Tuesday, September 27, 2011

#HPN updates #ProVision software with several high end features


Provision software release K.15.06.0006 has been posted on the web for download.  This release contains several key features for the ProVision switches such as

  •  BGP
  • PBR
  • Uplink Failure Detect
  • 6in4 tunneling
  • MLDv2



This is really huge for current and future customers as many of these features are only found in higher end products.  HP continues to include as many features as possible in our base software contrary to how most of our competitors charge you extra for.

https://h10145.www1.hp.com/downloads/SoftwareReleases.aspx?ProductNumber=J8697A&lang=en&cc=us&prodSeriesId=1827663

#RightNow is the best time ever to be a buyer of #ethernet switching as #HP drives costs down

HP is really making an impact by driving down costs of ethernet switching.  So says an article...

http://www.channelweb.co.uk/crn-uk/news/2111989/cisco-hp-turf-war-hits-switching-asps

Monday, September 26, 2011

#HPN 3800 Access Layer is better than #Cisco & #Juniper

Tolly just released a great report working through several comparison points comparing the HP 3800 series with the Cisco 3750x and Juniper EX4200.

Line rate from 64bytes to 9k frames
Latency from 3usec to 12usec  (from 50-100% faster than CSCO & JNPR)
Over Double the number of 10gb interfaces in a stack
From 8x to 25x more buffering capability than CSCO & JNPR
10% greater power efficiency than CSCO or JNPR in a stack or standalone
Over three years that efficient can save a customer from ~$300 to ~$800 per stack !!!

http://www.tolly.com/DocDetail.aspx?DocNumber=211127

#FCOE funny math

Check out this blog post from Cisco.

http://blogs.cisco.com/datacenter/3-for-2-the-fcoe-bandwidth-bonus/

It suggest that FCOE gives 50% greater bandwidth than 8gb FC.  While...theoretically true..its not in practice.

Here's why not..simple..not a long argument.  While I can transit FCOE at 10gb between server across a FCOE network... at the other end I can only hook up to 8gb FC compliant storage.  So..even if FCOE is more efficient and has more bandwidth, I can only talk to the end devices at 8gb.  This is especially true because FC is all about token credits..you can only send to me what I allow you to.  And...you cant buffer.  End result... no more throughput than your storage system has.

16gb FC is the next generation.  There are no converged adapters for 16gb FC yet.  So..for now that means 16gb FC is faster than 10gb FCOE.

Hence the problem with FCOE... its always going to be this disjointed architecture.

ISCSI today supports 10gb ethernet... it can utilize multi LAG 10gb...it can utilize 40gb.  It scales and it doesnt have the issues that FCOE has.

Thursday, September 22, 2011

#FlatterNetworks from #HP not only save on network costs, but server costs as well

I loved the articile from The Register talking about how networks need to become flatter.

Here's why... we already believe that networks have too many tiers.  Modern devices today can deliver both L2 & L3 functionality at a reasonable cost.  Also, we can deliver switching platforms with high density 1gb and 10gb ports.  The reason to have three or four tiers disappears.

The benefit to the customer... somewhat obvious.  You bought less equipment.  You have less oversubscription.  Your latency drops.

But..there are benefits that most dont realize.  Multi-tier architectures have greater latency...right?   With federated applications there is a great amount of server to server traffic.  Added latency in the network is multiplied and amplified.  sort of like that famous bridge in Washington that vibrated itself to death.

If I have 5-10ms of extra latency between devices..and a single user transaction means several servers working togetether...than every transaction can have many extra ms of latency.  The app starts to slow down.

But more imporant...what is a server doing while its waiting for return traffic from another server?  Nothing.  That means idle clock cycles on servers.. and less server utilization.  Multi-tier networks will require you to have more servers than you need.

And some customer are probably masking this problem as well by buying faster cpus, more cache, SSDs, or caching systems.  Costs also amplify we use other products to try to mask a bad network architecture.

   

#TheRegister weighs in on flatter networks

Tuesday, September 20, 2011

Overview of Cisco Nexus

#Cisco #UCS #Fallacy

http://www.networkworld.com/community/blog/cisco-claims-7400-ucs-customers-august?source=nww_rss

Does Cisco really have 7400 customers?  How many of those "customers" actually choose UCS?  and how many got some free units bundled into a networking deal?  How many of those 7400 bought UCS just to host CallManager?

My educated guess?  Less than 1,000 actually bought and paid real money to deploy UCS for a non-Cisco application.

Where's the vision #Cisco?

Cisco, what is your Vision?  Do you have one?  If you do...I think if you went to a shrink and tried to explain it..he would declare that you have multiple personalities.

Your latest announcement is VxLAN

http://newsroom.cisco.com/press-release-content?type=webcontent&articleId=467114

but how does a mac in udp tunnelling scheme jive with FCoE?

how does VxLAN jive with OTV?

how does VxLAN jive with FabricPath?

how does any of this jive with LISP?

I'm pulling my hair out thinking about it...


#TACACS on #HP #MSR #Router


=~=~=~=~=~=~=~=~=~=~=~= PuTTY log 2011.05.10 21:22:23 =~=~=~=~=~=~=~=~=~=~=~=
disp	
<HYD-HSA>display curr	
<HYD-HSA>display current-configuration 
#
 version 5.20, Release 1809P01
#
 sysname HYD-HSA
#
 domain default enable test
#
 telnet server enable
#
 dar p2p signature-file flash:/p2p_default.mtd
#
 port-security enable
#
 hwtacacs nas-ip 172.16.78.1
#
vlan 1
#
hwtacacs scheme test
 primary authentication (Tacacs Server IP)
 primary authorization (Tacacs Server IP)
 primary accounting (Tacacs Server IP)
 key authentication (Tacacs Server Key)
 key authorization (Tacacs Server Key)
 key accounting (Tacacs Server Key)
 user-name-format without-domain
#
domain system
 access-limit disable
 state active
 idle-cut disable
 self-service-url disable
domain test
 authentication login hwtacacs-scheme test
 authorization login hwtacacs-scheme test
 accounting login hwtacacs-scheme test
 access-limit disable
 state active
 idle-cut disable
 self-service-url disable
#
user-group system
#
local-user admin
 password cipher .]@USE=B,53Q=^Q`MAF4<1!!
 authorization-attribute level 3
 service-type telnet
  ---- More ----                #
interface Aux0
 async mode flow
 link-protocol ppp
#
interface Ethernet0/0
 port link-mode route
 ip address 10.237.60.2 255.255.255.252
#
interface Ethernet0/1
 port link-mode route
 ip address 172.16.78.1 255.255.255.192
#
interface Serial0/0
 link-protocol ppp
#
interface NULL0
#
 ip route-static 0.0.0.0 0.0.0.0 10.237.60.1
#
 load xml-configuration
#
 load tr069-configuration
  ---- More ----                #
user-interface aux 0
user-interface vty 0 4
 authentication-mode scheme
#
return
<HYD-HSA>

Is your network safe? #IPV6 might create new issues

Monday, September 19, 2011

#Tolly puts out a report on #HP #WLAN #MSM

http://tolly.com/DocDetail.aspx?DocNumber=211122

and reviews our current MSM controllers/software and 460/466 access points

#FCOE is the next #TokenRing


I think in 10 years we’re all going to be at a bar recounting funny technologies like FCOE…and for those of us that have been through this before…   I predict that FCOE will be lumped into the same category as Token Ring.

Yet another overly complex technology  that will not be very multi-vendor compatible that was more expensive than competing technologies and didn’t have a well defined path to higher performance.

#FCoE vs #ISCSI a #deathcage match

So much at stack... so many different opinions...  I love this tongue in cheek post...

http://datacenteroverlords.com/2011/09/14/fibre-channel-and-ethernet-the-odd-couple/

Crude...but pretty accurate.

FCOE is a solution looking for someone that likes overly complicated expensive architectures.

FCOE is this decades Token Ring.  

#NVGRE Network Virtualization without any new protocols?

#VxLAN will it suceed or #fail ?

Etherealmind doesnt think so...

http://etherealmind.com/top-5-things-vxlan-fail/

Please comment to let me know what you think....

Monday, September 12, 2011

#ComWare DHCP Option 43 with #HP E-MSM WLAN Solutions


ComWare DHCP Option 43 with HP E-MSM WLAN Solutions

Four discovery methods are available. The following table summarizes their
features and recommended applications.

If controlled APs are behind a firewall or NAT device.

Discovery is performed whenever an AP:
 Is restarted (or reset to factory defaults)
 Loses connectivity with its controller
 Is removed and rediscovered using an action on the Controlled APs >>
Overview > Discovered APs page.


Unprovisioned APs

Once an unprovisioned AP has received its IP address from a DHCP server, it
attempts to discover a controller using the following methods, in order:
 UDP broadcast
 DHCP
 DNS

When configured as DHCP client (which is the factory default setting for all APs),
an AP can obtain the IP addresses of controllers on the network from any DHCP
server configured to support the Colubris Vendor Class (DHCP option 43).
Note: If you have two (2) or more E-MSM Controllers, the AP will only connect to
one (1) controller at a time based on the priority listing. In the event that the EMSM
AP loses its connectivity to the active controller, discovery process is
performed again.
Green=1000Mbps, Yellow =10/100Mbps
SYS
RPS
SLOT1
SFP+
49 50 51 52
Green=10Gbps, Yellow=1Gbps
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48
10/100/1000Base-T
Duplex: Green=Full Duplex, Yellow=Half Duplex PoE: Green=Delivering Power, Yellow=Fault, Flashing Gren=Over Budget
H3C S5800 Series
Mode
Green=Speed
Yellow=Duplex
Unit
Flashing=PoE


1. These are the assumptions that we will be using in our
config example using two (2) E-MSM Controllers.
E-MSM Controller #1 192.168.1.12 (hex C0A8 010C)
E-MSM Controller #2 192.168.1.13 (hex C0A8 010D)
Subnet – 192.168.5.0/255.255.255.0
Domain Name – cinci.rr.com
DNS servers – 209.18.47.62
Gateway – 192.168.5.1
#
dhcp enable
#
dhcp server ip-pool 192.168.5.0-net
network 192.168.5.0 mask 255.255.255.0
gateway-list 192.168.5.1
dns-list 209.18.47.62
domain-name cinci.rr.com
option 43 hex 0108 C0A8010C C0A8010D

Notes:
01 - Colubris option code 1 as defined in the DHCP server
08 - Option code 1 is 8 bytes long which are the IP addresses of the controller in
hex C0A8010C & C0A80CD


2. These are the assumptions that we will be using in our
config example using two (3) E-MSM Controllers.
E-MSM Controller #1 192.168.1.12 (hex C0A8 010C)
E-MSM Controller #2 192.168.1.13 (hex C0A8 010D)
E-MSM Controller #2 192.168.1.14 (hex C0A8 010E)
Subnet – 192.168.5.0/255.255.255.0
Domain Name – cinci.rr.com
DNS servers – 209.18.47.62
Gateway – 192.168.5.1
#
dhcp enable
#
dhcp server ip-pool 192.168.5.0-net
network 192.168.5.0 mask 255.255.255.0
gateway-list 192.168.5.1
dns-list 209.18.47.62
domain-name cinci.rr.com
option 43 hex 010C C0A8010C C0A8010D C0A8010E
Notes:
01 - Colubris option code 1 as defined in the DHCP server
0C - Option code 1 is 12 bytes long which are the IP addresses of the controller
in hex C0A8010C, C0A8010D & C0A8010E

The MSM controller can handle up to five (5) IP addresses for IP
provisioning.

If there is a DNS Server lists defined on the DHCP IP pool, the AP(S) will try to
connect first via DNS before sending unicast packets to each individual controller
sent via DHCP. The AP appends the default domain name returned by a DHCP
server (when it assigns an IP address to the AP) to the controller name.
 cnsrv1.<domain-name>
 cnsrv2.<domain-name>
 cnsrv3.<domain-name>
 cnsrv4.<domain-name>
 cnsrv5.<domain-name>

#PDS to build a cloud data center here in #Milwaukee

One of HP's largest resellers in Wisconsin is about to roll out a large cloud  based data center...

http://www.jsonline.com/business/129626778.html