cvc2009 moscow repeater+ica fabian kienle final

39
Citrix WAN Optimisation with Citrix Repeater 5.0 and ICA Acceleration Fabian Kienle Business Development Manager CE-E

Upload: highload-2009

Post on 16-Apr-2017

603 views

Category:

Technology


0 download

TRANSCRIPT

Page 1: Cvc2009 Moscow Repeater+Ica  Fabian Kienle Final

Citrix WAN Optimisationwith Citrix Repeater 5.0 and ICA Acceleration

Fabian KienleBusiness Development Manager CE-E

Page 2: Cvc2009 Moscow Repeater+Ica  Fabian Kienle Final

Six Keys to Successful Application Delivery

Citrix EdgeSight™

Monitor End UserExperience

Citrix Repeater™

Accelerate Appsto Branch Users

Citrix AccessGateway™

Enable SecureApplication

Access

Citrix® NetScaler®

Deliver Web Applications

Citrix XenAppServer™

Deliver Windows Applications

Citrix XenDesktop™

Deliver Desktops

AppsUsers

Page 3: Cvc2009 Moscow Repeater+Ica  Fabian Kienle Final

Citrix Repeater helps with “the last mile”

WANScaler areas of operation

– TCP Flow Control

– Multi-Level Compression

– Protocol Optimization

Page 4: Cvc2009 Moscow Repeater+Ica  Fabian Kienle Final

Protocol Optimization

Page 5: Cvc2009 Moscow Repeater+Ica  Fabian Kienle Final

What is CIFS?•Common Internet File System

– Running on top of SMB “Server Message Blocks”•CIFS is used for

– Directory Browsing– File Transfer– UNC paths– Open/Read/Write/Close operations

•Common trait– Many roundtrips per transaction– Lots of meta data in relation to desired files

Page 6: Cvc2009 Moscow Repeater+Ica  Fabian Kienle Final

How Does WANScaler Accelerate CIFS?•Anticipate requests based on learned behavior

•Read ahead in anticipation of the next data block

•Avoid compressing meta data

– CIFS engine communicates with compression module

Page 7: Cvc2009 Moscow Repeater+Ica  Fabian Kienle Final

Multi-Level Compression

Page 8: Cvc2009 Moscow Repeater+Ica  Fabian Kienle Final

How Does WANScaler Compression Work?• Compression

– Replace a large data chunk with a small token. Send token instead – acts as pointer

– WANScaler Methods:– Disk Based Compression– Memory Based Compression

•Unlike a web cache, WANScaler is not object or file aware. It is only bit stream aware for TCP connections.

• The memory overwrites automatically when the history is full (FIFO).

Page 9: Cvc2009 Moscow Repeater+Ica  Fabian Kienle Final

WANScaler Compression Advantages•Compression is configurable per service class though not

required•WANScaler compression is application independent•Requires zero configuration:

– Automatically chooses the best compression method dynamically:– Disk-based compression (DBC)– Memory-based compression

Page 10: Cvc2009 Moscow Repeater+Ica  Fabian Kienle Final

Multi-Level Compression• Nested compression engines

– Disk-based compression: delivers up to 3500:1 compression for disk matches.

– Memory-based compression: delivers 300:1 compression for memory matches .

– Zlib– LZS

• Automatic – nothing to configure. WANScaler algorithms use the best available based on the situation

Page 11: Cvc2009 Moscow Repeater+Ica  Fabian Kienle Final

Flow Control

Page 12: Cvc2009 Moscow Repeater+Ica  Fabian Kienle Final

Typical TCP Flow Control

•Flow Control– TCP does not know what the bandwidth of the link is!

1TCP Slow Start - packet sending

rate is increased after each round trip.

2TCP Congestion Control -Packet Loss penalty = sending rate cut

by 50%.

Ethernet LAN, 10Mb/s, low latency and loss

Slow Start

Congestion ControlAlgorithm

1

2

x x x x x x

X = packet loss

x x x xx

Page 13: Cvc2009 Moscow Repeater+Ica  Fabian Kienle Final

TCP On the WAN

1High latency means a slower

recovery period during congestion control.

2Feedback (packet loss) is too

infrequent and ambiguous to be accurate.

T3, 45Mb/s, high latency and loss

Slow Start

Congestion Control

1 2

X = packet loss

x x x

Page 14: Cvc2009 Moscow Repeater+Ica  Fabian Kienle Final

TCP On the WAN

1. TCP Distance Bias–Short distance sessions may have packet loss but recover quickly

–Long distance sessions are impacted by packet loss but recover slowly

2. The Result is Low Throughput and Random Application Delays

Slow Start

Long Distance

x x x x x x

X = packet loss

x x x xx

Time (Milliseconds)

Per

form

ance

(Mbp

s) Short Distance1.

Page 15: Cvc2009 Moscow Repeater+Ica  Fabian Kienle Final

Typical WAN Communication

WAN RouterWAN

WAN RouterServer

WAN RouterWAN

WAN RouterClient

Switch Switch

120 ms

Page 16: Cvc2009 Moscow Repeater+Ica  Fabian Kienle Final

Typical WAN Communication

WAN RouterWAN

WAN RouterServer

WAN RouterWAN

WAN RouterClient

Switch Switch

SYN

120 ms

Page 17: Cvc2009 Moscow Repeater+Ica  Fabian Kienle Final

Typical WAN Communication

WAN RouterWAN

WAN RouterServer

WAN RouterWAN

WAN RouterClient

Switch Switch

SYN

SYN + ACK

ACK

120 ms

Page 18: Cvc2009 Moscow Repeater+Ica  Fabian Kienle Final

Typical WAN Communication

WAN RouterWAN

WAN RouterServer

WAN RouterWAN

WAN RouterClient

Switch Switch

SYN

SYN + ACK

ACK

HTTP GET

120 ms

Page 19: Cvc2009 Moscow Repeater+Ica  Fabian Kienle Final

Typical WAN Communication

WAN RouterWAN

WAN RouterServer

WAN RouterWAN

WAN RouterClient

Switch Switch

SYN

SYN + ACK

ACK

HTTP GET

120 ms

ACK+Data

Page 20: Cvc2009 Moscow Repeater+Ica  Fabian Kienle Final

Typical WAN Communication

WAN RouterWAN

WAN RouterServer

WAN RouterWAN

WAN RouterClient

Switch Switch

SYN

SYN + ACK

ACK

HTTP GET

120 ms

ACK+Data

ACK

Page 21: Cvc2009 Moscow Repeater+Ica  Fabian Kienle Final

TCP Flow Control with WANScaler

WAN Server

WAN Router

WANClient

Switch120 ms

SwitchWAN Router

Page 22: Cvc2009 Moscow Repeater+Ica  Fabian Kienle Final

TCP Flow Control with WANScaler

WAN Server

WAN Router

WANClient

Switch120 ms

SwitchWAN Router

WANScaler WANScaler

Page 23: Cvc2009 Moscow Repeater+Ica  Fabian Kienle Final

TCP Flow Control with WANScaler

WAN Server

WAN Router

WANClient

Switch120 ms

SwitchWAN Router

WANScaler WANScaler

SYN SYN+AC

KACK

HTTP GET

ACK AC

KACK

Fast Side Fast SideSlow Side Slow Side

Page 24: Cvc2009 Moscow Repeater+Ica  Fabian Kienle Final

WANScaler Flow Control

Each Segment has its own flow control:– Commonly deployed TCP Windows are 64kB max.– On the WAN side, WANScaler increases the Window to 8MB

(RFC 1323)– WANScaler acknowledges packets on the LAN side, so server

keeps sending Use rate-based sender on the WAN segment. Never

send faster than the configured link speed

WAN Server

WAN Router WANClient

Switch120 ms

SwitchWAN Router

WANScaler WANScaler

Page 25: Cvc2009 Moscow Repeater+Ica  Fabian Kienle Final

What’s new with version 5.0 ?ICA Acceleration

Page 26: Cvc2009 Moscow Repeater+Ica  Fabian Kienle Final

Location based ICA Optimisation

Citrix Repeater

Corporate LAN/Data Center

Branch Office/Regional Site

Based on the well known optimisation technologies, already known from the ICA ClientCaching function taken away from the client, and taken into the data center (Citrix Repeater).So it can be used by ANY user at the data center location !

Page 27: Cvc2009 Moscow Repeater+Ica  Fabian Kienle Final

Optimisation of GUI-, print- and data transfer

Branch Office/Regional Site

Corporate LAN/Data Center

Page 28: Cvc2009 Moscow Repeater+Ica  Fabian Kienle Final

System requirements for ICA AccelerationFor the first release:• XenApp Client

• 32-bit Windows PC* • 11.0

• XenApp Server• ONLY 32-bit Windows Server 2003• XA Server Version 4.5 or 5.0• HRP03 with post Acceleration HF

• Supported Appliances• Branch Repeater (Windows Server)

Version 1.5• Branch Repeater (Linux) on Version 5.0• WANScaler 8000 Series with version 5.0• WANScaler Client offers ONLY TCP

Acceleration**

Page 29: Cvc2009 Moscow Repeater+Ica  Fabian Kienle Final

Scenario # 1:

Example: Two users open the same document, but data will only be transferred once.

• ICA user uses the same Citrix Repeater Box:

1. Different users, by using the same ICA window size and resolution, that access the same texts, numbers etc. (z.B. Word or Excel)

Page 30: Cvc2009 Moscow Repeater+Ica  Fabian Kienle Final

Scenario # 2:

Example: Different users are using form or browser based appplications with identical background objects. Redundant data will not be transferred.

• ICA user uses the same Citrix Repeater Box:

2. Different users, that use the same ICA window size and resolution, scrolling within the same document.

Page 31: Cvc2009 Moscow Repeater+Ica  Fabian Kienle Final

Support of mixed environmentsHQ/Data Center

Branch Office

Citrix RepeaterBranch Repeater or WANScaler

Branch Office

Internet/WAN

For users WITHOUT WanScaler /Branch Repeater Box or with an older ICA Client it all stays the same!

Page 32: Cvc2009 Moscow Repeater+Ica  Fabian Kienle Final

Case Study #1Example of a non optimised environment

WANMPLS

10 - 45 Mbps> 180 ms

50 ICA users15% ICA data traffic

85% non ICA data traffic40% line utilisation ng

XenApp Server 4.5Mainly usage of the

Internet Explorer

Environment:• Slow applications through a slow network (high latency) • But nearly no usage of same/redundant ICA data• No repeatable application scenario (Usage of same data/documents)

Page 33: Cvc2009 Moscow Repeater+Ica  Fabian Kienle Final

Case Study #2Company with branch offices and a centralised serverfarm

WANDSL or fractional T1 (256 kbps – 768 kbps)

50 ms20 ICA users

>95% ICA data traffic>90% line utilisation

XenApp Server 4.5MS OfficePrinting

File sharing

Environment:• Users use same Word and Excel files• Printing is an important, and frequently used function • Users very often copy files (via ICA) from or to centralised file servers

Page 34: Cvc2009 Moscow Repeater+Ica  Fabian Kienle Final

Case Study #3Central server farm, big amount if print data

WANMPLS

1.5 Mbps~ 50 ms

30 ICA users>75% ICA data traffic>90% line utilisation

XenApp Server 4.5ERM Software

Web PortalPrinting of PDF files

Environment:• Slow Response times (applications) by large, and frequently used PDF-Print files• Usage of WEB based Portals and ERM software. Access on same Datensätze.• No usage of MS Office applications and Client Drive Mapping

Page 35: Cvc2009 Moscow Repeater+Ica  Fabian Kienle Final

ICA Acceleration - facts Ideal for print data, CDM file transfer and MS Office as well as applications with many

identical window contents

NOT ideal for Adobe/CAD applications as there are only few opportunities for compression and nearly no redundant data

ICA screen resolution makes the difference. Same resolution is optimal. Differentiating resolution means more (per resolution) data traffic and a larger amount of data within the Citrix Repeater Cache.

Page 36: Cvc2009 Moscow Repeater+Ica  Fabian Kienle Final

ICA Acceleration - facts• Optimal, when many ICA users utilise the same Citrix Repeater Box:

Different users are accessing under usage of the same ICA window size and resolution based on identical data (example: Word or Excel)

Same or different users are blättern within the same file under usage of the same ICA window size and resolution

Different or identical users are printing the same files via ICA

Duplication of same files under usage of ICA Client Drive Mapping. Example: local storage of mail attachments from a published Outlook

Page 37: Cvc2009 Moscow Repeater+Ica  Fabian Kienle Final

Six Keys to Successful Application Delivery

Citrix EdgeSight™

Monitor End UserExperience

Citrix Repeater™

Accelerate Appsto Branch Users

Citrix AccessGateway™

Enable SecureApplication

Access

Citrix® NetScaler®

Deliver Web Applications

Citrix XenAppServer™

Deliver Windows Applications

Citrix XenDesktop™

Deliver Desktops

AppsUsers

Page 38: Cvc2009 Moscow Repeater+Ica  Fabian Kienle Final

Data Center

Transform your datacenter into a delivery center

Page 39: Cvc2009 Moscow Repeater+Ica  Fabian Kienle Final