Hello team
what is best solution for tunning performance
like different Hardrive for MySQL/RRD and System?
What about performance SSD Raid or RAM Disk?
We are using VMware, with iSCSI SAS and SSD , for System SAS drive for RRD/MySQL two drive from SSD Raid
any idea? like HW Raid without iSCSI?
Now we have more than 500 devices, 4000 ports and 3000 sensors
Tom
[email-header-praha1]
Tomas Chott technický ředitel chott@praha1.netmailto:chott@praha1.net Phone: +420 245 004 005 Mobile: +420 733 123 135
Metropolitní síť Praha 1 z.s.p.o. Jindrisska 18 110 00 Praha 1 Czech Republic www.praha1.nethttp://www.praha1.net [Popis: Facebook]http://www.facebook.com/praha1net http://www.linkedin.com/pub/tomas-chott/49/97a/1aa
4000 ports and 3000 sensors is no problem at all for a server with a SSD. Use distributed pollers to just scale up the cpu that you need with small VMs
/Markus
2017-02-19 20:10 GMT+01:00 chott@praha1.net chott@praha1.net:
Hello team
what is best solution for tunning performance
like different Hardrive for MySQL/RRD and System?
What about performance SSD Raid or RAM Disk?
We are using VMware, with iSCSI SAS and SSD , for System SAS drive for RRD/MySQL two drive from SSD Raid
any idea? like HW Raid without iSCSI?
Now we have more than 500 devices, 4000 ports and 3000 sensors
Tom
[image: email-header-praha1]
*Tomas Chott technický ředitel*
chott@praha1.net Phone:
- +420 245 004 005 <+420%20245%20004%20005> *Mobile:* +420 733 123 135
<+420%20733%20123%20135>*
*Metropolitní síť Praha 1 z.s.p.o.* Jindrisska 18 110 00 Praha 1 Czech Republic www.praha1.net
[image: Popis: Facebook] http://www.facebook.com/praha1net http://www.linkedin.com/pub/tomas-chott/49/97a/1aa
observium mailing list observium@observium.org http://postman.memetic.org/cgi-bin/mailman/listinfo/observium
For your comparison, my installation shows as follows:
Statistics
DB size
2.31GB
RRD size
29.7GB
Devices
762
Ports
9863
IPv4 Addresses
1641
IPv4 Networks
227
IPv6 Addresses
53
IPv6 Networks
19
Services
0
Applications
78
Processors
658
Memory pools
553
Storage Entries
380
Disk I/O Entries
303
HR-MIB Entries
2850
Entity-MIB Entries
12279
Syslog Entries
6053897
Eventlog Entries
788703
Sensors
3574
Printer Supplies
15
Netscaler VServers
0
Netscaler Services
0
Virtual Machines
69
IP SLAs
19
My observium VM has 8 3.16GHz cores, 16GB of RAM, and the ESX datastore is using a NFS partition on a ZFS array of several 750GB SATA disks with an Intel 750 NVMe SSD acting as the SLOG for the array. I have the poller wrapper configured for 32 threads and on average, it takes roughly 92 seconds to complete the polling of all 762 devices. I didn't have to do any special performance tuning other than the increased poller threads in order to keep the polling time well under 300 seconds.
The greatest performance increase I had gained over a year ago was when I had changed the ESX datastore from iSCSI to NFS, and adding the very fast ZFS SLOG device.
.Ron
From: observium [mailto:observium-bounces@observium.org] On Behalf Of chott@praha1.net Sent: Sunday, February 19, 2017 1:11 PM To: observium@observium.org Subject: [Observium] Observium tunning
Hello team
what is best solution for tunning performance
like different Hardrive for MySQL/RRD and System?
What about performance SSD Raid or RAM Disk?
We are using VMware, with iSCSI SAS and SSD , for System SAS drive for RRD/MySQL two drive from SSD Raid
any idea? like HW Raid without iSCSI?
Now we have more than 500 devices, 4000 ports and 3000 sensors
Tom
Tomas Chott technický ředitel
chott@praha1.net mailto:chott@praha1.net Phone: +420 245 004 005 Mobile: +420 733 123 135
Metropolitní síť Praha 1 z.s.p.o. Jindrisska 18 110 00 Praha 1 Czech Republic www.praha1.net http://www.praha1.net
http://www.facebook.com/praha1net
I'd recommend reducing poller threads to ~16 or so.
No use spiking I/O load for 1/3rd of the time!
Adam.
Sent from BlueMail
On 19 Feb 2017, 20:38, at 20:38, Ron Marosko ron@rjr-services.com wrote:
For your comparison, my installation shows as follows:
Statistics
DB size
2.31GB
RRD size
29.7GB
Devices
762
Ports
9863
IPv4 Addresses
1641
IPv4 Networks
227
IPv6 Addresses
53
IPv6 Networks
19
Services
0
Applications
78
Processors
658
Memory pools
553
Storage Entries
380
Disk I/O Entries
303
HR-MIB Entries
2850
Entity-MIB Entries
12279
Syslog Entries
6053897
Eventlog Entries
788703
Sensors
3574
Printer Supplies
15
Netscaler VServers
0
Netscaler Services
0
Virtual Machines
69
IP SLAs
19
My observium VM has 8 3.16GHz cores, 16GB of RAM, and the ESX datastore is using a NFS partition on a ZFS array of several 750GB SATA disks with an Intel 750 NVMe SSD acting as the SLOG for the array. I have the poller wrapper configured for 32 threads and on average, it takes roughly 92 seconds to complete the polling of all 762 devices. I didn't have to do any special performance tuning other than the increased poller threads in order to keep the polling time well under 300 seconds.
The greatest performance increase I had gained over a year ago was when I had changed the ESX datastore from iSCSI to NFS, and adding the very fast ZFS SLOG device.
.Ron
From: observium [mailto:observium-bounces@observium.org] On Behalf Of chott@praha1.net Sent: Sunday, February 19, 2017 1:11 PM To: observium@observium.org Subject: [Observium] Observium tunning
Hello team
what is best solution for tunning performance
like different Hardrive for MySQL/RRD and System?
What about performance SSD Raid or RAM Disk?
We are using VMware, with iSCSI SAS and SSD , for System SAS drive for RRD/MySQL two drive from SSD Raid
any idea? like HW Raid without iSCSI?
Now we have more than 500 devices, 4000 ports and 3000 sensors
Tom
Tomas Chott technický ředitel
chott@praha1.net mailto:chott@praha1.net Phone: +420 245 004 005 Mobile: +420 733 123 135
Metropolitní síť Praha 1 z.s.p.o. Jindrisska 18 110 00 Praha 1 Czech Republic www.praha1.net http://www.praha1.net
http://www.facebook.com/praha1net
-- This message has been scanned for viruses and dangerous content by http://www.efa-project.org E.F.A. Project, and is believed to be clean. Click here to report this message as spam. http://smtp.rjr-services.com/cgi-bin/learn-msg.cgi?id=2387110049D.A7717&tok en=4c0e07d4ac1328c78fa26915b1e42fb8
observium mailing list observium@observium.org http://postman.memetic.org/cgi-bin/mailman/listinfo/observium
Yeah, the 32 was from when it was running on an ESX box with Xeon X5355 processors. Newer ESX box has Xeon E5-2643v3 processors. :-)
From: observium [mailto:observium-bounces@observium.org] On Behalf Of Adam Armstrong Sent: Sunday, February 19, 2017 2:50 PM To: Observium Network Observation System observium@observium.org Cc: Observium Network Observation System observium@observium.org Subject: Re: [Observium] Observium tunning
I'd recommend reducing poller threads to ~16 or so.
No use spiking I/O load for 1/3rd of the time!
Adam.
Sent from BlueMail http://www.bluemail.me/r?b=8831
On 19 Feb 2017, at 20:38, Ron Marosko <ron@rjr-services.com mailto:ron@rjr-services.com > wrote:
For your comparison, my installation shows as follows:
Statistics
DB size
2.31GB
RRD size
29.7GB
Devices
762
Ports
9863
IPv4 Addresses
1641
IPv4 Networks
227
IPv6 Addresses
53
IPv6 Networks
19
Services
0
Applications
78
Processors
658
Memory pools
553
Storage Entries
380
Disk I/O Entries
303
HR-MIB Entries
2850
Entity-MIB Entries
12279
Syslog Entries
6053897
Eventlog Entries
788703
Sensors
3574
Printer Supplies
15
Netscaler VServers
0
Netscaler Services
0
Virtual Machines
69
IP SLAs
19
My observium VM has 8 3.16GHz cores, 16GB of RAM, and the ESX datastore is using a NFS partition on a ZFS array of several 750GB SATA disks with an Intel 750 NVMe SSD acting as the SLOG for the array. I have the poller wrapper configured for 32 threads and on average, it takes roughly 92 seconds to complete the polling of all 762 devices. I didn’t have to do any special performance tuning other than the increased poller threads in order to keep the polling time well under 300 seconds.
The greatest performance increase I had gained over a year ago was when I had changed the ESX datastore from iSCSI to NFS, and adding the very fast ZFS SLOG device.
…Ron
From: observium [mailto:observium-bounces@observium.org] On Behalf Of chott@praha1.net mailto:chott@praha1.net Sent: Sunday, February 19, 2017 1:11 PM To: observium@observium.org mailto:observium@observium.org Subject: [Observium] Observium tunning
Hello team
what is best solution for tunning performance
like different Hardrive for MySQL/RRD and System?
What about performance SSD Raid or RAM Disk?
We are using VMware, with iSCSI SAS and SSD , for System SAS drive for RRD/MySQL two drive from SSD Raid
any idea? like HW Raid without iSCSI?
Now we have more than 500 devices, 4000 ports and 3000 sensors
Tom
Tomas Chott technický ředitel
chott@praha1.net mailto:chott@praha1.net Phone: +420 245 004 005 Mobile: +420 733 123 135
Metropolitní síť Praha 1 z.s.p.o. Jindrisska 18 110 00 Praha 1 Czech Republic www.praha1.net http://www.praha1.net
participants (4)
-
Adam Armstrong -
chott＠praha1.net -
Markus Klock -
Ron Marosko