Scribd
Upload
40 views

Trill: A High-Performance Incremental Query Processor For Diverse Analytics

The document describes Trill, a high-performance incremental query processor for diverse analytics. Trill provides fast streaming analytics that is 2-4 orders of magnitude faster than traditional systems. It supports a range of scenarios including real-time, offline, and progressive queries. Trill also enables strong integration with programming languages and cloud services through its library implementation and code generation techniques.

Uploaded by

Dot-Insight
Copyright
© © All Rights Reserved
Available Formats
Download as PPTX, PDF, TXT or read online on Scribd
40 views

Trill: A High-Performance Incremental Query Processor For Diverse Analytics

The document describes Trill, a high-performance incremental query processor for diverse analytics. Trill provides fast streaming analytics that is 2-4 orders of magnitude faster than traditional systems. It supports a range of scenarios including real-time, offline, and progressive queries. Trill also enables strong integration with programming languages and cloud services through its library implementation and code generation techniques.

Uploaded by

Dot-Insight
Copyright
© © All Rights Reserved
Available Formats
Download as PPTX, PDF, TXT or read online on Scribd
You are on page 1/ 15

Trill: A High-Performance Incremental

Query Processor for Diverse Analytics


Badrish Chandramouli, Jonathan Goldstein, Mike Barnett, Rob DeLine,
Danyel Fisher, John C. Platt*, James F. Terwilliger, John Wernsing

Microsoft Research
Contact: [email protected] Twitter: @badrishc

*
Current affiliation: Google. This work was performed at Microsoft Research.
Diverse Scenarios for Analytics
• Real-time
• Monitor app telemetry (e.g., ad clicks) &
raise alerts when problems are detected
Real-Time

• Real-time with historical Dashboard

• Correlate live data stream with historical


activity (e.g., from 1 week back) Engine
+ Fabric
• Offline
• Develop initial monitoring query using logs
• Back-test monitoring query over historical logs

• Progressive
• Non-temporal analysis (e.g., BI) over large dataset,
stream data, get quick approximate results
Interactive Query Authoring
Three Key Requirements
• Performance
Scenarios
• High throughput: critical for large offline datasets
• Low latency & overhead: Important for real time monitoring • monitor
telemetry & raise
alerts
• Fabric & language integration • correlate real-
• Cloud app/service acts as driver, uses the analytics engine time with logs
• develop initial
• Need rich data-types, integrate custom logic seamlessly
monitoring query
• back-test over
historical logs
• Query model • offline analysis
• Need to support real-time and offline data, temporal and (BI) with early
results
relational queries, early results for exploratory offline queries
Trill: Fast Streaming Analytics Library
• Performance
• 2-4 orders of magnitude faster than traditional SPEs
• For relational queries, comparable to best DBMS
• User-controlled latency specification
• explicit latency vs. throughput tradeoff

• Fabric & language integration


• Built as high-level language (HLL) library component
• Works with arbitrary HLL data-types & libraries

• Query model
• Extended LINQ syntax based on tempo-relational query model
• Supports broad & rich analytics scenarios (relational, progressive, time-based)
Trill’s Use Cases
• Azure Stream Analytics
Cloud service
• With Scope for Bing Ads
• With Orleans for Halo game
monitoring & debugging
• …

• Key enabler: performance +


fabric & language integration +
query model
Example (simplified)
• Define event data-type in C#
struct ClickEvent { long ClickTime; long User; long AdId; }
• Define ingress Application time Latency specification
var str = Network.ToStream(e => e.ClickTime, Latency(10secs));
• Write query (in C# app) Lambda expression
var query =
str.Where(e => e.User % 100 < 5)
.Select(e => { e.AdId }) Grouping key selector
.GroupApply( e => e.AdId,
Operator s => s.Window(5min).Aggregate(w => w.Count()));

• Subscribe to results Grouped sub-query


query.Subscribe(e => Console.Write(e)); // write results to console
Latency-Throughput Spectrum
• Data organized as stream of batches
• Purely physical (no impact on query results)

…
𝑜𝑝 2
• Users specify latency constraint
𝑜𝑝 1
(10 secs)
…

• Batch up to 10 secs of data


…
• Small batches  low latency
• Large batches  high throughput
• More load  larger batches  better throughput
+ Columnar
• Columnar format within each batch

…
• Timestamps as arrays
• Bitvector to indicate row absence
class DataBatch { 𝑜𝑝 2
long[] SyncTime;
...
Bitvector BV;
𝑜𝑝 1 …

}
• One array per payload field
…
class UserData_Gen : DataBatch {
long[] c_ClickTime;
long[] c_User; timestamp payload columns
long[] c_AdId; bitvector
}
• Batch classes are generated & compiled on-the-fly (under the hood)
• Enables efficient QP & serialization
+ Fabric & Language Integration Application

• But, user view is row-oriented Receive(results)

• Dynamically generate and compile HLL code


Trill
for operators
• Our example: filter (where) On(Batch b) {
• User writes str.Where(e => e.User % 100 < 5) for i = 0 to b.Size {
if !(b.c_User[i]%100 < 5)
set b.bitvector[i]
}
• General technique }
next-operator.On(b)

• Generate tight loops over batches, with inlined


expressions (done using reflection)
• Avoid method calls within loops
• Timestamps are columns – accessed only if needed
• Pointer-swing where possible Send(events)
...
str.Where(e => e.User % 100<5);
See Paper …
• Property-based operator codegen specialization
• Grouped & batched operator algorithms
• Library-mode & multi-core scheduler
• Efficient serialization & string support
• Rich query language
• SQL queries with progressive (early) results
• Temporal queries
• Sliding, hopping, tumbling, data-dependent session windows
• Temporal joins, set difference
• Powerful high-perf expression-based user-defined aggregate framework
Evaluation (sample)
• Pre-loaded datasets in main memory
• 16-core machine
• Streaming filter
Evaluation (sample)
• Pre-loaded datasets in main memory
• 16-core machine
• Relational queries
Evaluation (sample)
• Pre-loaded datasets in main memory
• 16-core machine
• Temporal queries
Conclusions
• Trill is a fast & expressive library for analytics
• 2-4 orders-of-magnitude faster than traditional streaming engines
• Comparable to columnar databases for offline SQL queries
• But with progressive (early result) support
• Expressive query language and support for HLL type-system & code
• Library that can be easily embedded in a variety of settings (distributed fabrics,
servers, Cloud applications, devices, …)

• Trill is being used across diverse analytics scenarios


• More info @ http://aka.ms/trill
© 2013 Microsoft Corporation. All rights reserved. Microsoft, Windows, Windows Vista and other product names are or may be registered trademarks and/or trademarks in the U.S. and/or other countries.
The information herein is for informational purposes only and represents the current view of Microsoft Corporation as of the date of this presentation. Because Microsoft must respond to changing market conditions, it should not be interpreted to be a commitment on
the part of Microsoft, and Microsoft cannot guarantee the accuracy of any information provided after the date of this presentation. MICROSOFT MAKES NO WARRANTIES, EXPRESS, IMPLIED OR STATUTORY, AS TO THE INFORMATION IN THIS PRESENTATION.

You might also like