A Year in Google Cloud

把现有的基础设施带到谷歌云上是一个挑战,我们的小团队以惊人的速度完成了这项任务。我们在那次行动中幸存下来,并且留下了伤疤来证明这一点。我们将回顾一下我们是如何做到这一点的,以及我们在数据中心之外所面临的挑战。在本课中,我们将讨论我们为高可用性、灾难恢复解决方案和数据库托管所做的决策。
CloudSQL与自管理实例 IOPS限制
代理SQL 培根储蓄和我们曾经遇到的问题。
IP别名 灾难恢复解决方案
*更多!

展开查看详情

1.Our Year in Google Carmen Mason :: VitalSource Technologies Allan Mason :: Pythian Group, Inc.

2.Who we are Allan and Carmen Mason 2

3.Digital learning platforms that drive outcomes. 3

4.We Provide Solutions For: Campus Publishers Institutions Executive Associations & Corporate Stores Education Certifying Bodies Learning 4

5.Key Statistics: Last 12 Months 15 Million 7 Thousand 22 Million STUDENTS SERVED INSTITUIONS BOOKS AND COURSES DELIVERED 5

6.6

7.VitalSource Global Scale

8.HELPING BUSINESSES BECOME MORE DATA-DRIVEN © The Pythian Group Inc., 2018 8

9. PYTHIAN A global IT company that helps businesses leverage disruptive technologies to better compete. Our services and software solutions unleash the power of cloud, data and analytics to drive better business outcomes for our clients. Our 20 years in data, commitment to hiring the best talent, and our deep technical and business expertise allow us to meet our promise of using technology to deliver the best outcomes faster. © The Pythian Group Inc., 2018 9

10. 20 Years in Business 400+ Pythian Experts 350+ Current Clients in 35 Countries Globally © 2017 Pythian. Confidential 10

11. We’re Hiring! https://www.pythian.com/careers/ © 2018 Pythian. Confidential

12.Agenda Things we’ll cover today: • Motivations • Considerations that drove us • Decisions we faced • Our current stack • What’s next 12

13.Motivations

14.Local Data Center HA Configuration Solution Provides • Saved Binary Logs • Differential Relay Logs • VIP Handling • Notification 14

15.Failover in the Data Center • 10 – 20 seconds • Automatically assigns most up to date replica as the new master • VIP tied to CNAME • Replicas are slaved to the new master • A CHANGE MASTER statement is logged to bring the old master in line once it’s fixed. • Alerts by email, and notifications in Slack, with decisions that were made during the failover. 15

16.Ownership • DNS – Managed by Parent Company’s Windows team. • Load Balancer – Managed by Parent Company’s Network team. • App Servers – VMs managed by Parent Company’s VMware team. • Covers – An expensive Isilon managed by Parent Company’s Storage team. • All our stuff – Network managed by Parent Company’s Network team. 16

17.A Lazy Day in May 17

18.“Firemen just ran into our datacenter… with a hose.” 18

19.Do things the right way 19

20.Decisions

21.Why The Cloud? • Flexibility to scale • “Always on” mentality • Easy to manage 21

22.22

23.Why Google Cloud Platform • Google Cloud Platform regions are connected using Google’s private network. • They wrote Kubernetes; they support it well. • Excellent, hands on support. • Better stability and performance, in our experience. 23

24. Cost Comparison Cloud Provider Instance Type vCPU RAM SSD Max IOPS Max Cost/mo. Throughput n1-standard-16 16 60G 1024G 32,000 based on 480 IO block $ 562.44 cores storage sizes of 256 KB € 489.88 25,000 based on vCPU m4.4xlarge 16 64G 1024G 30,000 based on 320 IO $ 2,368.11 cores storage. € 2.062,68 as of 2018-10-22 24

25.Google Compute Engine vs Google CloudSQL

26.Self Managed Instances vs. Google CloudSQL Google Compute Engine CloudSQL (GCE) 26

27. Self Managed Instances: The Pros • Customizable instances • Full control of the OS and installed packages • MySQL built in max values on variables • EG. Max Connections • Choose Percona Server, MariaDB, etc. 27 Google Compute Engine

28. Self Managed Instance: The Cons • We manage the backups • We manage patches and upgrades • We manage failovers • Requires more time on Systems Administration • Leaving less time for Database work 28 Google Compute Engine

29. CloudSQL: The Pros • Routine patching • Automatic failovers • No resource management needed. 29 Google CloudSQL