Avaya P333R-LB User guide

Avaya P333R-LB
Load Balancing Stackable Switch
Load Balancing
Application Guide
May 2001

Communication without boundaries
Table of Contents:
Section 1: Introduction
Section 2: Application 1 — Server Load Balancing
Section 3: Application 2 — Firewall Load Balancing
Section 4: Application 3 — Application Redirection
Section 5: Application 4 — Policy-Based Load Balancing

Communication without boundaries
1
Introduction
Today, businesses rely on the ability to conduct
transactions and support customer relationships
over the Internet. Delay-free access to information
is the key to businesses running smoothly.
Traditional network equipment and design is not
suitable to respond to the increasing demand on
availability and performance. The two key factors
in ensuring you can keep up with your customers’
demands are:
High Availability—Your clients should not be con-
cerned with your network problems. All they want
is 24/7 access without excuses. Load Balancing
provides transparent reliability—once configured
the P333R-LB gives worry-free, availability.
Removing Bottlenecks—High Availability is only
half the story; some network element along the way
may slow everything down. Any software-based
entity in your network, such as a firewall or server
can become a bottleneck. By adding more elements
in parallel and Load Balancing, you dramatically
boost bandwidth.
Load balancing allows you to simply divide
demand among multiple resources to ensure that
information flows freely.
The applications in this document show how
to deploy Load balancing using the P333R-LB
to maximize network performance by increasing
availability and removing bottlenecks.
What is driving the need for
Internet Traffic Engineering?
The Growth of eBusiness Applications: The explosive
growth in critical eBusiness applications has created
the need to distribute application services over
multiple servers for scalability and enhanced
availability. The problem with traditional network
devices is that many eBusiness applications, such
as eCommerce transactions (e.g., shopping carts),
search engine sessions, and individualized services
(e.g., changing your bank account information
online) require that the client communicate with
the same server for the entire session. This requires
content-aware switching, the ability to examine
the packet, and intelligently bind clients to a server
using a variety of criteria that may span multiple
TCP connections. This goes beyond the capability
of traditional L3-4 switches.
Security: Along with increased connectivity and
access to the corporate network from extranets
and eBusiness applications, comes greater risk
to network security. Traditionally, routers and
firewalls have been used to protect critical corporate
server resources, but as traffic demands grow, their
software-based filter processing capabilities hamper
access and response. This has created a need for
switches that can process an Access Control List
(ACL) at hardware-based speeds, while at the same
time mitigating external threats from Denial of
Service and spoofing attacks.
The Keys to a Better Network:
• High availability—
Users always have
access to resources
• Removing bottlenecks—
To ensure that servers,
firewalls, and other key
components are not
overloaded and limit
network performance.

Communication without boundaries
The P333R-LB
The P333R-LB enhances the P330 stackable switch
solution by providing full load balancing and
routing functionality in a single switch.
Recent trends in network equipment in general,
and switches in particular, have been towards
ease of use and deployment.
Avaya leads these trends: the P333R-LB continues
the P330 family tradition of being powerful, simple,
flexible and offering an unprecedented price/
performance ratio.
Flexible
The P333R-LB has the same flexibility as the P330
system: The P333R-LB has both expansion module
to allow a variety of uplinks (fiber Ethernet,
GE copper fiber or GBIC and even ATM) and
a stackability option.
This stackability option allows scalability and
investment protection for customers needing more
ports and load balancing throughput, increasing
investment protection on any network design.
Transparent
Once configured, the P333R-LB “disappears” as far
as your users are concerned. As a result, you do not
need to perform a large-scale reconfiguration to
benefit from increased availability and throughput.
Price/Performance
However good a product is, it must deliver the
performance you need at a viable price. The P333R-LB
provides all your load balancing needs in a single,
compact switch.
With the P333R-LB, you can now afford to implement
full Layer 2 and 3 switching and Layer 4 load bal-
ancing anywhere in your network. Now applications
such as Server load balancing, Firewall Load
Balancing, Application Redirection and Policy-based
load balancing have become cost-effective.
The P333R-LB provides all this in one switch.
Compare this with other solutions that require
several expensive switches to perform the
same functionality.
2

Communication without boundaries
3
• Round Robin—This
scheme directs traffic
to the next server
in the list.
•
Hash—This statistically-
based scheme is
based on a pre-defined
algorithm.
• Response Time-based—
This scheme directs
traffic to the least
busy server.
Application 1—Server Load
Balancing
Server load balancing allows you to provide high
availability and overcome bottlenecks caused by
limitations in the servers.
Configuring mirrored servers via load balancing
means that, if one server is taken off-line for any
reason, users will not experience any delays or
interruptions. You can extend this availability to
a number of levels, from servers to server farms
to load balancing switches.
Load balancing can also overcome bottlenecks
created by “server session overload.” Without load
balancing, the number of sessions that the server
can handle may become insufficient for all the users
who are interacting with your organization.
By intelligently dividing the workload among
your servers, you ensure that users can continue
to interact in a bottleneck-free environment.
Installing one or more P333R-LB stackable load
balancing switches as the server farm switch allows
you to easily implement Server load balancing.
The server farm appears as a single Virtual IP (VIP)
address to the outside world, so you can implement
multiple servers for increased throughput and
redundancy with no disruption.
Load Balancing Schemes: You can choose the load
balancing scheme according to your needs. The
P333R-LB offers Round Robin, Hash and Response
Time-based schemes.
Persistency. Persistency is important since it ensures
that users receive consistent and uninterrupted
responses to their requests.
Persistency is especially important in eCommerce
where persistency of a session with the same
server ensures that transactions between the user
and a specific server will not be lost. Persistency
ensures that the SSL encryption session will not be
cut by moving to a different server. The P333R-LB
offers two persistency schemes: Hash, Source/
destination IP-based (and destination Port), and
activity timer-based.
Health Checking: In order to make the load balancing
more effective in terms of availability, we have
incorporated a health checking mechanism into
the load balancer that ensures that traffic is only
directed to “healthy” servers. This vital component
of load balancing is performed by “pinging” the
servers and checking for a response.
Graceful Shutdown: This allows you to take down
a server to perform maintenance without your users
feeling a thing. Old sessions continue until they are
completed, and new sessions are directed to another
server. Once all old sessions are completed, you can
take down the server.

Communication without boundaries
The Load Balancing Process
The P333R-LB selects the best available physical
server upon receiving a new session request based
on the following:
• Server health (based on application health,
not just server status)
• Number of open sessions
• Round Robin
• Hash
• Persistence: Subsequent packets in the same
session sent to the same server until the
session terminates.
• Session address translation performed on
every packet
This process is transparent both to the user and
to the server, so no configuration is required (unlike
proxy servers, for example).
Unlike simple load-balancers, the Avaya load
balancing solution supports multiple persistence
methods. This enables the P333R-LB to fully
support eBusiness applications that require
continuous communication between the client
and server throughout a transaction session. The
following persistence methods are supported:
• IP address hashing
• Source address with inactivity timer
High Availability with Cajun
The special configuration below shows how each
element is fully backed up to provide maximum
uptime:
• Server fails—The P333R-LB senses the loss
of the server and directs all new sessions to
the other servers that are available in the farm.
• P333R-LB fails—enhanced VRRP ensures
transparent switchover to the second switch.
A special topology ensures the server
connection to the network.
• Uplink fails—STA and port redundancy bring
the backup LAG on-line instantaneously.
4
Gigabit Ethernet
with LAG
Redundant
LAG
Gigabit Ethernet
with LAG
Avaya P333R-LB
Avaya P333R-LB
Avaya P882
Avaya P580
Avaya P330
Server Farm
IP1IP2IP3IP4
P880
Virtual IP A
*future

Avaya P333R-LB
WAN Router
Avaya P333R-LB
Gigabit Ethernet
with LAG Avaya P333R-LB
VRRP
LAN Internet
WAN Router
Avaya P333R-LB
Gigabit Ethernet
with LAG Avaya P333R-LB IPa
IPb
VRRP
LAN Internet
Communication without boundaries
Application 2—Firewall
Load Balancing
Firewalls can inherently constitute a bottleneck
since they are software-based. There is also the
issue of transparent availability: firewalls can act
as a single point of failure, causing severe problems
with Internet access.
The P333R-LB can be used to load-balance across
multiple routers and firewalls and overcome
these problems.
Implementing the P333R-LB removes the bottleneck
since the load is distributed at hardware speed over
multiple firewalls.
To ensure availability, the P333R-LB switches on
both sides of the firewall perform continuous health
checks on the links to the firewall, the firewall itself,
and each other.
The two applications below show firewall load
balancing in systems both with and without NAT
(Network Address Translation).
No single point of failure. The doubling up of the key
components ensures that communication between
the LAN and Internet is always maintained.
If one P333R-LB fails, due to VRRP, the second
switch can instantaneously take over all load
balancing functions. If a firewall fails, then the
P333R-LB will transparently redirect all traffic
through the second firewall.
“No NAT” Application. In this case, there are pairs
of P333R-LB switches on each side of the firewalls.
This is necessary since sessions must travel across
the same firewall. If the session is sent to the
second firewall, it will be disconnected by the
“statefull” firewall. It is therefore important to
have the same load balancing decisions on both
sides of the firewall.
NAT Application. In this case, P333R-LB switches
are only required on the LAN side of the firewalls.
The session traffic coming from the Internet will
have the specific IP address of the firewall from
which the session started.
5
• NAT (Network Address
Translation)—allows
you to use any IP
address within your
organization while
only using legal
IP addresses outside.
This is useful when you
have a limited number of
legal IP addresses (e.g.,
for ISPs).
Firewall Load Balancing – no NAT
Firewall Load Balancing – with NAT

Communication without boundaries
Application 3—Application
Redirection
Application redirection is the diversion of traffic
from certain applications to a different direction
than the original one without the user/source
being aware or requiring any reconfiguring.
Cache Redirection: One of the most common uses
of application redirection is “cache redirection.”
Today, many organizations use proxy servers
for Internet access. This approach, however, has
problems which makes it less than ideal:
• It is only possible to deploy a single proxy
server which is both a bottleneck and a single
point of failure.
• The network manager needs to configure each
user to access the Internet via the proxy server IP.
• Users can bypass the proxy without supervision.
Solution to the Problem: With cache redirection, the
user does not need to configure anything: they are
convinced they are accessing the Internet directly.
Load balancing identifies Internet traffic, “snatches”
the packets and diverts them to the cache.
The cache server checks whether the information is
in its memory. If it is not, the cache server retrieves
the information from the Internet. The cache server
then forwards the information to the user who is
unaware of the entire process.
It is possible to deploy more than one server and
to implement Server load balancing among them
(see also Application 1—Server Load Balancing).
6
WAN Router
a.com
b.com
c.com
...
c.com
a.com
Internet
Avaya P333R-LB
Gigabit Ethernet
with LAG Avaya P333R-LB
VRRP
LAN
b.com
Cache
Server Farm Cache Server Farm

Avaya P333R-LB
Cache
Server
Farm
Firewall
Firewall
Router
Server
Farm
“Gold”
“Silver”
Internet
“Gold”
“Silver”
WAN Router
Avaya P333R-LB
LAN
Internet
Cache
Server Farm Server Farm
“Gold”
“Silver”
Communication without boundaries
Application 4—Policy-Based
Load Balancing
Policy-based load balancing allows you to provide
tiered services. Preferred users receive better
service/access and you can charge those customers
according to the level of service they receive. This
process of maximizing the use of current resources
is transparent to the user.
Implementation: The first stage is to define groups
of users and servers (gold, silver, etc.). The servers
provide identical services. “Gold” users are directed
to the “Gold” servers, “Silver” users are directed
to the “Silver” servers, etc.
Advantages: Other load balancers on the market
redirect traffic based only on a division of users and
servers to “Gold” and “Silver” groups. However,
Avaya’s policy-based load balancing gives the best
use of current resources by continuously checking
the status of the “Gold” and “Silver” groups. If the
“Gold” group is more loaded than the “Silver”
group then the “Gold” user will receive service
from the “Silver” group. Thus “Gold” users will
always receive the best service.
Policy-Based Load Balancing and Service Providers
The same application can be implemented with
cache servers, so Service Providers can give the
best service to preferred customers. The customers
only notice improved service—without any
reconfiguration and complications.
7
• Users receive
service according
to their level:
“gold,” “silver”
or “bronze.”

© 2001 Avaya Inc.
All rights reserved.
Printed in the U.S.A.
05/01 •EF-EIS1417 avaya.com
For a contact in your area, go to: www.avaya.com/contactus