KEMBAR78
Web Performance Acceleration with Strangeloop AS1000 | PDF
Web Performance Acceleration
          with Strangeloop AS1000

              Thomas Stensitzki



1. European Day of DotNetNuke @ Paris - Session 159
Thomas Stensitzki
• Senior Consultant at iCOMcept GmbH, Aachen
    – www.icomcept.de
    – Microsoft Gold Partner
    – SPLA Hosting Provider
•   MCITP, Enterprise Administrator, Exchange 2007 & 2010
•   MCTS, Sharepoint 2007, Virtualization, BDD
•   MCPD, ASP.NET 3.5
•   Founder .NET Usergroup Aachen (OecherCoder, www.oechercoder.net)
•   Member European DotNetNuke Professionals
•   Blog: http://www.sf-tools.net
•   E-Mail: thomas.stensitzki@icomcept.de



                                                                       2
Strangeloop AS1000




                     3
Overview


  Company    Strangeloop Networks - Vancouver, Canada (2006)

   Product   Strangeloop AS1000™ Application Scaling Appliance

Clients




Awards

                                          Best in Optimization 2009
                                         www.networkproductsguide.com



                                                                        4
The Challenge of Dynamic Web Applications

                                               Web 2.0 : Dynamic Web Applications
                                                                                     Collaboration
                                                                                         Apps
                                                                                  Web Services

                                                                               User Created
                                                                                 Content


                                                                           Interactive

                                                                   Personalized

                                                               Real-time
                                                                 data

                                                       Wikis
                                               Video
                            Brochure   Voice
Web 1.0: Static Web Sites     sites

  Rich applications place new, unmet performance demands on IT infrastructure

                                                                                              5
Front-End Performance Challenge

10~20% of the load time is spent on the back-end
80~90% of the load time is spent on the front-end.
                                               Front-End
                       Performance Challenges             $ Spent on Performance


                                          Back-end
                                          Front-end                     Back-end
                                          Bandwidth                     Front-end
                                          Servers                       Bandwidth
                                          CDN                           CDN




                                               Front-End

                                                      Empty Cache   Primed Cache
                        www.wikipedia.org                82%            80%
                        www.microsoft.com                75%            72%
                        www.yahoo.com                    86%            81%
                        www.cnn.com                      85%            82%
                        www.facebook.com/                97%            96%
                        www.aol.com                      92%            85%
                        www.nytimes.com                  72%            68%

                                                                                    6
Solving the Performance Challenge

 Development approach
                                 Strangeloop AS1000
+ Tried and true optimization      Strangeloop AS1000                IT / Network approach
  techniques                                                        + Offload tasks to network
+ Lever software /platform        Optimize applications             + Predictable, measurable
  features                        automatically, in real time,        ROI
- Hand tune components,           with a network device             - No knowledge of
  pages, code for                                                     application or data
  performance                                                         status
- No knowledge of user
- Demands rare and costly
  development resources




            Joining optimization techniques with a network device to change how
                            applications are developed and deployed
                                                                                            7
Solution for Front-End Performance Problem




                                             8
Demo


       9
Strangeloop – Dynamic Choreography Engine


  Pipeline Manager
  In Memory Database
  Pipeline Mirrors & Data Replication with High Availability



                        Dynamic Choreography Engine

                         Application           Network              Browser
                           - Facing             - Based             - Facing
                         Treatments           Treatments          Treatments



   Web                                                                         Browsers
Applications                           c

                              Configuration                Analysis



                                                                                    10
Deployment Options




                     11
Strangeloop Application Acceleration Treatments




                                             12
Strangeloop AS1000 Additional Features




                                         13
Http Compression


                  Automatic Compression
                                                Web Servers
 Client Aware Compression

 Adaptive Compression

 No End-User Agents
     • Gzip, Deflate




                                            Client Browser

                                                              14
RCache


                                                    Web Servers
 Pre-Compressed Content

 Static / Dynamic Content Caching

 Multi-Store Caching




                                              Clients request contents




                                                                         15
Predictive Browser Caching

    Intelligent Pre-Fetching
    Browser Idle Time Fetching




                   political.aspx    435
BuildASign Home                                 Political Signs        colbert.aspx   158   Yard Template
                   signs.aspx        130
                                                                       wallet.aspx    42
                   bumpers.aspx      75
                                                                       nation.aspx    17
                   banners.aspx      26
                                                                       lincoln.aspx   13
                   magnetics.aspx    15
                                                                       roberts.aspx   6
                   accessories       8
                                                                       signs.aspx     3
                   all.aspx          7
                                                                       all.aspx       1
                   buildasign.aspx   0




                                           Political Sign & Template


                                                                                                        16
Connection Maxomizer


 Maximising TCP Connection




                                                  Clients
   Web            Strangeloop AS1000
Applications



                                       Internet




                                                            17
Image Consolidation

            Real Time Browser Based Image Consolidation


 Different techniques based on browsers   Web Servers
                                                          Multiple
                                                          Images




                                                           Single
                                                           Image




                                                                     18
JavaScript & CSS Consolidation

             Intelligent JavaScript and CSS Consolidation

                                                    Web Servers

 Adaptive Consolidation
     • JavaScript and CSS Consolidation based
     on browser




                                                   Client Browser


                                                                    19
ASP.Net ViewState

Automated Tokenization for ASP.NET ViewState



                  ViewState




                                 (ViewState)
        Server



                  ViewState




         Automatic ViewState
         Removal & Re-Insert


         Client

                                               20
Multiple Hosting Services




                                             Strangeloop AS1000              192.168.10.101
                     Internet                                                192.168.10.102
Clients                                                                      192.168.10.103
                                       VIP : 10.10.10.11
                                       www.strangeloop.com 192.168.10.101
                                                            192.168.10.102
                                                            192.168.10.103
                                       blog.strangeloop.com 192.168.11.101   192.168.11.101
                                                            192.168.11.102   192.168.11.102
                                                            192.168.11.103   192.168.11.103
          Domain Name Server
          10.10.10.11 www.strangeloop.com
          10.10.10.11 blog.strangeloop.com                                   Web Servers




                                                                                              21
UI

  Overview           Diagnosis




Configuration   High Availability




                                    22
AS1000 Appliance - Models


                      AS1000 Basic                    AS1000               AS1000 Enterprise
                       Single Dual Core           Single Quad Core             Dual Quad Core
     CPU
                          Zeon 5500                  Zeon 5500                   Zeon 5500

   Memory                   4 GB                        8 GB                       16 GB

   Network        10/100/1000 Mbps Ethernet   10/100/1000 Mbps Ethernet   10/100/1000 Mbps Ethernet

     HDD                     80G                        160G                        160G

    Profile                  1U                          1U                          2U

    SSL CPS                 5,000                      15,000                      24,000

SSL Throughput            500 Mbps                     1Gbps                       1Gbps

Max Concurrent
                           32,000                      100,000                     250,000
 Connection

HTTP Requests &
                          4,000 TPS                  10,000 TPS                  20,000 TPS
  Responses




                                                                                                 23
AS1000 Virtual Appliance


  Benefits Of Virtual Deployment

  A virtual AS1000 deployment carries with it the benefits of
  virtual environments, which include:
   Ease of deployment without any network interruption,
    rewiring, or impact
   Use of existing resources and available processing
    resources
   Increased hardware utilization
   Decrease in costs, including hardware, power consumption,
    cooling, and operational expenses
   Scalability and elasticity


          Platform           Supported OS          OS Version
   VMWare Infrastructure 3        Yes          ESX 3.0 (3.0.0+)
     VMWare vSphere 4             Yes          ESX 4.0 (4.0.0+)
         XenServer                Yes            XenServer 5.5




                                                                  24
Sirius Benefit Plans (Canada)

Performance Challenge
•   With large numbers of participants
    simultaneously using their online benefits
    management application, Sirius Benefit Plans has
    application performance as a top
    priority.

Solution
•   The Strangeloop AS1000 improves the
    performance and scalability of dynamic Web
    applications by intelligently reading requests and
    applying ASP.NET and AJAX software
    optimization techniques in the network.

Benefits
•   Serve 300% more simultaneous participants
•   64% bandwidth savings across all pages
•   Over 260 hour savings of end-user time per
    month

                                                                   27
Millennium Pharmacy (USA)


Performance Challenge
•   Ensure that a data-intensive application
    delivers consistently fast response times
•   Reduce time spent on performance
    optimization
•   Focus on developing competitive new
    service features

Solution
•   Deploy Strangeloop AS1000 to minimize
    data traffic and accelerate application
    performance

Benefits
•   Improved developer productivity
•   More development resources for new
    feature development



                                                            28
Autoanything.com (USA)


Goals
•   Maximize Revenue
•   Increase User Satisfaction

Solution
•   Strangeloop AS1000 Network Appliance

Results
•   33% increase in revenue
•   8% increase in conversion rate
•   5% reduction in bounce rate




                                                              29
Disaboom (USA)


Business
•   Online interactive community for people with
    disabilities
•   > 76,000 subscribers
•   > 600,000 visitors / month in October
•   # of visitors doubled from July to October

Performance Challenge
•   Small IT staff
•   Need to deliver new features and make sure
    application scales
•   > 7% of users on dial-up
•   Expect continued growth

Results
•   Reduced payload by 50%, perceptibly
    improving user response times
•   Offloaded servers significantly


                                                           30
The Future

                                              Personalized Optimization
                                                  By User and Flow




Each browser and browser version
is different and must be optimized
             differently


                                                                                Mobile
                                                                              Optimization
                                                      Real Time
                                                      Behaviour
                                                      Heuristics
                      Auto Browser Detect
                       and Optimization

                               J     S C G
                                     J
                                     W S HI
          Auto Cabinets        P
                               G
                                     S
                                     F S TF
                                         M
                                     T
                                         L

                                                                              AS1000 Automatically
                                                                          Optimizes for the Mobile Web


                                                                                                         31
Q&A
           DE: webbooster.de
     EN: strangeloopnetworks.com
European inquiries: info@webbooster.de
                                         32

Web Performance Acceleration with Strangeloop AS1000

  • 1.
    Web Performance Acceleration with Strangeloop AS1000 Thomas Stensitzki 1. European Day of DotNetNuke @ Paris - Session 159
  • 2.
    Thomas Stensitzki • SeniorConsultant at iCOMcept GmbH, Aachen – www.icomcept.de – Microsoft Gold Partner – SPLA Hosting Provider • MCITP, Enterprise Administrator, Exchange 2007 & 2010 • MCTS, Sharepoint 2007, Virtualization, BDD • MCPD, ASP.NET 3.5 • Founder .NET Usergroup Aachen (OecherCoder, www.oechercoder.net) • Member European DotNetNuke Professionals • Blog: http://www.sf-tools.net • E-Mail: thomas.stensitzki@icomcept.de 2
  • 3.
  • 4.
    Overview Company Strangeloop Networks - Vancouver, Canada (2006) Product Strangeloop AS1000™ Application Scaling Appliance Clients Awards Best in Optimization 2009 www.networkproductsguide.com 4
  • 5.
    The Challenge ofDynamic Web Applications Web 2.0 : Dynamic Web Applications Collaboration Apps Web Services User Created Content Interactive Personalized Real-time data Wikis Video Brochure Voice Web 1.0: Static Web Sites sites Rich applications place new, unmet performance demands on IT infrastructure 5
  • 6.
    Front-End Performance Challenge 10~20%of the load time is spent on the back-end 80~90% of the load time is spent on the front-end. Front-End Performance Challenges $ Spent on Performance Back-end Front-end Back-end Bandwidth Front-end Servers Bandwidth CDN CDN Front-End Empty Cache Primed Cache www.wikipedia.org 82% 80% www.microsoft.com 75% 72% www.yahoo.com 86% 81% www.cnn.com 85% 82% www.facebook.com/ 97% 96% www.aol.com 92% 85% www.nytimes.com 72% 68% 6
  • 7.
    Solving the PerformanceChallenge Development approach Strangeloop AS1000 + Tried and true optimization Strangeloop AS1000 IT / Network approach techniques + Offload tasks to network + Lever software /platform Optimize applications + Predictable, measurable features automatically, in real time, ROI - Hand tune components, with a network device - No knowledge of pages, code for application or data performance status - No knowledge of user - Demands rare and costly development resources Joining optimization techniques with a network device to change how applications are developed and deployed 7
  • 8.
    Solution for Front-EndPerformance Problem 8
  • 9.
  • 10.
    Strangeloop – DynamicChoreography Engine Pipeline Manager In Memory Database Pipeline Mirrors & Data Replication with High Availability Dynamic Choreography Engine Application Network Browser - Facing - Based - Facing Treatments Treatments Treatments Web Browsers Applications c Configuration Analysis 10
  • 11.
  • 12.
  • 13.
  • 14.
    Http Compression Automatic Compression Web Servers  Client Aware Compression  Adaptive Compression  No End-User Agents • Gzip, Deflate Client Browser 14
  • 15.
    RCache Web Servers  Pre-Compressed Content  Static / Dynamic Content Caching  Multi-Store Caching Clients request contents 15
  • 16.
    Predictive Browser Caching  Intelligent Pre-Fetching  Browser Idle Time Fetching political.aspx 435 BuildASign Home Political Signs colbert.aspx 158 Yard Template signs.aspx 130 wallet.aspx 42 bumpers.aspx 75 nation.aspx 17 banners.aspx 26 lincoln.aspx 13 magnetics.aspx 15 roberts.aspx 6 accessories 8 signs.aspx 3 all.aspx 7 all.aspx 1 buildasign.aspx 0 Political Sign & Template 16
  • 17.
    Connection Maxomizer  MaximisingTCP Connection Clients Web Strangeloop AS1000 Applications Internet 17
  • 18.
    Image Consolidation Real Time Browser Based Image Consolidation  Different techniques based on browsers Web Servers Multiple Images Single Image 18
  • 19.
    JavaScript & CSSConsolidation Intelligent JavaScript and CSS Consolidation Web Servers  Adaptive Consolidation • JavaScript and CSS Consolidation based on browser Client Browser 19
  • 20.
    ASP.Net ViewState Automated Tokenizationfor ASP.NET ViewState ViewState (ViewState) Server ViewState Automatic ViewState Removal & Re-Insert Client 20
  • 21.
    Multiple Hosting Services Strangeloop AS1000 192.168.10.101 Internet 192.168.10.102 Clients 192.168.10.103 VIP : 10.10.10.11 www.strangeloop.com 192.168.10.101 192.168.10.102 192.168.10.103 blog.strangeloop.com 192.168.11.101 192.168.11.101 192.168.11.102 192.168.11.102 192.168.11.103 192.168.11.103 Domain Name Server 10.10.10.11 www.strangeloop.com 10.10.10.11 blog.strangeloop.com Web Servers 21
  • 22.
    UI Overview Diagnosis Configuration High Availability 22
  • 23.
    AS1000 Appliance -Models AS1000 Basic AS1000 AS1000 Enterprise Single Dual Core Single Quad Core Dual Quad Core CPU Zeon 5500 Zeon 5500 Zeon 5500 Memory 4 GB 8 GB 16 GB Network 10/100/1000 Mbps Ethernet 10/100/1000 Mbps Ethernet 10/100/1000 Mbps Ethernet HDD 80G 160G 160G Profile 1U 1U 2U SSL CPS 5,000 15,000 24,000 SSL Throughput 500 Mbps 1Gbps 1Gbps Max Concurrent 32,000 100,000 250,000 Connection HTTP Requests & 4,000 TPS 10,000 TPS 20,000 TPS Responses 23
  • 24.
    AS1000 Virtual Appliance Benefits Of Virtual Deployment A virtual AS1000 deployment carries with it the benefits of virtual environments, which include:  Ease of deployment without any network interruption, rewiring, or impact  Use of existing resources and available processing resources  Increased hardware utilization  Decrease in costs, including hardware, power consumption, cooling, and operational expenses  Scalability and elasticity Platform Supported OS OS Version VMWare Infrastructure 3 Yes ESX 3.0 (3.0.0+) VMWare vSphere 4 Yes ESX 4.0 (4.0.0+) XenServer Yes XenServer 5.5 24
  • 25.
    Sirius Benefit Plans(Canada) Performance Challenge • With large numbers of participants simultaneously using their online benefits management application, Sirius Benefit Plans has application performance as a top priority. Solution • The Strangeloop AS1000 improves the performance and scalability of dynamic Web applications by intelligently reading requests and applying ASP.NET and AJAX software optimization techniques in the network. Benefits • Serve 300% more simultaneous participants • 64% bandwidth savings across all pages • Over 260 hour savings of end-user time per month 27
  • 26.
    Millennium Pharmacy (USA) PerformanceChallenge • Ensure that a data-intensive application delivers consistently fast response times • Reduce time spent on performance optimization • Focus on developing competitive new service features Solution • Deploy Strangeloop AS1000 to minimize data traffic and accelerate application performance Benefits • Improved developer productivity • More development resources for new feature development 28
  • 27.
    Autoanything.com (USA) Goals • Maximize Revenue • Increase User Satisfaction Solution • Strangeloop AS1000 Network Appliance Results • 33% increase in revenue • 8% increase in conversion rate • 5% reduction in bounce rate 29
  • 28.
    Disaboom (USA) Business • Online interactive community for people with disabilities • > 76,000 subscribers • > 600,000 visitors / month in October • # of visitors doubled from July to October Performance Challenge • Small IT staff • Need to deliver new features and make sure application scales • > 7% of users on dial-up • Expect continued growth Results • Reduced payload by 50%, perceptibly improving user response times • Offloaded servers significantly 30
  • 29.
    The Future Personalized Optimization By User and Flow Each browser and browser version is different and must be optimized differently Mobile Optimization Real Time Behaviour Heuristics Auto Browser Detect and Optimization J S C G J W S HI Auto Cabinets P G S F S TF M T L AS1000 Automatically Optimizes for the Mobile Web 31
  • 30.
    Q&A DE: webbooster.de EN: strangeloopnetworks.com European inquiries: info@webbooster.de 32