Where to Use the TaskSmart
Like other NICS products, the TaskSmart can simultaneously operate as
a forward, transparent, and reverse proxy. We tested the TaskSmart in
all three modes.
For transparent mode, we deployed the TaskSmart as an in-line router,
sitting between inside subnets and one outside subnet. Alternatively,
HTTP can be redirected to the TaskSmart with a layer 4 switch or
a WCCP (Cisco Web Cache Coordination Protocol) -enabled router (not
tested).
The TaskSmart supports ICP and CERN inter-cache protocols for workload
distribution in a hierarchical fashion between child, peer, and
parent caches. We configured our TaskSmart as an ICP child (see
right), forwarding requests to an ICP parent located at our
ISP.
TaskSmarts can be clustered for fault tolerance. Clusters
assign virtual IP addresses to specified services; requests sent to
the virtual IP are handled as long as any member of the cluster is
operational.We were unable to test fail-over without a second TaskSmart,
but sent requests to a virtual address we defined for a cluster consisting
of our active TaskSmart and a mythical standby.
In cluster mode, back-end server responses are re-written to ensure
cluster transparency. But clustering does not provide load balancing;
deploy the TaskSmart with a layer 4 switch if that's what you're seeking.
Compaq's Deployment Guide (available via FTP from Compaq's website)
outlines strategies for client (forward / transparent proxy) and server
(reverse proxy) acceleration. This guide is a "must read" for evaluating
where and how to use a cache. It identifies points of client acceleration
and aggregation that influence proxy placement, firewall considerations,
load balancing alternatives, and performance / fault-tolerance trade-offs.
Hints assist with TaskSmart scaling and cache cluster or hierarchy design.
According to Chuawiwat, Compaq is revising this Guide to add lessons learned
during the TaskSmart's first six months in the field.