7. #gcconf2019 #GameStackJP
• Choice for How and When You Play
• すでにクインシーのデータセンターにはXboxが搭載
• 近日中に、54データセンターに拡大予定
Microsoft クラウド サービス② | xCloud
8. #gcconf2019 #GameStackJP
Microsoft クラウド サービス③ | Azure
• 常時接続 常時稼働
• 無制限のリソース
• フレキシビリティ
• オープン
• 豊富なサポート
2
regions
in Japan
140
available in
140 countries
54
regions
worldwide
11. #gcconf2019 #GameStackJP
マイクロソフト海底に潜る
太平洋80Tbps
大西洋160Tbps
データセンター間NW
1.4 million miles 敷設
海底DC
United
States
United
States
Canada
Mexico
Venezuela
Colombia
Peru
Bolivia
Brazil
Argenti
na
Atlanta
Ocean
Algeria
MaliNiger
Nigeria
Cha
d
Liby
a
Egy
pt
Sud
an
Ethi
opia
Dr
Cong
oAngol
a
Zambi
a
Namb
ia
South
Africa
Greenla
nd
Svalbar
d
Sweden
NorwayUnited
Kingdo
m
France
PolandUkraine
Turkey
Saudi
Arabia
Iran
Kazakist
an
India
Russia
China
Myanm
ar
(Burma)
Indian
Ocean
Indonesi
a
Australi
a
Pacific
Ocean
Pacific
Ocean
14. #gcconf2019 #GameStackJP
Pearl Abbys : 黒い砂漠
•5 million registrations in one month
•3 million downloads at launch
https://blogs.msdn.microsoft.com/sqlserverstorageengine/2018/07/23/pearl-abyss-massive-scale-using-azure-sql-database/
On February 28, 2018, Pearl Abyss
launched Black Desert Mobile on
Microsoft Azure virtual machines
utilizing 7,500 cores and 46 instances of
Azure SQL Database in Korea Central.
65. #gcconf2019 #GameStackJP
後発だが、頑張っている Azure IaaS
Below the MQ for IaaS on August 2013 Below the MQ for May 2014.
GARTNER MAGIC QUADRANT
FOR INFRASTRUCTURE
AS A SERVICE (IAAS) 2018
81. #gcconf2019 #GameStackJP
Azure Blobs Today High Throughput Blobs Premium Blobs
Limited Public Preview
htbbfeedback@microsoft.com
Central US, East US 2, West US, North Europe (October), East US (October)
http://aka.ms/premiumblob
40X
200X
Massive Azure Blob Scalability
82. #gcconf2019 #GameStackJP
昨年さらに進化した冗長構成の選択肢
• データセンターをまたがる「可用性ゾーン」という考え方が登場
Azure Site Recovery / Region PairsAvailability ZonesAvailability SetsPremium Storage
Zone 2
Zone 3
Zone1
Region 1
Zone 1
Zone 2
Zone 3
Region 2
Data
center 1
Data
center 2
Hardware failure Rack level failure Datacenter outage
Infrastructure
(applicationsanddata)
Single VM Rack 1 Rack 2
Consult SLAs based on
business needs
Select regions based on available services
Acknowledge
shared responsibilities
Educate HA, DR and Backup for
independent reliability needs
Natural Disaster
85. #gcconf2019 #GameStackJP
Availability 99.995%
*AIR: 3
Availability 99.999%
*AIR: 75
*AIR: 15
Availability 99.99%
Nov Storage outage
Downtime
*AIR = Annual Interruption Rate (incl. Planned Maintenance)
Full reboot for every host update
Memory preserving updates introduction (<30sec)
Self-maintenance for low impact in-place updates
Planned Maintenance for decom
Update Using Live Migration
Kernel Soft Reboot Updates
Reliability investments
Platform capability, Customer controllability
Project Tardigrade: VMs survive on most underlying HW or SW failures
Batched updates for ultra-sensitive VMs
Live Migration support extended
Dedicated Hosts
Gest OS update for VMSS VMs
Self-maintenance for rebooting updates
Scheduled Events GA
Live Migration for healing
Project Rainier: Azure Wide effort to reduce reboot count
Live migration from degraded hardware
Impact reduction for OS memory preserving update
Control healing from degraded hardware
Scheduled Events for Live Migration
サーバー性能と同じくらい大事な信頼性