[124153] in North American Network Operators' Group
=?iso-8859-1?Q?Experiences_with_A10_AX_series_Load_Balancers=3F=20?=
daemon@ATHENA.MIT.EDU (Babak Pasdar)
Thu Mar 25 08:27:50 2010
To: Bryan.Welch@arrisi.com
From: "Babak Pasdar" <bpasdar@batblue.com>
Date: Thu, 25 Mar 2010 08:30:04 -0400
Cc: nanog@nanog.org
Errors-To: nanog-bounces+nanog.discuss=bloom-picayune.mit.edu@nanog.org
Bryan,
We have built out our infrastructure around A10 and after doing our own =
due diligence on them, we like them for the following reasons:
1. The performance is outrageous (in a good way) in terms of:
a. Total number of sessions supported
b. Ramp Rate (New connections per second)
c. Throughput, up to 37+ gig in our tests
Quite frankly A10 is the only company we found that could help us delive=
r on a secure 100Gig plus network that we designed for a client. Their =
specs were 100Gig today with scalability to 200Gig. The AX-5200 deliver=
ed as advertised for us. F5 didn't even come close to scaling to these =
levels in our testing.
2. The interface is reasonably decent, but could stand some refinement. =
The aPolicy is for the most part a one-for-one replacement of the F5 TC=
L iPolicy. So transitioning should be pretty straight -forward.
3. The company is a good company with good support. Lee Chen, one of th=
e founders of Foundry Networks started A10. I went to California and me=
t with him and his engineering team to see the type of folks we would be=
engaging with. They were extremely pleasant to deal with, unlike a lot=
of the other companies out there today. =20
The foundation of the system is that they came up with a process to make=
Parallelization far more efficient than the standard high-performance O=
Ses like Linux and Unix. This lets them throttle through scenarios that=
bog down other systems. We have seen first-hand how the systems perfor=
mance has addressed DNS DoS scenarios where F5 fell down on the job.=20
4. Finally, the cost is far more pleasant than the nickle and diming tha=
t F5 does for this license and that license. This is the one area F5 s=
hould really get dinged on. I don't at all care for the In our scenario=
it came in at roughly 40% less than F5.
All-in-all my advice is you can't go wrong with the A10 unless there is =
some specific feature that is missing. We have not found any for our pu=
rposes. Quite frankly I feel that many people are unnecessarily afraid =
to migrate away from F5. Not that F5 is a bad product or company, but c=
ompetition is good and A10 is a good competitor. We are very pleased w=
ith our decision to go with A10.=20
I hope this is useful feedback for you.
Best Regards,
Babak
--
Babak Pasdar
President & CEO | Certified Ethical Hacker
Bat Blue Corporation | Integrity . Privacy . Availability
(p) 212.461.3322 x3005 | (f) 212.584.9999 | (w) www.BatBlue.com
Receive Bat Blue's Daily Security Intelligence Report
Bat Blue's AS: 25885 | BGP Policy | Peering Policy=20
Bat Blue's Legal Notice
Reducing IT Security Budget, Burden & Risk - Video | Article
> -----Original Message----- =20
> From: Welch, Bryan [mailto:Bryan.Welch@arrisi.com] =20
> Sent: Tuesday, March 23, 2010 8:35 PM =20
> To: nanog@nanog.org =20
> Subject: Experiences with A10 AX series Load Balancers=3F =20
> =20
> Does anyone have any experiences good/bad/indifferent with this comp=
any and =20
> their products=3F They claim 2x the performance at =3F the cost and=
am a bit =20
> leery as you can imagine. =20
> =20
> We are looking to replace our aging F5 BigIP LTM's and will be evalu=
ating =20
> these along with the Netscaler and new generation F5 boxes. =20
> =20
> =20
> =20
> =20
> Regards, =20
> =20
> Bryan =20
> =20
> =20
> =20
