Nomadix AG 2300, Руководство пользователя, Страница: 40 / 294

A
CCESS
G
ATEWAY
Introduction 28
Link Availability Detection Method and Time
Load balancing and failover requires some form of monitoring of each ISP link to determine its availability
for executing load balancing and failover decisions. Generally, link monitoring is accomplished by two
different methods:
1.
Periodic probing of predefined hosts using HTTP or ICMP ping requests.
2.
Periodic DNS queries to the DNS servers provided by each ISP.
The period between successive link tests is usually configured, and is typically set to between 30 seconds and
60 seconds. This represents the maximum time for which a user will remain connected to a failed ISP
connection before being re-routed to a working ISP link in an ISP failure scenario.
Traffic Balancing and Weighting
Load balancers have some form of weighting of traffic between links to achieve a desired balance scenario.
With the Nomadix NSE, traffic is balanced by individual subscriber numbers, and weighted according to the
speed of the ISP connected to each port. For example, if an NSE has 2 x 10M links connected and currently
has 100 active subscribers, then 50 users would be connected to each link. If the ISP links were 10 Mbps and
40Mbps, then 20 users would be connected to the 10M link and 80 users to the 40M link, and so on.
Load Rebalancing upon Link Recovery
Load balancing and failover with well-configured link availability detection provides fast and effective
recovery from ISP link failure occurrences. Additional consideration must be made as to what actions should
be taken when a failed ISP link recovers. The Nomadix approach is to rebalance as the ISP links change, thus
making sure the maximum level of service is always provided. There is a small yet important waiting time to
ensure changing links is kept to a minimum.
Load Balancing and Failure Considerations
1.
Is load balancing or just ISP failover required?
2.
Is aggregation of multiple low-speed links required?
3.
How reliable are different local ISP services?
4. What are the relative costs of different ISP services?
5.
Do ISP links need to be shared between guest and back-office users?
6.
Is there a requirement to have certain users connected to a particular ISP?
1. It may be a requirement to provide just a backup service to the primary ISP service in the case that the main
HSIA ISP fails. The backup service may be on a pay-to-use basis through a 3G or 4G wireless modem, or
be a low-cost, lower-tier service, such as a cable modem service, that is only used when the main ISP link is
down, on the basis that providing a reduced HSIA service is better than no service at all when the main ISP
link is down. Alternatively, the organization may have multiple ISP links, and wants to be able to fully
utilize all of them under normal conditions. The Nomadix NSE supports both failover only and combined
load balancing with failover.
2. In some instances, suitable high-speed internet services required to meet the aggregate needs of the
organization may not be available or are simply too expensive. In this case it may be desirable to aggregate
multiple lower-cost, lower-speed lines together. The Nomadix AG2400 and AG5600 can aggregate services
from up to three ISP links; the AG5800 and AG5900 can handle up to five links.
3. It is important to consider the relative quality of each ISP link. If a second link is much lower quality than
the main ISP link, then it should only be used as a back-up link in failover mode, and not in a load-balanced
environment. If the quality of the links is much the same, then load balancing with failover should be used.
4. It is important to consider the relative cost of links. If all links have a fixed monthly charge, then ideally
they should be used in a load-balanced mode, so that costly links are not sitting unused most of the time.