2 of 25

UII Journey

Into the Hybrid Cloud

3 of 25

Welcome to Google DevFest 2019

Nashihun Amien

SRE & CSIRT

Badan Sistem Informasi

Universitas Islam Indonesia

linkedin.com/in/nashihunamien

4 of 25

UII Infrastructure Fact

Total Users: 33.000.

Total datacenter: 1

Total bandwidth: 4.2GBps

Total Access Points: 743

Full FO inter-building connection

5 of 25

Cerita lama

Monolith architecture

30 Apache web server with manual scaling process for 1 application (Load Balancer with HAProxy).

Version 1.0 and always be 1.0

Standalone database

KRS always become obstacle for all civitas UII

6 of 25

Jaman now

Microservices architecture

Semi-Automate autoscale infrastructure

From project to product mindset

Cluster infrastructure

Hyper Converged Infrastructure

First run able to serve 3k concurrent users

7 of 25

Target

30k concurrent user

and

99% SLA

8 of 25

HOW?

9 of 25

Private cloud with Hyper Converged Infrastructure

Hyper Converged Infrastructure (HCI) combines common datacenter hardware using locally attached storage resources with intelligent software to create flexible building blocks that replace legacy infrastructure consisting of separate servers, storage networks, and storage arrays.

11 of 25

Kubernetes in action

Minimum production setup

3,5,7 etcd
2 master
N workers

12 of 25

Provisioning in baremetal

Kubeadm
RKE
Ansible
Kubernetes the hard way
etc . . .

Hey, Minikube is not for production!

14 of 25

We do have 192 cores, 2.95 TB memory and 42.66 TB storage in one cluster.

15 of 25

But We need more compute!

16 of 25

All Hail to GCP (GKE)

17 of 25

Hybrid cloud

Private + Public = PERFECT!

GCP can serve more nodes than on-prem resources.
Provide autoscale provision for for EKS.

18 of 25

Infrastructure target

Target:

Seamless data transfer between infrastructure.

Sync database

Solution:

Interconnection between on-prem and cloud ( Upto 500Mbps )

20 of 25

Summary

Google Cloud

Managed by Google
We can expect minimal downtime
At some point developer can manage the infrastructure by itself
Pay as you go

On-Prem

Self managed datacenter
We need to maintain our datacenter for minimal downtime
You need operational team to make sure everything works
On-prem hardware is expensive

21 of 25

Just another summary

To integrate the on-prem and the gcp, you need to understand network infrastructure. Unless you are going to publish the apps and put under CDN network
Know your apps behaviour and load
Monitoring system is a must
If you are going to build a startup, and you need an infrastructure for your apps, Google Cloud is a good option
Kubernetes can save you from million traffic if you know what to do with it. But it can be hard if you are not understand the concept

22 of 25

Not recommended books

23 of 25

Recommended books

24 of 25