G08 - Spatial Database Group

Download Report

Transcript G08 - Spatial Database Group

Jiazhang Liu;Yiren Ding
Team 8
[10/22/13]
1
Traditional Database
• Servers
• Database Admin
• DBMS
Traditional Database
1
• High cost
• Lack of elasticity
• Hard to maintain
2
Introduction
• Relational Cloud: “database-as-a-service” (DBaaS )
• Why is it attractive?
– Hardware and energy cost much lower
– The cost is proportional to actual use (pay-per-use)
• So, how to make Relational Cloud more attractive?
– Efficient multi-tenancy
– Elastic scalability
– Database privacy
3
Efficient multi-tenancy
• Given a set of databases and workloads, what
is the best way to serve them from a given set
of machines?
Efficient multi-tenancy
3
• Solution:
– uses a single database server on each machine
which hosts multiple logical databases.
– Relational Cloud periodically determines which
databases should be placed on which machines
using a novel non-linear optimization formulation.
– a cost model that estimates the combined
resource utilization of multiple databases running
on a machine.
4
Elastic scalability
• A good DBaaS must support database and
work- loads of different sizes.
• The challenge arise when a database workload exceeds the capacity of a single machine.
• Must support scale-out, where responsibility
for query processing is partitioned among
multiple nodes to achieve higher throughput.
5
Database Privacy
• Encrypt all the data stored in the DBaaS.
– privacy concerns would largely be eliminated.
• However, any impact on processing encrypted
data?
5
Database Privacy
– Developed CryptDB: to provide privacy with an acceptable impact on
performance (22.5% reduction in throughput)
6
Database partitioning
• to scale a single database to multiple nodes,
useful when the load exceeds the capacity of a
single machine.
• to enable more granular placement and load
balance on the back-end machines compared
to placing entire databases.
7
Database Placement
• Resource allocation is a major challenge when
designing a scalable, multi-tenant Service like
Relational Cloud.
7
Database Placement
• Resource Monitor
Monitor the resource requirements of each
workload
• Combined Load Predictor
Predicting the load multiple workloads will
generate when run together on a server
• Consolidation Engine
Assigning workloads to physical servers
8
SYSTEM DESIGN
• Relational Cloud Architecture
9
10
Flipboard
• Greg Scallan, Chief Architect at Flipboard says,
“Our service currently runs 100% on AWS in
multiple availability zones.”
• “We chose AWS because they were able to
provide a majority of the solution we needed
as we built our data center. Also, we
appreciated the flexibility as we tried out
various solutions to our business vision.”
11
Sources
• Relational Cloud: A Database-as-a-Service for the
Cloud:
http://people.csail.mit.edu/nickolai/papers/curin
o-relcloud-cidr.pdf
• AWS Case Study:
Flipboardhttps://aws.amazon.com/solutions/case
-studies/flipboard/
• AWS
Documentationhttp://docs.aws.amazon.com/AW
SEC2/latest/UserGuide/concepts.html
12
Questions?