2. Developing High Performance and Scalable ColdFusion Applications
Using Terracotta Ehcache
By: Shailen Prasad
Product Management & Strategy
In-Memory Computing
Software AG USA, Inc.
3. What is this?
Growth will happen when efficiency increases. Plow did what
"In-Memory computing is doing to the human race today"
5. What will be covered in this presentation?
How to scale – options (pros and cons)
Caching basics (various options available)
Recent updates of Open source Ehcache project.
Ehcache, Terracotta OSS and BigMemory
The benefits of distributed caching for building applications where latency and performance is
crucial.
Advance caching techniques for scaling your current CF application using Terracotta distribution
caching (BigMemory)
To conclude, highlights on some customer use cases where caching was mission critical
In-Memory caching and data management is becoming mainstream for accelerating
business applications. This session will introduce :
6. SCALE OUT
SCALEUP
Scale your existing application
Also knowns as “Vertical Scaling”
Pros:
• Less power consumption than
running multiple servers
• Generally less challenging to
implement
• Less licensing costs
Cons:
• $$$ VERY EXPENSIVE
• Greater risk of hardware failure
causing bigger outages
• Generally severe vendor lock-in
and limited upgradeability in the
future.
Also knowns as “Horizontal Scaling”
Pros:
• $ Much cheaper than scaling
vertically
• Easier to run fault-tolerance
• Generally easier to upgrade
Cons:
• Bigger datacenter foot print
• Higher power consumptions
• Possibly more network resource
dependency
AddingmoreRAM/CPU
Adding more machines
7. In-Memory Caching – In a nutshell
On Heap Cache (also known as In-Process Cache) or L1 cache
ColdFusion Application
CF Internals
On Heap
Cache
JVM
• Fastest among all other caching tier
• Doesn’t require marshalling and un-marshalling of the
data
• Limited by Max JVM heap size (limited on 32 bit systems)
• Garbage collection – still remains the challenge
8. Caching – In a nutshell
Local Off Heap Cache (in-process caching) or L1 cache
ColdFusion Application
CF Internals
Local On-heap
Cache
JVM
Local Off-heap Cache
Direct memory buffers
• Access Memory outside of Application heap
• Can Scale as it is not limited by JVM heap size even on a 32
bit machine
• Application doesn’t have to worry about Garbage collection
for the data stored in off-heap
• Slower than on-heap cache, entries has to serialized and de-
serialized.
9. JVM
Local Off-heap Cache
Caching – In a nutshell
Distributed Off Heap Cache (also known as Out-of-Process
Cache) or L2 cache
ColdFusion
Application
CF Internals
Local On-heap
Cache
JVM
L1Off-heap Cache
JVM
L2 Off-heap Cache
• Runs outside of the Application Server JVM
• Slower than local offheap – reads/writes are over the
network
• Highly scalable with its distributed design
• Adds resiliency with more fault tolerance
13. Open Source
Current in CF2016 - Ehcache 2.10.0
Ehcache 3.x (complete overhaul with lots of improvements!!
Revamped API that leverages Java generics and simplifies Cache interactions
Full compatibility with javax.cache API (JSR-107)
Offheap storage capabilities, including offheap only caches
Out of the box Spring Caching and Hibernate integration thanks to the javax.cache support
Significant improvement in performance over all its previous versions
And many more ... (more at www.ehcache.org)
14. 90% of Data in
Memory
MODERNIZE Database
90% of Data in
Database
Memory
App Response Time
Milliseconds
App Response Time
Microseconds
24. Terracotta 4.x Open Source Offering/Architecture
Standard Java
Proven TBs scale capacity
Not managed by the JVM
No Garbage Collections
Predictable latencies
No specialized appliance needed
29. Any kind of data can be stored in Terracotta BigMemory to make
an application speed up & scale out
Terracotta BigMemory has different tiers for data storage
that can be configured based on an application’s
requirements
Terracotta In-Memory Data grid - Rich data storage & Access
39. Terracotta OSS Setup is just a few steps…
1. Download Terracotta OSS (latest terracotta-4.3.2.tar.gz) at
http://www.terracotta.org/downloads/open-source/catalog
2. Extract to the location of your choice
3. Ensure JAVA_HOME is set
4. Navigate to <TERRACOTTA_INSTALL>/server/bin
5. Start with default single node config by executing:
start-tc-server.sh (or .bat)
6. Terracotta process is now accessible at IP:9510
Note: If setting up Terracotta in active/mirror setup, tc---config.xml must be created and referenced
at startup:
start-tc-server.sh (or .bat) -f <path-to-config>/tc-config.xml –n <server-name-to-start>
40. Connecting ColdFusion to Terracotta in few steps…
1. Copy Ehcache + Terracotta libs to <CF_HOME>/cfusion/lib
<TERRACOTTA_INSTALL>/apis/ehcache/lib/ehcache-2.10.1.jar
<TERRACOTTA_INSTALL>/apis/toolkit/lib/terracotta-toolkit-runtime-4.3.2.jar
2. Add terracotta-specifics configurations in CF ehcache configs:
<CF_HOME>/cfusion/lib/ehcache.xml
<CF_HOME>/cfusion/lib/auth-ehcache.xml
3. Restart CF
4. Notice Terracotta connection in CF logs
41. All the below CF caches are now in Terracotta
• CF Authentication:
auth---ehcache.xml: authcache, authtokenmappingcache
• Internal Caching (CF templates, component paths)
• <cfcache> - Cache fragments of html
• <cfquery> - Cache DB calls
<cfquery name="myAccount“ cachedwithin=#createTimeSpan( 0, 1, 0, 0 )#>
• ORM with Ehcache 2nd level caching: Caching Hibernate queries
Entityload('BlogPost',{},{cacheable=true})
• CF Cache functions: Direct Ehcache calls
CacheGet / CachePut / CacheRemove / CacheGetAllIds
CacheGetMetadata
CacheGetProperties / CacheSetProperties
• Custom CF JAVA components using Ehcache library directly
42. Still so much used and Distributed
The Picture says it all !!! – The power of distributed In-Memory computing.
44. Success Stories: Fortune 500 online payment processor
Radically Improving Profitability With Better, Faster
Fraud Detection
SPEED
What they wanted
Before BigMemory
• Dramatically boost bottom-line profit through faster, more accurate
fraud detection
• Lost 30 cents on every $100 to fraud
• With Oracle Exadata, failed to meet 800 ms
SLA around 10% of time
• Limited to 50 rules, even though each new rule generated
$12 million in profit
45. Success Stories: Fortune 500 online payment processor
SPEEDSPEED
After BigMemory:
Savings of tens of millions of dollars in reduced costs from missed
SLAs and fraudulent charges
Meeting stricter 650-millisecond SLA 99% of time
Savings of $1 million annually in reduced database licenses
Plans to expand from 4TB to 150TB for new applications and to
achieve 250 millisecond SLA
47. “The team began almost immediately to
cache the data. The result was encouraging:
the site's overall response time--the time it
took a page to load--dropped on the evening
of Oct. 22 from eight seconds to two. That
was still terrible, of course, but it represented
such an improvement that it cheered the
engineers. They could see that HealthCare.gov
could be saved instead of scrapped.”
Success Stories: Healthcare.gov
48. Challenges
• All data access to backend database (many round-trips)
• 10+ seconds response times
• Numerous down-times due to concurrent users
Benefits
• Provide in-memory data access such as subscriber data and provider
comparison information
• Session replication of user profile info
• Performance & Scalability
Success Stories: Healthcare.gov
49. App
Server
App Server
App Server
App ServerApp Server
Ehcache
App Server
App Servers
App Server
Ehcache
Security Gateway
“Presentation Zone” “Application Zone”
App Server
App Servers
JMS
Individual
& Families
Issuers
3rd Parties
(B2B)
SOR