[176615] in North American Network Operators' Group
Re: Juniper MX Sizing
daemon@ATHENA.MIT.EDU (Youssef Bengelloun-Zahr)
Fri Dec 5 19:15:00 2014
X-Original-To: nanog@nanog.org
From: Youssef Bengelloun-Zahr <youssef@720.fr>
In-Reply-To: <168CDB82-08FE-4947-95A3-FFD773B06EEB@gmail.com>
Date: Sat, 6 Dec 2014 01:14:50 +0100
To: Brad Fleming <bdflemin@gmail.com>
Cc: "nanog@nanog.org" <nanog@nanog.org>
Errors-To: nanog-bounces@nanog.org
Hi,
Running MLXe with MR2 and/or CER-RT as MPLS PEs depending on POP size. We al=
so run the later as route reflectors.
They behave beautifully when it comes to churning BGP full feeds, convergenc=
e is around 30-45s (full RAM). Routing capacity is also amazing.
I'm particularly amazed by the CER-RT from a price/performance/footprint per=
spective. So I would advice it unless the OP has some specific technical req=
uirements (flowspec support, etc.).
Best regards.
> Le 5 d=C3=A9c. 2014 =C3=A0 22:52, Brad Fleming <bdflemin@gmail.com> a =C3=A9=
crit :
>=20
> We have both Brocade CER and XMR (predecessor to the MLXe) in our environm=
ent today. We find both platforms attractive from a price and power consumpt=
ion standpoint. They will both handle the IPv4 and IPv6 unicast routing tabl=
es today.* The MLXe with MR2 cards is quite a formidable box; lots of power a=
nd pretty light-weight OS (compared to Junos). We found our XMR nodes with o=
riginal mgmt cards and Gen1 line cards converge pretty quick; we=E2=80=99ve n=
ever timed one officially but my gut feeling is RIB+FIB convergence is rough=
ly 45sec assuming your peer is RTT nearby. The CER is a little slower to con=
verge in our experience; however, we have them in non-critical portions of t=
he network so I can=E2=80=99t really attest to their convergence performance=
. Sorry.. not much in the way of lab readings for our Brocade gear.
>=20
>=20
>=20
>> On Dec 5, 2014, at 2:09 PM, Ammar Zuberi <ammar@fastreturn.net> wrote:
>>=20
>> What=E2=80=99s a cheaper alternative to the MX104s?
>>=20
>> We take a full BGP table and are on the AMS-IX and DE-CIX and are looking=
for a new router. The MX series looks a bit out of budget but we=E2=80=99re=
currently looking into the Brocade MLX series. We push under 10Gbps, but we=
do need 10Gbps routing due to capacity issues during attacks.
>>=20
>> Sorry for being a bit off-topic here.
>>=20
>> Ammar
>>=20
>> This email and any files transmitted with it are confidential and intende=
d solely for the use of the individual or entity to whom they are addressed.=
If you have received it by mistake, please let us know by e-mail reply and d=
elete it from your system; you may not copy this message or disclose its con=
tents to anyone. Please note that any views or opinions presented in this em=
ail are solely those of the author and do not necessarily represent those of=
the company. Finally, the recipient should check this email and any attachm=
ents for the presence of viruses. The company accepts no liability for any d=
amage caused by any virus transmitted by this email.
>>=20
>>> On Dec 6, 2014, at 12:01 AM, Brad Fleming <bdflemin@gmail.com <mailto:bd=
flemin@gmail.com>> wrote:
>>>=20
>>> Then you should look for something other then the MX104.
>>>=20
>>> In our testing an MX104 running Junos 13.3R4 with a single, full feed to=
ok about 4min 25sec to (1) converge the RIB from a router sitting 0.5ms RTT a=
way and (2) update the FIB with all entries. This performance was observed w=
ith single RE and dual RE and without any excess services running. If we add=
ed inline-flow sampling to the device full convergence took closer to 5min 4=
5sec in our lab. Efforts to bring the convergence time down (without filteri=
ng ingress advertisements) with the assistance of JTAC proved unsuccessful.
>>>=20
>>> We decided to =E2=80=9Cbite the bullet=E2=80=9D and procure MX480s inste=
ad but obviously that=E2=80=99s not possible for everyone. If the MX480 is o=
ut of the question a Brocade CER Premium is an option. We have 3 in producti=
on and see very attractive convergence times; however, they have a more limi=
ted feature set and you=E2=80=99ll want to understand how their FIB memory s=
cales. Apologies, I don=E2=80=99t know the Cisco equivalent from the ASR lin=
e these days but I=E2=80=99m sure others on the list could help out.
>>>=20
>>>=20
>>>> On Dec 5, 2014, at 11:45 AM, Graham Johnston <johnstong@westmancom.com <=
mailto:johnstong@westmancom.com>> wrote:
>>>>=20
>>>> Shawn,
>>>>=20
>>>> It's more about FIB than RIB as I am concerned about the time it takes u=
ntil MPCs have updated route information after large scale changes in routes=
learned via BGP.
>>>>=20
>>>> Graham Johnston
>>>> Network Planner
>>>> Westman Communications Group
>>>> 204.717.2829
>>>> johnstong@westmancom.com <mailto:johnstong@westmancom.com>
>>>> =EF=81=90=EF=80=A0think green; don't print this email.
>>>>=20
>>>> -----Original Message-----
>>>> From: Shawn Hsiao [mailto:phsiao@tripadvisor.com]=20
>>>> Sent: Friday, December 05, 2014 11:30 AM
>>>> To: Graham Johnston
>>>> Cc: nanog@nanog.org
>>>> Subject: Re: Juniper MX Sizing
>>>>=20
>>>>=20
>>>> Is your sizing concern just for the RIB, or also for FIB to sync up? T=
he latter was a problem for us, but not the former. We also have inline-jf=
low turned on and that is still a work-in-progress in terms of impacting per=
formance.
>>>>=20
>>>> We are using MX104 for similar purposes for many months now, and with s=
ome tweaks in our procedures and configurations we found it to be acceptable=
. MX104 may not be able to process routes as fast as MX480, but MX480 is a=
lso not instantaneous either so similar risks exist.
>>>>=20
>>>>=20
>>>>=20
>>>>> On Dec 5, 2014, at 11:59 AM, Graham Johnston <johnstong@westmancom.com=
> wrote:
>>>>>=20
>>>>> I am wondering if anyone can provide their real world experience about=
sizing Juniper MX routers as it relates to BGP. I am needing a device that=
has a mix of layer 2 and 3 features, including MPLS, that will have a very l=
ow port count requirement that will primarily be used at a remote POP site t=
o connect to the local IX as well as one or two full route transit providers=
. The MX104 has what I need from a physical standpoint and a data plane sta=
ndpoint, as well as power consumption figures. My only concern is whether t=
he REs have enough horsepower to churn through the convergence calculations a=
t a rate that operators in this situation would find acceptable. I realize t=
hat 'acceptable' is a moving target so I would happily accept feedback from p=
eople using them as to how long it takes and their happiness with the produc=
t.
>>>>>=20
>>>>> For those of you that deem the MX104 unacceptable in this kind of role=
and moved up to the MX240, what RE did you elect to use?
>>>>>=20
>>>>> Thanks,
>>>>> Graham Johnston
>>>>> Network Planner
>>>>> Westman Communications Group
>>>>> 204.717.2829
>>>>> johnstong@westmancom.com<mailto:johnstong@westmancom.com>
>>>>> P think green; don't print this email.
>=20
