slide1 n.
Download
Skip this Video
Loading SlideShow in 5 Seconds..
@ andy_pavlo PowerPoint Presentation
Download Presentation
@ andy_pavlo

Loading in 2 Seconds...

play fullscreen
1 / 27

@ andy_pavlo - PowerPoint PPT Presentation


  • 94 Views
  • Uploaded on

On Predictive Modeling for D istributed D atabases. @ andy_pavlo. VLDB - August 28 th , 2012. Databases?. Evan Jones?. Putin is going to get re-elected!. Romney has a Swiss bank account!. Muammar Gaddafi is in trouble!. High-Volume. Transaction Processing. z. ,. ?.

loader
I am the owner, or an agent authorized to act on behalf of the owner, of the copyrighted work described.
capcha
Download Presentation

PowerPoint Slideshow about '@ andy_pavlo' - brita


An Image/Link below is provided (as is) to download presentation

Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author.While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server.


- - - - - - - - - - - - - - - - - - - - - - - - - - E N D - - - - - - - - - - - - - - - - - - - - - - - - - -
Presentation Transcript
slide1

On Predictive Modeling for

Distributed Databases

@andy_pavlo

VLDB - August 28th, 2012

slide2

Databases?

Evan Jones?

slide3

Putin is going to get re-elected!

Romney has a Swiss bank account!

Muammar Gaddafi is in trouble!

slide5

High-Volume

Transaction Processing

z

,

?

slide6

Main Memory•Parallel•Shared-Nothing

H-Store: A High-Performance, DistributedMain Memory Transaction Processing SystemProc. VLDB Endow., vol. 1, iss. 2, pp. 1496-1499, 2008.

slide7

Fast

Repetitive

Small

slide8

Transaction

Execution

Proc. Name

Input Params

Transaction

Result

Client

Database Cluster

Database Cluster

slide9

?

P1

P2

P3

P4

?

?

Client

Database Cluster

Database Cluster

slide11

Pro Tip:

Canadians do notlike unnecessary surgeries.

slide12

Main Idea:

Use models to predict transaction behavior before execution.

On Predictive Modeling for OptimizingTransaction Execution in Parallel OLTP SystemsProc. VLDB Endow., vol. 5, iss. 2, pp. 85-96, 2011.

slide13

Client

Database Cluster

Database Cluster

slide15

Step #1:

Estimate the path

that the transaction will take.

slide16

Current State

Input Parameters:

w_id=0

i_w_ids=[0,0] i_ids=[1001,1002]

?

?

GetWarehouse:

SELECT * FROM WAREHOUSEWHERE W_ID = ?

slide17

Step #2:

Determine which optimizations to enable in the DBMS.

slide18

Input Parameters:

+1

w_id=0

i_w_ids=[0,0] i_ids=[1001,1002]

+1

Optimizations:

  • Best Partition?
  • Touched Partitions?

+1

  • Finished Partitions?

+1

+1

slide19

Current State

Input Parameters:

w_id=0

i_w_ids=[0,1] i_ids=[1001,1002]

?

?

CheckStock:

SELECT S_QTY FROM STOCKWHERE S_W_ID = ?AND S_I_ID = ?;

?

InsertOrder:

X

INSERTINTO ORDERS

(o_id, o_w_id)

VALUES (?, ?);

slide21

w_id=0

i_w_ids=[0,1] i_ids=[1001,1002]

HashValue(w_id)

=0

=1

ArrayLength(i_w_ids)

ArrayLength(i_w_ids)

=1

=2

=1

=2

slide22

Experimental

Evaluation

slide23

94.9%

+1.86%

95.0%

+1.17%

90.2%

+8.15%

slide24

(txn/s)

Houdini

Assume Single-Partitioned

TATP

TPC-C

AuctionM

+57%

+126%

+117%

slide25

Conclusion:

Scaling your OLTP DBMS must come from within.

slide26

http://hstore.cs.brown.edu

https://github.com/apavlo/h-store