Performance - was Re: LDAP Groups to Freeradius and then Ruckus Wireless?
Hi Arran, To be fair, I have read a lot of performance problems in FreeRadius, and I have yet to see any. We have been running here for more than a year now, Debian 7/FreeRadius from 2.1.12 to 2.2.5, MySQL, detail files and AD authentication, roaming/proxying too, in a virtual machine with 64 bits, 1GB RAM and 2 CPUs, 3500 daily EDUROAM users, and I have yet to see the CPU loads go up more than 0.05-0.10. Even when I have major events that bring down the wifi intra-structure, like yesterdays where a switch went crazy, I hardly register an heavy load. What are your mileage, guys? Regards, Rui Ribeiro Regards,
Message: 4 Date: Tue, 10 Jun 2014 07:09:30 +0100 From: Arran Cudbard-Bell <a.cudbardb@freeradius.org> To: FreeRadius users mailing list <freeradius-users@lists.freeradius.org> Subject: Re: LDAP Groups to Freeradius and then Ruckus Wireless? Message-ID: <B5471695-B7D2-4970-B4C8-6A4905489FCC@freeradius.org> Content-Type: text/plain; charset="us-ascii"
On 10 Jun 2014, at 06:36, Fajar A. Nugraha <list@fajar.net> wrote:
On Mon, Jun 9, 2014 at 9:43 PM, Arran Cudbard-Bell <a.cudbardb@freeradius.org> wrote: ......
Would it make sense to enable the flag on all bundled package recipes? IIRC 2.x debian/rules has it (official debian/ubuntu package has it as well), but 3.0.x doesn't. Haven't checked others though.
No, there's a noticeable performance impact. One of the other users said it was as much as 15%.
-Arran
Arran Cudbard-Bell <a.cudbardb@freeradius.org> FreeRADIUS Development Team *************************************************
On 10 Jun 2014, at 11:46, Rui Ribeiro <ruyrybeyro@gmail.com> wrote:
Hi Arran,
To be fair, I have read a lot of performance problems in FreeRadius, and I have yet to see any. We have been running here for more than a year now, Debian 7/FreeRadius from 2.1.12 to 2.2.5, MySQL, detail files and AD authentication, roaming/proxying too, in a virtual machine with 64 bits, 1GB RAM and 2 CPUs, 3500 daily EDUROAM users,
3500 users (over a day) is on the small side in terms of FreeRADIUS deployments. Your concurrent users are probably a quarter of that.
and I have yet to see the CPU loads go up more than 0.05-0.10. Even when I have major events that bring down the wifi intra-structure, like yesterdays where a switch went crazy, I hardly register an heavy load.
What are your mileage, guys?
Try authenticating 2 Million subscribers after a major BNG outage. That 15% will make a big difference. The Eduroam community, whilst one of the most visible users of FreeRADIUS, probably only represents about 10% of the users authenticated daily using FreeRADIUS, possibly even less. -Arran Arran Cudbard-Bell <a.cudbardb@freeradius.org> FreeRADIUS Development Team FD31 3077 42EC 7FCD 32FE 5EE2 56CF 27F9 30A8 CAA2
3500 authentications is what some setups have to handle in like 20 seconds :) That said, databases are almost always the bottleneck when measuring capacity. FreeRADIUS memory usage seem really negligible (at least when some dumb user does not add a memory leak... duh). CPU usage can be quite intensive when handling heavy load. Still, that's not an issue, and it's laughably low compared to performance tests played on apache HTTP servers (for instance). -----Message d'origine----- De : freeradius-users-bounces+nicolas.chaigneau=capgemini.com@lists.freeradius.org [mailto:freeradius-users-bounces+nicolas.chaigneau=capgemini.com@lists.freeradius.org] De la part de Arran Cudbard-Bell Envoyé : mardi 10 juin 2014 13:19 À : FreeRadius users mailing list Objet : Re: Performance - was Re: LDAP Groups to Freeradius and then Ruckus Wireless? On 10 Jun 2014, at 11:46, Rui Ribeiro <ruyrybeyro@gmail.com> wrote:
Hi Arran,
To be fair, I have read a lot of performance problems in FreeRadius, and I have yet to see any. We have been running here for more than a year now, Debian 7/FreeRadius from 2.1.12 to 2.2.5, MySQL, detail files and AD authentication, roaming/proxying too, in a virtual machine with 64 bits, 1GB RAM and 2 CPUs, 3500 daily EDUROAM users,
3500 users (over a day) is on the small side in terms of FreeRADIUS deployments. Your concurrent users are probably a quarter of that.
and I have yet to see the CPU loads go up more than 0.05-0.10. Even when I have major events that bring down the wifi intra-structure, like yesterdays where a switch went crazy, I hardly register an heavy load.
What are your mileage, guys?
Try authenticating 2 Million subscribers after a major BNG outage. That 15% will make a big difference. The Eduroam community, whilst one of the most visible users of FreeRADIUS, probably only represents about 10% of the users authenticated daily using FreeRADIUS, possibly even less. -Arran Arran Cudbard-Bell <a.cudbardb@freeradius.org> FreeRADIUS Development Team FD31 3077 42EC 7FCD 32FE 5EE2 56CF 27F9 30A8 CAA2 This message contains information that may be privileged or confidential and is the property of the Capgemini Group. It is intended only for the person to whom it is addressed. If you are not the intended recipient, you are not authorized to read, print, retain, copy, disseminate, distribute, or use this message or any part thereof. If you receive this message in error, please notify the sender immediately and delete all copies of this message.
Rui Ribeiro wrote:
To be fair, I have read a lot of performance problems in FreeRadius, and I have yet to see any.
The complaints about performance problems are largely things like: - overloaded SQL databases - no indexes, - long-running administrative queries blocking FreeRADIUS - system overload - a small system being asked to authenticate 100K users in 10s - many PEAP authentications per second on a small system I've done tests with 1M accounts in the "users" file. The server takes 3 seconds to start, and then does 50K+ authentications per second. Which is about the same speed as when there's 10 accounts in the "users" file. There are no performance problems inside of FreeRADIUS. Alan DeKok.
participants (4)
-
Alan DeKok -
Arran Cudbard-Bell -
Chaigneau, Nicolas -
Rui Ribeiro