On 16 March 2015 at 11:07, Berthold Gunreben <[email protected]> wrote:

>
> - It gives useful, or at least nice error pages when a service is down.
>

I hope you mention that just for giggles, right? ;-)   HA proxy when the
service can have long enough outages to care...


> If you want to have a load balancer, I also would recommend to have a
> closer look at latencies. I don't know if the Mainframe is the correct
> machine to perform this type of task. This would be something that Rob
> would be better to comment (well, he already told his opinion).
>
> Wish it were that simple. With Linux on z/VM, most of us are not in the
billion hits per day business, so criteria are different. When you're using
a shared resource environment, you should not be too picky about latency.
After all, latency is the only thing you can sacrifice to achieve higher
resource efficiency. When you're willing to wait for a few ms rather than
get serviced instantly, you can share with others. The challenge is the
long tail in the distribution of response times. When your end user
experience involves a lot of transactions, you need to be careful not to
run into one of the rare slow ones each time.

There may be good reasons to need a load balancer. Just as there may be
very good reasons to move the content to a CDN or use dedicated firewall
devices. Some these environments are complicated enough that you should not
insist in doing it all yourself with Linux on z/VM. Focus on scenario's
where you simplify infrastructure by virtualizing it, not complicate it.

Rob van der Heij
http://www.velocitysoftware.com/

----------------------------------------------------------------------
For LINUX-390 subscribe / signoff / archive access instructions,
send email to [email protected] with the message: INFO LINUX-390 or visit
http://www.marist.edu/htbin/wlvindex?LINUX-390
----------------------------------------------------------------------
For more information on Linux on System z, visit
http://wiki.linuxvm.org/

Reply via email to