},"content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"告警：设备上报的告警；","attrs":{}}]}]},{"type":"listitem","attrs":{"listStyle":null},"content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"Topo：组网Topo结构；","attrs":{}}]}]},{"type":"listitem","attrs":{"listStyle":null},"content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"故障传播图：告警间的影响关系。","attrs":{}}]}]}],"attrs":{}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"strong","attrs":{}}],"text":"流程介绍：","attrs":{}}]},{"type":"bulletedlist","content":[{"type":"listitem","attrs":{"listStyle":null},"content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"降噪：过滤原始告警中的闪断、震断等数量多又无效告警；","attrs":{}}]}]},{"type":"listitem","attrs":{"listStyle":null},"content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"聚合：对告警进行划分，将Topo不相关的告警分开，可能相关（属于同一故障）的告警聚合到一起，得到多个告警组；","attrs":{}}]}]},{"type":"listitem","attrs":{"listStyle":null},"content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"识别定位：结合Topo、故障传播图，对每个告警组进行分析，识别出每个告警组中有几个故障，每个故障的根因网元和根因告警；","attrs":{}}]}]},{"type":"listitem","attrs":{"listStyle":null},"content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"诊断：对于每个故障告警诊断出故障的类型，例如：电源中断。","attrs":{}}]}]}],"attrs":{}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"strong","attrs":{}}],"text":"输出：","attrs":{}}]},{"type":"bulletedlist","content":[{"type":"listitem","attrs":{"listStyle":null},"content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"故障的根因","attrs":{}}]}]},{"type":"listitem","attrs":{"listStyle":null},"content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"故障设计的告警","attrs":{}}]}]},{"type":"listitem","attrs":{"listStyle":null},"content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"故障类型","attrs":{}}]}]},{"type":"listitem","attrs":{"listStyle":null},"content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"故障恢复建议","attrs":{}}]}]}],"attrs":{}},{"type":"heading","attrs":{"align":null,"level":2},"content":[{"type":"text","text":"（五）AIOps框架实现算法流程","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"以上讲解了整个的算法流程，接下来，我们看看如果使用华为AIOps框架去实现算法流程。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"strong","attrs":{}}],"text":"1、快速配置数据源，编排流程","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"配置数据源：将无线、传输、核心网等多个域的告警接入，接入网络拓扑数据；","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"流程编排：通用已有的原子能力，快速进行流程编排。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"image","attrs":{"src":"https://static001.geekbang.org/infoq/e0/e0020afef040fab0fabe69104dfd1f9a.jpeg","alt":null,"title":null,"style":[{"key":"width","value":"75%"},{"key":"bordertype","value":"none"}],"href":null,"fromPaste":true,"pastePass":true}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"​经过上述过程，可以完成“事件通知”功能，并将结果保存到记录集（即，数据库），用于大屏展示。效果图如下：","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"image","attrs":{"src":"https://static001.geekbang.org/infoq/96/9648304bec083c3e6d8230d2e6aee09f.jpeg","alt":null,"title":null,"style":[{"key":"width","value":"75%"},{"key":"bordertype","value":"none"}],"href":null,"fromPaste":true,"pastePass":true}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"打开其中一条告警，可以看到如下信息：","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"image","attrs":{"src":"https://static001.geekbang.org/infoq/bc/bc7b0928cdde2701d53815b2e965e9eb.jpeg","alt":null,"title":null,"style":[{"key":"width","value":"75%"},{"key":"bordertype","value":"none"}],"href":null,"fromPaste":true,"pastePass":true}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"heading","attrs":{"align":null,"level":1},"content":[{"type":"text","text":"AIOps部署建议","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"根据前述的实践，我们可以总结以下内容：","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"strong","attrs":{}}],"text":"1、选定成熟场景，循序渐进部署AIOps","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"经过长期实践，我们对AIOps部署失败的主要原因做了如下总结：","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"strong","attrs":{}}],"text":"数据上不来：","attrs":{}},{"type":"text","text":"数据分散在各个独立系统之上，缺乏综合采集管理手段。数据缺失，数据质量低下是造成AIOps效果欠佳的主要原因；","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"strong","attrs":{}}],"text":"命令下不去：","attrs":{}},{"type":"text","text":"缺乏自动化运维工具，不能进行主动检测，恢复操作；","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"strong","attrs":{}}],"text":"模型不智能：","attrs":{}},{"type":"text","text":"不能有效的积累日常运维中的标注信息，不能实现模型自学习。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"因此，在部署失败的基础上，我们可以得出，如果要成功部署AIOps，我们需要：","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"从具备条件的","attrs":{}},{"type":"text","marks":[{"type":"strong","attrs":{}}],"text":"成熟场景","attrs":{}},{"type":"text","text":"出发，","attrs":{}},{"type":"text","marks":[{"type":"strong","attrs":{}}],"text":"循序渐进","attrs":{}},{"type":"text","text":"推进AIOps部署；","attrs":{}}]},{"type":"bulletedlist","content":[{"type":"listitem","attrs":{"listStyle":null},"content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"strong","attrs":{}}],"text":"数据上的来","attrs":{}},{"type":"text","text":"，全面收集各种运维数据，提高数据质量；","attrs":{}}]}]},{"type":"listitem","attrs":{"listStyle":null},"content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"strong","attrs":{}}],"text":"命令下得去","attrs":{}},{"type":"text","text":"，AIOps后端对接现在自动运维工具，增强诊断手段和自动恢复能力；","attrs":{}}]}]},{"type":"listitem","attrs":{"listStyle":null},"content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"strong","attrs":{}}],"text":"有效积累标注数据","attrs":{}},{"type":"text","text":"，让AIOps模型能不断收到反馈，具备自学习能力。","attrs":{}}]}]}],"attrs":{}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"strong","attrs":{}}],"text":"2、选择成熟的AIOps服务","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"针对不同类型的企业，AIOps服务的选择也是不尽相同，具体见下表：","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"image","attrs":{"src":"https://static001.geekbang.org/infoq/ac/ac6a831e2e121d12ed105fe3147a3cfa.jpeg","alt":null,"title":null,"style":[{"key":"width","value":"75%"},{"key":"bordertype","value":"none"}],"href":null,"fromPaste":true,"pastePass":true}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"image","attrs":{"src":"https://static001.geekbang.org/infoq/d5/d5559c6065a3a35a79e221e197b5332b.jpeg","alt":null,"title":null,"style":[{"key":"width","value":"75%"},{"key":"bordertype","value":"none"}],"href":null,"fromPaste":true,"pastePass":true}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"​华为AlOps服务降低网络AI应用开发门槛，加速网络AI应用落地。沉淀了10+开箱即用的智能APP，覆盖运营商网络、园区网络、数据中心网络和IT应用等应用领域。预集成丰富的AI原子能力，覆盖故障预测、检测、诊断、识别等环节。支持用户零编码开发AI应用，提升运维效率。","attrs":{}}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"感兴趣就一起来体验一下吧~ ","attrs":{}},{"type":"link","attrs":{"href":"https://www.hwtelcloud.com/products/aio","title":"","type":null},"content":[{"type":"text","text":"https://www.hwtelcloud.com/products/aio","attrs":{}}]}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"link","attrs":{"href":"https://bbs.huaweicloud.com/blogs?utm_source=infoq&utm_medium=bbs-ex&utm_campaign=other&utm_content=content","title":"","type":null},"content":[{"type":"text","text":"点击关注，第一时间了解华为云新鲜技术~","attrs":{}}]}]}]}