scalability of hobbit/xymon ?
Hi,
In my company, we are currently monitoring ~7000 devices (unix, microsoft, network, firewall, storage) in 13 datacenters in different countries (Spain, Italy, Germany, Belgium, France, Amsterdam, Belgium, UK).
The architecture is simple: one hobbit cluster using Heartbeat in each datacenter (= 13 clusters) reporting to a central cluster by bbproxy and some custom scripts to concatenate/consolidate the bb-hosts files correctly.
My question is for the future, we need to support 20 000 devices and i would like to know if hobbit is able to do that. (scalability). Indeed, my company doesn’t trust that it’s possible, and they want to move to HP BSM or EMC Smarts .... :-(
Any experience about that is welcome.
Regards, Nico
Although I'm far from knowing what hobbit can do but doesn tthis paragraph
Quote : http://www.hswn.dk/hobbit/help/about.html *Xymon can handle monitoring lots of systems. *
*Big Brother is implemented mostly as shell-scripts, and performance suffers badly from this. In large networks where you need to monitor hundreds or thousands of hosts, processing of the data simply cannot keep up. Another problem with BB is that it stores all status-information in individual files; when you have lots of hosts and statuses, the amount of disk I/O triggered by this severely limits how many systems you can monitor with one BB server. Xymon avoids these performance bottlenecks by keeping most of the ever-changing data in memory instead of on-disk, and by being implemented in C rather than shell scripts.*
I'd imagine if the above statements is correct the power of the server and available memory is the only thing holding you back?
On Wed, Jul 8, 2009 at 2:31 PM, nico <nico at crysto.org> wrote:
Hi,
In my company, we are currently monitoring ~7000 devices (unix, microsoft, network, firewall, storage) in 13 datacenters in different countries (Spain, Italy, Germany, Belgium, France, Amsterdam, Belgium, UK).
The architecture is simple: one hobbit cluster using Heartbeat in each datacenter (= 13 clusters) reporting to a central cluster by bbproxy and some custom scripts to concatenate/consolidate the bb-hosts files correctly.
My question is for the future, we need to support 20 000 devices and i would like to know if hobbit is able to do that. (scalability). Indeed, my company doesn’t trust that it’s possible, and they want to move to HP BSM or EMC Smarts .... :-(
Any experience about that is welcome.
Regards, Nico
To unsubscribe from the hobbit list, send an e-mail to hobbit-unsubscribe at hswn.dk
To resolve this issue of unknown capacity, I think we need to come up a test/torture script for xymon hobbit server. This way we can demonstrate/simulate the load capacity of xymon server before deployment.
I have some idea on how to do the coding(using perl's test module), it will be great if your company can allocate resource to do this programming task together. otherwise I am interested to do it myself but a lower priority.
T.J. Yang
Date: Wed, 8 Jul 2009 15:00:26 +0100 From: tainted.soul69 at googlemail.com To: hobbit at hswn.dk Subject: Re: [hobbit] scalability of hobbit/xymon ?
Although I'm far from knowing what hobbit can do but doesn tthis paragraph
Quote : http://www.hswn.dk/hobbit/help/about.html
Xymon can handle monitoring lots of systems.
Big Brother is implemented mostly as shell-scripts, and performance suffers badly from this. In large networks where you need to monitor hundreds or thousands of hosts, processing of the data simply cannot keep up. Another problem with BB is that it stores all status-information in individual files; when you have lots of hosts and statuses, the amount of disk I/O triggered by this severely limits how many systems you can monitor with one BB server.
Xymon avoids these performance bottlenecks by keeping most of the ever-changing data in memory instead of on-disk, and by being implemented in C rather than shell scripts.
I'd imagine if the above statements is correct the power of the server and available memory is the only thing holding you back?
On Wed, Jul 8, 2009 at 2:31 PM, nico wrote:
Hi,
In my company, we are currently monitoring ~7000 devices (unix, microsoft, network, firewall, storage) in 13 datacenters in different countries
(Spain, Italy, Germany, Belgium, France, Amsterdam, Belgium, UK).
The architecture is simple: one hobbit cluster using Heartbeat in each datacenter (= 13 clusters) reporting to a central cluster by bbproxy and
some custom scripts to concatenate/consolidate the bb-hosts files correctly.
My question is for the future, we need to support 20 000 devices and i would like to know if hobbit is able to do that. (scalability).
Indeed, my company doesn’t trust that it’s possible, and they want to move to HP BSM or EMC Smarts .... :-(
Any experience about that is welcome.
Regards, Nico
To unsubscribe from the hobbit list, send an e-mail to
hobbit-unsubscribe at hswn.dk
Hotmail® has ever-growing storage! Don’t worry about storage limits. http://windowslive.com/Tutorial/Hotmail/Storage?ocid=TXT_TAGLM_WL_HM_Tutoria...
Thanks for the proposition. The hobbit clusters are in production but i think i can run some scripts on a standby node. (morever we got preprod).
It'll be a pleasure to try your scripts and share results with the ML.
Regards, Nico
On 08/07/09 mercredi 8 18:41, "T.J. Yang" <tj_yang at hotmail.com> wrote:
To resolve this issue of unknown capacity, I think we need to come up a test/torture script for xymon hobbit server. This way we can demonstrate/simulate the load capacity of xymon server before deployment.
I have some idea on how to do the coding(using perl's test module), it will be great if your company can allocate resource to do this programming task together. otherwise I am interested to do it myself but a lower priority.
T.J. Yang
Date: Wed, 8 Jul 2009 15:00:26 +0100 From: tainted.soul69 at googlemail.com To: hobbit at hswn.dk Subject: Re: [hobbit] scalability of hobbit/xymon ?
Although I'm far from knowing what hobbit can do but doesn tthis paragraph
Quote : http://www.hswn.dk/hobbit/help/about.html
Xymon can handle monitoring lots of systems.
Big Brother is implemented mostly as shell-scripts, and performance suffers badly from this. In large networks where you need to monitor hundreds or thousands of hosts, processing of the data simply cannot keep up. Another problem with BB is that it stores all status-information in individual files; when you have lots of hosts and statuses, the amount of disk I/O triggered by this severely limits how many systems you can monitor with one BB server.
Xymon avoids these performance bottlenecks by keeping most of the ever-changing data in memory instead of on-disk, and by being implemented in C rather than shell scripts.
I'd imagine if the above statements is correct the power of the server and available memory is the only thing holding you back?
On Wed, Jul 8, 2009 at 2:31 PM, nico wrote:
Hi,
In my company, we are currently monitoring ~7000 devices (unix, microsoft, network, firewall, storage) in 13 datacenters in different countries
(Spain, Italy, Germany, Belgium, France, Amsterdam, Belgium, UK).
The architecture is simple: one hobbit cluster using Heartbeat in each datacenter (= 13 clusters) reporting to a central cluster by bbproxy and
some custom scripts to concatenate/consolidate the bb-hosts files correctly.
My question is for the future, we need to support 20 000 devices and i would like to know if hobbit is able to do that. (scalability).
Indeed, my company doesn¹t trust that it¹s possible, and they want to move to HP BSM or EMC Smarts .... :-(
Any experience about that is welcome.
Regards, Nico
To unsubscribe from the hobbit list, send an e-mail to
hobbit-unsubscribe at hswn.dk
Hotmail® has ever-growing storage! Don¹t worry about storage limits. http://windowslive.com/Tutorial/Hotmail/Storage?ocid=TXT_TAGLM_WL_HM_Tutoria... Storage_062009 To unsubscribe from the hobbit list, send an e-mail to hobbit-unsubscribe at hswn.dk
participants (3)
-
nico＠crysto.org
-
tainted.soul69＠googlemail.com
-
tj_yang＠hotmail.com