Welcome message from author
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
2
Analyzing big databases on CPUs is expensive and complex (and slow)
SQream DB is software that runs
big database analytics on standard hardware with GPUs
3
Big Data Challenges of Enterprises
FAST INSERTSWe generate so much data, our databases struggle with the speed of events
TOO MUCH DATAWe have almost 10x-100x more data than we had 2 years ago
COSTS TOO MUCHWe have less money to spend on big data nowadays
TOO SLOWOur solutions are way too slow to do everything we need, so we compromise
SO COMPLEXThese new technologies are a real pain to work with and hirepeople for
4
Analytics is your competitive edgeIt use to be a few terabytes, and analytics were
more around operational activity.Today the data grows to 10x TBs to 100x TBs and analyzing it
effectively is key for businesses to survive and to compete.
Hundreds of TBSometimes even petabytes of data, coming in at a rate of a
few terabytes per day
Up to 10TB
Up to 1-4TB
2013 2014-2015 2016
5
SQream is software that runs a GPU based SQL Big Database
Scale from TB to PB with ease
Patented technologies
Massively Parallel enginebuilt from scratch
Supercomputer on a chip –Nvidia GPU with 5000 cores
1000x more compute per cm2
6
SQream solves the Big Data Challenges of Enterprises
Grows with your dataSQream’s architecture is designed to be extremely scalableIt easily grows up to hundreds of TB
GPU based DatabaseWith standard hardware beefed up by powerful
Nvidia GPUs, SQream can do 1000x more compute per cm2, at significantly reduced costs.
Supercomputer capabilitiesExtremely fast, near real-time query response
times and unparalleled ingestion speeds
SimplicityEasy, simple and familiar SQL syntax, with industry standard BI connectors ( JDBC, ODBC)
7
Analyzing big data on CPUs is expensive, complex (and slow)Benchmark performed by a leading Telco:
Analyze subscriber network usage - 4.3 billion call records, 30 million subscribers
875
14
168
550
602
1
2
71
Teradata EDWFull 42U rack
SQream DBDell R720
Query time
Logical Servers
CPUs
Total time
37% faster querying
Less hardware
Power saving
Time efficient
SQream’s benefits
8
35 mins500 GB of uncompressed data
Indexes required
4 mins8.75x faster!
No indexing
Self Organizing Network (SON) optimizations Analyzing CDR rollup
Use Case – Cellcom
vs
Preparation CycleCDR download, mediation and insert into SQream DB
Almost 9x faster Less DBA work necessary 4:1 compression ratio
9
Vending machine use-case
• 6000 machines, generating numerous telemetry events per hour
• Transmission via GPRS/3G/LTE to central SQream DB server
• Near real-time analysis on over 100,000,000 rows, spanning nearly 4 years of operation.
10
HLS - Geospatial Analysis
SQream Technologies Ltd - Confidential
Load and analyze constantly streaming data from a UAV for detection of
object movement - geospatial queries.
(point in polygon, distance calculation, line-crossings)
57
25
85
120
0
20
40
60
80
100
120
140
1 2
Tim
e (m
illis
eco
nd
s)
Insertion and querying time for ~50,000 entries per sec.
Series1 Series2
SQream is
4.8x faster
300
5000
0
1000
2000
3000
4000
5000
6000
1
Tim
e (m
illis
eco
nd
s)
Geospatial querying of 1.6 billion entries
Series1 Series2
SQream is
16x faster
11
Homeland SecurityReal time threat detection
Limited field conditions
Sensor fusion - Millions of entries every minute
Location based threat analysis –airports, city centers, borders, …
Act on threats before they act on you
12
Field-level threat detection
Seconds
Hours
Real time, location-based threat detection
Big Database in a suitcaseHighest density compute per cm2
High precision,Real-time performance
Multi-sensor fusion
13
SIEM - Results• Analysis period extended from
single digit days to one year +
• Ingestion of 1M generated events every 10 minutes
• Queries run 10x-80x faster compared to ArcSight
• 8 TB of events data per year (47 Billion entries)
BeforeSQream DB
WithSQream DB
14
Extend analytics period from few days to many months
40s 4s
ArcSight (HP SIEM) SQream DB
Query time1 day of data
Much faster results –10X performance
SQream’s benefits
*** NOT POSSIBLE *** 20sQuery time
1 year of dataNew analysis capability
not previously achievable
946s 14sQuery time
1 month of dataSQream gives higher
resolution with approx. 65x better performance
All three graphs are not to scale
Extend ArcSight (HP SIEM) with faster, larger retention capabilities
15
SQream and Sheba medical centercut cancer cure research time
from years to weeks
200 GBAverage size of a single human genome sequencing
2 MonthsTime it takes a genome researcher to compare a handful of sequences
1 PBThe amount of storage needed by Sheba genome research institute
2 HoursTime it takes a researcher to compare up to hundreds of sequences with SQream
x100Factor of improvement over existing methods
16
Help more cancer patients, faster
The manual methodComparing a handful of samples
The SQream methodComparing many thousands of samples concurrently
Related Documents
