[PPT] - No Reservations: A First Look at Amazons Reserved Instance PowerPoint Presentation

SLIDE 1

Pradeep Ambati, David Irwin, and Prashant Shenoy University of Massachusetts Amherst

No Reservations: A First Look at Amazon’s Reserved Instance Marketplace

SLIDE 2

Cost vs. Risk Tradeoffs in IaaS Clouds

2

Cost (per hour)

Cheap Expensive

Rejection Risk, Non-Revocable Rejection Risk, Revocable Demand Risk, Non-Revocable

Risk

On-Demand Reserved Spot

SLIDE 3

Reserved Instance (RI) Risks

RI’s only cheaper than on-demand if highly utilized
Accurately forecasting demand over long periods is challenging
Unforeseen events like COVID-19 can substantially change demand
Reserved VMs expose users to substantial demand risk
Due to the gap between forecasted and actual demand
To mitigate demand risk…
…Amazon operates the Reserved Instance Marketplace (RIM)

3

SLIDE 4

Reserved Instance Marketplace (RIM)

RIM allows users to sell VM reservations at a price they set
After listing reserved VMs, EC2 posts them on the RIM
Enables purchasing variable and shorter-term VM reservations
AWS only offers 1-year and 3-year reservations on the market

4

SLIDE 5

Comparing RIM with the Spot Market

RIM is a competitive market with multiple buyers and sellers
In contrast, EC2’s spot market has only a single seller (Amazon)
Spot price not set based on instantaneous supply/demand
RIM is akin to the housing market with many unique listings
Listing value defined by a large set of attributes
E.g., VM type, number of instances listed, term duration, supply/demand, etc.
Spot market is akin to the stock market
Uniform pricing of many identical assets (VMs) across regions and AZs

5

SLIDE 6

RIM Pricing

Clearly, RIM’s usefulness is a function of its prices
But, EC2 does not automatically archive RIM price data
We started monitoring and archiving RIM prices in 2018/9
Query RIM using EC2’s python Boto3 API every 30 minutes
Collect data for all VM types in 69 AZs across 22 regions
Publicly released data at UMass Trace Repository
http://traces.cs.umass.edu/index.php/Main/Cloud

6

SLIDE 7

A First Look at RIM Data

Analyze RIM data from 2018/9 to 2020/5
Reveal important market characteristics
Identify potential reserved VM optimization opportunities
Key Market Characteristics
Market volume
VM type
Term duration
Time-on-market
Comparison with on-demand and spot

7

SLIDE 8

Market Volume across Regions

us-east-1 and us-west-2 are largest regions

8

400 800 1200 1600

u s

e

a s t

1

u s

w

e s t

2

u s

e

a s t

2

e u

w

e s t

1

e u

c

e n t r a l

1

u s

w

e s t

1

a p

n
r

t h e a s t

1

s a

e

a s t

1

a p

s
u

t h e a s t

1

0K 6K 12K 18K 24K

Average Market Volume (#instances listed)

Average Market Volume (#ECUs listed)

Region

Instances ECUs

SLIDE 9

Market Volume across Regions

Lower market volume may increase risk of using RIM

9

400 800 1200 1600

u s

e

a s t

1

u s

w

e s t

2

u s

e

a s t

2

e u

w

e s t

1

e u

c

e n t r a l

1

u s

w

e s t

1

a p

n
r

t h e a s t

1

s a

e

a s t

1

a p

s
u

t h e a s t

1

0K 6K 12K 18K 24K

Average Market Volume (#instances listed)

Average Market Volume (#ECUs listed)

Region

Instances ECUs

SLIDE 10

Market Volume by VM Type

c4.large and c5.large are most popular VM types

10

250 500 750 1000

c4.large c5.large m4.large i3.xlarge c5.4xlarge r5.large r4.large c5d.large m5.2xlarge c5d.9xlarge

0K 2K 4K 6K 8K

Average Market Volume (#instances listed)

Average Market Volume (#ECUs listed)

Instance Type

Instances ECUs

us-east-1

SLIDE 11

Market Volume by VM Type

RIM lagging indicator of instance type popularity

11

250 500 750 1000

c4.large c5.large m4.large i3.xlarge c5.4xlarge r5.large r4.large c5d.large m5.2xlarge c5d.9xlarge

0K 2K 4K 6K 8K

Average Market Volume (#instances listed)

Average Market Volume (#ECUs listed)

Instance Type

Instances ECUs

us-east-1

SLIDE 12

Listing Volume vs. Term Duration

Short and long term durations more plentiful

12

0K 3K 6K 9K

1

3

4

6

7

9

1

1

2 1 3

2

4 2 5

3

6

0K 20K 40K 60K 80K 100K

#Instances Listed

#ECUs Listed

Duration (months)

Instances ECUs

us-east-1

SLIDE 13

Listing Volume vs. Term Duration

Discrepancy at 25-36 months may indicate buyer’s remorse

13

0K 3K 6K 9K

1

3

4

6

7

9

1

1

2 1 3

2

4 2 5

3

6

0K 20K 40K 60K 80K 100K

#Instances Listed

#ECUs Listed

Duration (months)

Instances ECUs

us-east-1

SLIDE 14

Listing Price (c4.large) vs. Term Duration

14 2 4 6 8

1-3 4-6 7-9 10-12 13-24 25-36

1Y Reservation Price 3Y Reservation Price

Price (cents/hr) Duration (months)

Minimum Price Average Price

Average listing price decreases (nearly) linearly

us-east-1

SLIDE 15

Time-on-the-Market across Regions

Largest regions have the longest average time on the market

15

30 60 90

u s

e

a s t

1

u s

w

e s t

2

u s

e

a s t

2

e u

w

e s t

1

e u

c

e n t r a l

1

u s

w

e s t

1

a p

n
r

t h e a s t

1

s a

e

a s t

1

a p

s
u

t h e a s t

1

Average Time on the Market (Days)

Region

SLIDE 16

Time-on-the-Market across Regions

Indicates that demand is less relative to the supply

16

30 60 90

u s

e

a s t

1

u s

w

e s t

2

u s

e

a s t

2

e u

w

e s t

1

e u

c

e n t r a l

1

u s

w

e s t

1

a p

n
r

t h e a s t

1

s a

e

a s t

1

a p

s
u

t h e a s t

1

Average Time on the Market (Days)

Region

SLIDE 17

Time-on-the-Market by VM type

Popular VM types have longest average time on the market

17

30 60 90 120 150

c4.large c5.large m4.large i3.xlarge c5.4xlarge r5.large r4.large c5d.large m5.2xlarge c5d.9xlarge

Average Time on the Market (Days)

Instance Type

us-east-1

SLIDE 18

Comparing VM Purchasing Options

Users neither lose nor make money on average in the RIM

18

0.025 0.05 0.075 0.1 Reserved (1yr) RIM (<1yr) Reserved (3yr) RIM (>=1yr) Spot Ondemand

Effective Price ($/hr) Purchasing Option

us-east-1

SLIDE 19

Comparing VM Purchasing Options

RIM listings (>=1yr) price close to spot

19

0.025 0.05 0.075 0.1 Reserved (1yr) RIM (<1yr) Reserved (3yr) RIM (>=1yr) Spot Ondemand

Effective Price ($/hr) Purchasing Option

us-east-1

SLIDE 20

Conclusion

Provides a first look at Amazon RIM data
Analyzed RIM data from 2018/9 to 2020/5 to reveal key market characteristics
Publicly released data at UMass Trace Repository
http://traces.cs.umass.edu/index.php/Main/Cloud
Future work
More in-depth analysis of RIM data
Use RIM data to mitigate demand risk and optimize long-term cloud costs

20

SLIDE 21

Email: lambati@umass.edu

21