Wade,
I'm responding to the list as well so that other folks can get the
benefit of this conversation.
We use pfSense in a CARP cluster configuration for load balancing
because we cannot afford any downtime. Our server pool is tied to a
virtual server address which is actually a CARP type VIP shared between
our two pfSense boxes. This configuration works very well for us and
helped us replace two Cisco LocalDirector boxes saving us a whole bunch
of rack space, and giving us access to far more detailed reporting of
what was going on with that portion of our network.
-Gary
Wade Blackwell wrote:
Thanks Gary,
That is precisely what I will be using the LB for as well. If we go
this route we will be purchasing support through Centipede as well. Another
question, when you create the LB pools and virtual servers do you use VIPS
or the interface IPs on the PF? I am messing around with this at the moment
and am having some trouble. Thanks.
Wade B
Wade Blackwell
"Integrity is often more painful and always more profitable than perception
management"
-----Original Message-----
From: Gary Buckmaster [mailto:[EMAIL PROTECTED]
Sent: Wednesday, April 16, 2008 2:51 PM
To: support@pfsense.com
Subject: Re: [pfSense Support] Who has some good numbers to share for load
balancing?
Wade,
We use pfSense to load balance connections to our content filtering
database. Daily we get approximately 40 million connections with a peak
rate of close to 3Mb/s to a pool of 20 servers and our application requires
latency to be very minimal. Up until recently we ran that using 1.2Ghz
celeron boxes and they were perfectly capable of handling the load. Your
bottleneck will be your server pool's ability to process the connections
before you ever start reaching the abilities of the load balancers to handle
the traffic.
-Gary
Wade Blackwell wrote:
Good afternoon PFsense fans,
Greetings from the sunny central cost of California. I am
currently pricing out several load balancer solutions. The
requirements are pretty basic;
-Redundancy (CARP)
-Sticky
-intelligent load balancing of TCP services (fail a load balanced
node/server out of the pool when the service fails) -ability to
manually pull nodes out of the pool for maintenance without affecting
customers
So I know that PF supports all of these requirements and is a
good inexpensive candidate for the project. What I am now trying to
get a handle on is what can I expect for connections/sec? The proposed
hw platform for the PF's is;
CPU: Intel Pentium E2140 Dual-Core 1.60GHz, 1MB L2 Cache, 800MHz
LGA775
RAM: 1GB (2 x 512MB) Unbuffered ECC DDR2-667
NIC: Dual 10/100/1000 Mbps NICs (Intel 82573L + 82573V) - Integrated
PCIe x8: Intel PRO/1000 PT Dual Port Server Adapter - 2 x GbE (RJ45) -
PCIe x4 Fixed Drive - 1: 160GB Western Digital RE (3.0Gb/s, 7.2Krpm,
16MB Cache) SATA
Anyone on the list have some benchmarks they could pass along? TIA.
-W
---------------------------------------------------------------------
To unsubscribe, e-mail: [EMAIL PROTECTED] For additional
commands, e-mail: [EMAIL PROTECTED]
---------------------------------------------------------------------
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]