Technical Presentation - PowerPoint PPT Presentation

1 / 45
About This Presentation
Title:

Technical Presentation

Description:

Improves end users application response time and productivity. Offload application and ... Opening and closing TCP connections at a high rate, hoping to overwhelm the target TCP ... – PowerPoint PPT presentation

Number of Views:40
Avg rating:3.0/5.0
Slides: 46
Provided by: orya7
Category:

less

Transcript and Presenter's Notes

Title: Technical Presentation


1
Introduction
  • Technical Presentation

2

Crescendo Networks
Mission To Develop Best of Breed Application
Acceleration Solutions for the Worlds Busiest
and Most Demanding Applications
3
AFEs Answer Acute Business Requirements
AFE Sales Exceeding 1 Billion in 2009
1516
1433
Application Acceleration Market Worldwide,
2004-2007, Millions of U.S., end-user spending
57
Load-Balancing
67
AFE
1175
79
999
40
-30
108
783
1459
641
1366
248
1096
AFE CAGR
Load-BalancingCAGR
372
891
535
269
2004
2005
2006
2007
2008
2009
Source Gartner Dataquest (June 2005)
4
Agenda
  • Market Overview
  • Application Delivery Challenge
  • Solving the Challenge
  • Maestro Platform Functionality
  • Deployment Alternatives

5
Evolution of the AFE
Router
L4 Load-Balancer
L7 Load-Balancer
Application Front-End(AFE)
SSL Offload
Web-Acceleration (Comp/Caching)
6
AFE Position
Traffic shaping
Load balancing
Application Front End(AFE)
Technology
Network
Server
Application
Position
Offload application overhead with customized
End-to-End optimization
Target
Managing and Prioritizing traffic On the overall
network
Server accessibility bottle necks.
7
Application Delivery Challenges
  • Content generation compression
  • Consumes application resources
  • Increased application latency
  • TCP connection management
  • Consumes server resources
  • Increases application latency
  • Content encryption security
  • Incremental server load
  • Reduces application performance
  • Incremental users compound
  • Servers / bandwidth / load

ServerResources
100
ApplicationLatency
0
Traditional LoadBalancer
8
Solving The Challenge
  • Results
  • Accelerated applications
  • Decrease response time
  • Optimize infrastructure
  • Increased security
  • Conserve bandwidth
  • TCP Connection management
  • Terminate TCP connections
  • Compression
  • Reduce server compression overhead
  • Content encryption security
  • Offload content encryption

ServerResources
100
ApplicationLatency
0
CN Maestro
9
A Servers Point of View
  • Servers Work

Connection Management
Network Failures
Data
SSL
Compression
DDoS Attacks
Data
10
Platform Functionality
11
Maestro Platform Functionality
  • Key Performance Feature Concurrency!
  • SLT (Short Lived Transaction) Technology
  • TCP Offload
  • Connection Management
  • Request Response Buffering
  • Advanced Offload Modules
  • Compression
  • SSL
  • L4 Load Balancing
  • L7 Content Switching
  • Fast TCP

12
TCP The Problem
  • Servers TCP stack pain points
  • Multiple session setups/teardowns
  • Large number of simultaneous connections
  • Mixture of WAN clients (slow, fast, good/bad
    link)
  • Congestion control, dropped packets,
    retransmissions

13
TCP Offload The Solution
  • True TCP Termination
  • Connection Consolidation
  • Connection Management
  • Server LAN speed transactions
  • Client
  • Request Buffering
  • Response Buffering

Before
After
14
Request Buffering
t
Maestro
Server
Server
Client
Client
Open (SYN)
Open (SYN)
Memory, CPU
(SYN/ACK)
(SYN/ACK)
ACK
ACK
Req 1/2
Req 1/2
More CPU, More Memory
ACK
ACK
Req 2/2
Req 2/2
RTO
RTO
Req 2/2
Full Request
Req 2/2
Memory Alloc., Process, Release
ACK
Processing -gt Memory Released
ACK
ACK
Response
Response
15
Response Buffering
Maestro
Server
Server
Client
Client
t
Resp 1/2
Response Buffered
Response Buffered
Resp 1/2
Resp 1/2
ACK
Resp 2/2
ACK
ACK
Memory Released
ACK
Resp 2/2
Resp 2/2
RTO
RTO
CPU, Memory
Resp 2/2
Resp 2/2
ACK
ACK
Memory Released
16
Fast TCP
  • Optimized client side connection management
  • Fast Slow Start
  • Optimized congestion avoidance algorithm
  • Reduce fluctuations around maximum bandwidth
  • Minimize number of dropped packets

17
Compression
  • Dedicated hardware-based compression engine
  • Inline compression - Zeroed latency
  • Standard Deflate and GZIP algorithm support
  • Performance
  • 1 Gbps compressed throughput
  • Average compression ratio 60
  • Application layer visibility
  • Include compressible MIME types
  • Exclude already compressed MIME types

18
Compression Advantages
  • Optimization
  • Inbound/Outbound bandwidth reduction
  • Acceleration
  • Improves end users application response time and
    productivity
  • Offload application and server resources
  • Relives server from performing compression a
    resource intensive task

19
SSL
  • Dedicated hardware-based SSL engine
  • Symmetric SSL support
  • Client side
  • Server side
  • Full flexibility in all parameters
  • Different SSL Key sizes
  • Certificates Import, or self signed
  • Full range of Cipher algorithms and profiles

20
SSL Advantages
  • Offload servers SSL
  • SSL handshake
  • Different key sizes for server client side are
    optional
  • Full integration with Maestros distributed
    module architecture
  • SLT
  • Compression (compress encrypt)
  • Load Balancing
  • Etc.

21
L4 - Load Balancing
  • Load Balancing Algorithms
  • Round Robin (RR)
  • Weighted Round Robin (WRR)
  • Weighted Least Pending Requests (WLPR)
  • Connection persistency
  • IP based
  • Application level

22
L7 - Switching
  • Pure request-based load balancing
  • Requests are filtered based on
  • URI
  • Host header
  • File type
  • Language

23
DDoS Protection
  • Maestro provides protection against a wide range
    of Distributed Denial-of-Service
  • Few examples

24
The Maestro Platform Functionality
Load Balancing
DDoS
SSL
Application 1 Application 2
L7 Switching
SSL
Application 3
Layer 7 Processing
SLT
SLT
Redundancy
Compression
25
Deployment Alternatives
26
Deployment Alternatives
  • On LAN 1 Leg Topology
  • Routed Mode
  • Spoofed Mode
  • Mash Mode
  • Redundancy Mode

27
On LAN 1 Leg Topology
CN configuration Virtual-IP 1 ? Server
1 Virtual-IP 2 ? Server 2 Virtual-IP 3 ?
Server 3
  • CN resides on LAN
  • Traffic is redirected to the CNs virtual IPs
  • Servers in example are in different clusters

192.168.1.0 / 24
28
Routed
CN configuration Virtual-IP 1 ? Server
1 Virtual-IP 2 ? Server 2 Virtual-IP 3 ?
Server 3
  • CN resides between LANs
  • Traffic is redirected to the CNs virtual IPs
  • CN functions as a Router

Server 1
192.168.1.0 / 24
192.168.2.0 / 24
Server 2
Server 3
29
Spoofed Mode
  • Routed topology
  • Virtual IPs Real servers IPs
  • CN intercepts traffic destined to real servers
    IPs
  • CN functions as a Router

CN configuration Virtual-IP 1 Server
1 Virtual-IP 2 Server 2 Virtual-IP 3 Server 3
Server 1
192.168.1.0 / 24
192.168.2.0 / 24
Server 2
Server 3
30
Mash Mode
  • Large number of interfaces and Routing
    capabilities allow mixture of topologies

192.168.4.0 / 24 Spoofed
192.168.3.0 / 24 Routed
192.168.2.0 / 24 On LAN
Server 3
31
Redundancy Mode
  • Virtual Router Redundancy Protocol (VRRP) based
  • Using a virtual IP and MAC address
  • Deployment modes
  • Hot Stand-By (Active Passive)
  • Load Sharing (Active Active)
  • Enhanced failover decisions making
  • Upstream network device availability
  • Application server health and connectivity
  • Configurable thresholds

32
Product Management
  • GUI
  • JAVA based, simple to use, configure and monitor
  • Accessible standard browser
  • CLI
  • familiar look and feel, standard to L3 devices
  • Accessible via SSH or Telnet
  • SNMP
  • SNMP MIBs provide a flexible way to update the
    device and poll for statistics, used also to
    communicate with 3rd party tools
  • Upgrades
  • via a bundle uploaded via FTP or HTTP
  • Configuration scheme
  • A text based file which can be imported/exported
    via FTP or HTTP
  • Versioning
  • Downgrade/Upgrade up to 1 version backwards

33
Monitoring
  • Statistics
  • Real time statistics
  • History statistics
  • Statistics are available from the different
    interfaces
  • Reporting
  • SNMP Syslog support

34
Management Security
  • Separated secured physical management interface
  • Access-list restrictions over the management port
  • Configurable users and groups with different
    access level to the management interfaces
  • Read Only
  • Administration
  • Technician

35
Hardware Product Line
36
Overview
  • The Maestro architecture is a hardware based
    appliance
  • Hardware models are design to provide maximum
    performance for each of the CN features
  • Using a dedicated task specific processor with
    optimal distribution of the OSI layers over a
    multi gigabit bus platform, the Maestro can
    deliver 100 performance of each of its features
    while executed simultaneously

37
General Specifications
  • 1.5U / 2U box
  • 2/4/8/10 x 1GbE ports
  • 10Gbps bus architecture
  • Web UI SNMP based, Statistics Monitoring,
    Remote management (SSH, Telnet)
  • Redundancy VRRP-Based A/A A/P
  • Agnostic to servers and browsers types
  • Simple maintenance mechanism FTP/HTTP based
  • Co-exists with existing infrastructure LBs,
    switches and routers
  • Certified by FCC, VCCI, CE, CB, UL

38
Maestro Product Line
CN-5510
CN-5504
  • Micro Engines 81
  • Application Memory 4GB
  • Interfaces
  • 10 x SFP GbE ports (optical/copper) 10/100/1000
    Mbps speed
  • Management RJ-45 10/100Mbps Ethernet
  • Performance
  • 1M concurrent connections
  • 500K SSL connections
  • Throughput 10/4Gbps (Total/Optimized)
  • 1Gbps Compression
  • 2.4M SYN/sec
  • Micro Engines 81
  • Application Memory 4GB
  • Interfaces
  • 4 x SFP GbE ports (optical/copper) 10/100/1000
    Mbps speed
  • Management RJ-45 10/100Mbps Ethernet
  • Performance
  • 1M concurrent connections
  • 100K SSL connections
  • Throughput 4/4Gbps (Total/Optimized)
  • 1Gbps Compression
  • 2.4M SYN/sec

39
Distributed Architecture Max Performance
gt 81 micro engines
Buffering Caching
Compression
HTTP Parsing
Serial and Ethernet out of bound management ports
SSL
Multi-GigabitTCP Engine
Layer 1-410GbpsProcessing
4/10 Gigabit Ethernet Ports
40
Success Stories
41
Thank You
42
ALP Module
  • ALP stands for Application Layer Processing
  • The ALP is an innovative Crescendo proprietary
    development
  • Goals
  • Identify and solve bottlenecks within application
    tiers
  • Monitoring tiers capacity
  • Analyze and optimize business processes

43
ALP Components
  • Admission Control
  • Prevents the application to exceed its capacity,
    by queuing requests once tier capacity is reached
  • Request Scheduling
  • Controls and optimizes the order in which
    requests are sent to the application tiers (once
    queued
  • Application Insight
  • Multi-tier application insight with end-to-end
    transaction tracking

44
Oracle World October 2006 - PeopleSoft Impact
with ALP
Direct 78 sec per business process
With ALP 14 sec per process!

81 (526 improvement)
45
Application Optimization Across all Tiers
Maestros Current Optimization/Acceleration
Technology
HTTP centric Optimization (Web Tier)

New ALP Technology
  • Application Acceleration Across Multiple Tiers
  • True Application Processing ? Intelligent
    Transaction Control and Scheduling
  • Increased Visibility and Performance Compliance
    Monitoring
  • SLA Enforcement for Business Processes

Applications centric Optimization (App-Tier)
  • New Levels of Application delivery optimization
  • New level of Backend Tiers Visibility
    Availability
  • Revolutionary convergence of IT Networking

Write a Comment
User Comments (0)
About PowerShow.com