[squid-users] Squid Benchmark

Alex Rousskov rousskov at measurement-factory.com
Sat Sep 28 18:00:09 UTC 2019


On 9/28/19 12:59 PM, Emmanuel BILLOT wrote:

> This is probably a bring recurring subject, but users in our company
> frequently say that "Internet navigating is slow..."
> While "slow" is such a relative feeling, depending on computer health,
> user behaviour or network state, we have to be aware for bottlenecks.
> 
> We are using Squid servers behind a LB, providing Internet or Intranet
> access for up to 3000 users.
> 
> NAT is for sure a faster way to access web, but considering security and
> filtering, we have to use Squid and we want to be sure that Squid is not
> the main cause for slow-down.
> 
> Is there a strong way, and moreover an efficient tool to check that
> using Squid is not a bottleneck compared to a direct to Internet access
> (NAT) ?

The answer to your question does not require a tool: Squid does slow
things down in most modern "TLS everywhere" environments, possibly
including yours. And if you do not decrypt traffic, then the answer is
likely even simpler: Squid slows down your traffic.

However, you are asking the wrong question. Since you _have to_ use
Squid (as you have said), the questions to ask are:

1. Does Squid slow things down right now more than it usually does? You
can answer that question by collecting and plotting various performance
stats from access logs and cache manager. You will see (or, better,
automatically detect!) when things suddenly go south.

2. Can we make Squid faster? The answer to that question is usually
"yes", but it requires significant investment in bottleneck analysis
and, in some cases, development/optimization.

Alex.


More information about the squid-users mailing list