Citrix Netscaler Technical Overview

1
Citrix NetscalerTechnical Overview
Johann Wimmer Sales Manager App. Networking
Group Citrix Systems Central Europe
2
Only Citrix has the best application delivery
solution for all app types
LAN
Branch
Mobile User
3
NetScaler is proven in the worlds most demanding
environments
75
6,000
Enterprise Deployments
Of Internet Users
4
Trends driving the business problem
5
Traditional approaches miss the real issue
Latency?
6
Identifying the real problem
Internet
Enterprise Data Center
Lots of Round Trips
7
Todays approach too complex!
SSL Acceleration
Compression
Application server
Cache
Web server
Application server
Web server
L4-7 Traffic Mgr
DDoSProtection
VPN Gateway
Clear text
Web server
Application server
Encrypted
Too complex cannot solve complete problem
8
Application delivery solution
Custom Web apps
NetScaler
Web Servers
Application / Database Servers
Built to deliver applications
9
How NetScaler solves the problem
Internet
Enterprise Data Center
Lots of Round Trips
10
How we do it
15x Faster
11
Real enterprise applications are
complex
CGI/ASP/JSP
EJB Servlets SOAP
HTTP
SQL JDBC
Web Servers
SQLNet XML/SOAP
App Servers
SNA
Telnet SSH
MAPI
NetBios CIFS
Application Users in Multiple Locations
12
Web does not always equal Applications
CGI/ASP/JSP
EJB Servlets SOAP
HTTP
SQL JDBC
Web Servers
SQLNet XML/SOAP
App Servers
SNA
Telnet SSH
MAPI
NetBios CIFS
Application Users in Multiple Locations
13
NetScaler accelerates
all application traffic
CGI/ASP/JSP
EJB Servlets SOAP
HTTP
SQL JDBC
Web Servers
SQLNet XML/SOAP
App Servers
SNA
Telnet SSH
MAPI
NetBios CIFS
Application Users in Multiple Locations
14
What really happens behind the scenes
Example Oracle Financials
SQL JDBC
Web Servers
Oracle Databases
Oracle App Servers
Oracle User
15
The result faster response time
6.8 sec
11.9 sec
224 sec
2x
10x
15x
3.6 sec
1.2 sec
13.8 sec
After NetScaler
Before NetScaler
After NetScaler
Before NetScaler
After NetScaler
Before NetScaler
Simple Non-Web App over a WAN
Accessing Report Data over a WAN
Accessing Report Data on a LAN
Source TCS Consulting Services, Performance
Analysis of Oracle E-Business Applications, 2005
16
Not just for Oracle
17
Components needed for application
delivery
Acceleration
Security / Access
TCP Optimization
Web Compression
RBA
SSL Acceleration
Dynamic Caching
MP Compression
Intrusion
DDoS
Request Switching
18
Request Switching
Request Switching
Application 1 Application 2
Application 3
1) Receive terminate connections
2) Decrypt / authenticate / analyze every request
3) Queue dispatch valid requests
4) Switch requests multiplex over persistent
connections
19
What is TCP Offload?

• HTTP traffic is usually many short lived
  connections
• Servers perform poorly with short lived
  connections
• TCP Offload separates queries from connections
• This allows it to multiplex many queries on a few
  connections
• Clients do not know the difference
• Servers process much less TCP/IP and thus perform
  better
• TCP Offload alone can make a big difference in
  performance

20
TCP Offload (II)

• NetScaler receives and terminates client
  connection
• Client transmits application requests
• NetScaler establishes server connection

• NetScaler transmits client requests over
  NetScaler - server connection
• Additional clients follow same procedure
• Multiple client requests are transmitted across
  common server connection
• TCP connection processing is minimized

Clients
Server
21
What is Layer 7 Load Balancing?

• A HTTP request has several components
• A URL (e.g., http//www.foo.com/content.html)
• A Cookie
• Client Information (e.g., browser type, etc.)
• Load Balancing decisions can be based on HTTP!
• Direct requests to groups of servers based on
  their URL
• Keep users that have logged into an application
  going to the same server
• Used to send specific URLs to specific servers
• All requests to app.cgi goes to server1, app2.cgi
  go to server2, etc.

22
GSLB Global Server Load Balancing

• Distributes network traffic across multiple sites
• Reduces application latency
• Distributes server load across multiple sites
• Disaster Recovery

23
Different GSLB methods

• Basic Distribution
• Disaster Recovery
• Geographic Proximity
• Fastest Site

24
What is Caching?
Customers
Suppliers/Partners
Mobile Users
Remote Employees
Additional Requests
Initial Request
25
What about Dynamic Caching?

• Static content is easy images dont change
• But enterprise applications arent just images
• Content is dynamically generated based on the
  database
• Different users will see different content for
  many apps
• But Many pages/responses are common for users
• Many sales reports
• Summary statements
• Portal pages
• These pages can be cached too!

26
How Dynamic Caching works

• Break down the URI
• http//www.foo.com/dir/subdir/app.cgi?v1n1v2n2
  v3n3v4n4
• Host header
• Filename (/dir/subdir/app.cgi)
• Individual Query Parameters
• Variable v1 / Value n1
• Variable v2 / Value n2
• Variable v3 / Value n3
• Variable v4 / Value n4
• Similar pages seem to have different URLs
• Name / Value pairs tend to be the same except for
  user identifiers
• You can tell the cache to only use specific
  variables for cache hits

27
Cachings impact

• Eliminates Content Generation Time
• Frequently generated/accessed reports can be
  cached
• For example quarterly reports
• Works with Static and Dynamic Content
• Applies to packaged applications (Oracle, SAP,
  etc.)
• Works with custom applications (Tomcat, BEA,
  J2EE, etc.)
• Reduces Load
• Significant load reduction on application servers
  and databases

28
Netscaler DEMO ?Caching
29
What is HTTP (Web) Compression?

• NetScalers transparent content compression
• Maximize payload per packet
• Increase application performance

30
Whats inside compression?

• Modern browsers support gzip compression
• Upon receipt, they will automatically decompress
• End user doesnt know/care
• NetScaler determines compressability based on
  User Agent header
• Gzip is similar to Winzip in effectiveness
• Compression ratio varies from 31 to 51 on
  average
• Varies by content type

31
AppCompress Extreme(Differential Compression)
Data Center Applications
End Users
On subsequent requests only changed data is sent
Virtually instant response time for end users
32
AppCompress Extreme (II)

• Compression ratio of up to 15x not uncommon
• Think latency, not throughput
• Only works with HTML content
• Existing gzip will catch other cases

33
AppCompress MPMulti-Protocol Compression

• Uploads and Downloads
• Any Application
• Any User Location

Data Center Applications
End Users
Now also compresses large files sent out by users
34
Independent Test Lab Results
44X (0.3 sec)
Standard Web Page with no changes
Standard Web Page with moderate changes
Enterprise Application Report Query
22X (0.7 sec)
8X (2.0 sec)
7X (2.0 sec)
4X (4.3 sec)
4X (4.4 sec)
1X (16.5 sec)
1X (15.8 sec)
1X (15.1 sec)
No Acceleration
AppCompress Extreme
No Acceleration
AppCompress Extreme
No Acceleration
Std. HTTP Compression
AppCompress Extreme
Std. HTTP Compression
Std. HTTP Compression
Source The Tolly Group (June 2005) Test
Applications Google (web), Oracle (enterprise)
35
SSL Acceleration

• All SSL connections terminate on system
• Content is decrypted
• Layer 7 (HTTP) policies are applied
• I.e. switching, acceleration, etc.
• Connection forwarded to server
• Server does not have to take encryption load
• Significantly speeds up server infrastructure

36
SSL Acceleration (II)

• Backend encryption is supported
• NS can encrypt connections to backend server
• Still improves performance because of TCP Offload
• Typical in sensitive environments (HIPAA,
  Finance, etc.)

37
Netscaler DEMO ?Offloading
38
Surge Protection The Problem
Response Time
Incoming HTTP Traffic
Server Load
Response Rate
39
Surge Takeaways

• Peak server effectiveness is not at peak load
• If servers are overloaded
• End user experience degrades
• Once one server goes down, others will follow
• Goal Protect a server from getting moreload
  than it can handle without droppinglegitimate
  user requests

40
Surge Protection What We Want
Incoming HTTP Traffic
Maximum Server Load
Server Load
Response Rate
Response Time
41
What Surge Protection does
Surge Queue
Citrix NetScaler

• Maintains optimal traffic rate to the server
• Does not drop valid traffic surges
• Results in optimal end user response time

42
Additional Resources

• Citrix Netscaler products and solutions
• http//www.citrix.com/netscaler
• Citrix Knowledge Center (its getting better
  every day ? )
• http//support.citrix.com -gt Product Support
  Pages -gt Netscaler Application Delivery
• http//support.citrix.com -gt Product Support
  Pages -gt Netscaler Application Firewall

43
(No Transcript)