Jump to content

Atlanta, Florida, Virginia dedicated


Todd Holley

Recommended Posts

I wasn't aware f10 had sflow on their linecards. I don't keep up with vendors we don't use (at all). I apologize.

 

sFlow is actually patent-pending proprietary Force10 technology - they're the only ones that have it.

 

Fixedorbit reports, initially, 76 peers total. I know that doesn't include private peering, because of proxy aggregation etc, but it's a far cry from their marketing 400 peers :)

 

http://fixedorbit.com/AS/25/AS25973.htm

 

FixedOrbit is a fun tool for people that sit on webhostingtalk to extract little tidbits- but it's not remotely accurate. Private peers aside, it doesn't even track announced IP addresses correctly. For examples, here's the Nobis ASN:

 

http://fixedorbit.com/AS/15/AS15003.htm

 

Control of approx 1,786 IP addresses (0.00%) in 6 groups [more info...]

 

In reality, we have closer to 12,000 announced and probably 8,000 more than that provisioned. Here is our /20 (8,000 IP's) direct from the ARIN website:

 

http://ws.arin.net/whois/?queryinput=!%20NET-216-6-224-0-1

 

Plus all of our Mzima netblocks before ARIN approved us for our own /20:

 

http://ws.arin.net/whois/?queryinput=!%20NET-67-201-48-0-1

http://ws.arin.net/whois/?queryinput=!%20NET-72-37-145-0-1

http://ws.arin.net/whois/?queryinput=!%20NET-72-37-224-0-1

http://ws.arin.net/whois/?queryinput=!%20NET-72-37-237-0-1

http://ws.arin.net/whois/?queryinput=!%20NET-72-37-218-0-1

http://ws.arin.net/whois/?queryinput=!%20NET-72-37-222-0-1

http://ws.arin.net/whois/?queryinput=!%20NET-72-37-221-0-1

http://ws.arin.net/whois/?queryinput=!%20NET-72-37-242-0-1

http://ws.arin.net/whois/?queryinput=!%20NET-67-201-0-0-2

http://ws.arin.net/whois/?queryinput=!%20NET-67-201-40-0-1

http://ws.arin.net/whois/?queryinput=!%20NET-72-37-246-0-1

 

Thankfully ARIN's 75% utilization policy is tracked far more accurately than a couple little scripts on fixedorbit.com or cidr-report.org.

 

I don't see mzima on the equinix peering exchange as well. I also only see a couple of Tier1s, and I don't see any end user networks. End user networks are great to peer with, because you offset the cost of paying for transit costs. However, it's difficul to peer with anyone like that, unless you're on equinix peering exchange, or switch-and-datas peering fabric.

 

Regarding Equinix Peering Exchange:

 

http://www.mzima.net/pr20040922.html

Link to comment
Share on other sites

 

Fixedorbit reports, initially, 76 peers total. I know that doesn't include private peering, because of proxy aggregation etc, but it's a far cry from their marketing 400 peers :)

 

http://fixedorbit.com/AS/25/AS25973.htm

 

Are we assuming fixedorbit is 100% up-to-date? I know of far too many providers throughout the hosting industry that peer with a provider, or have tier1 transit and produce erroneous results on fixedorbit. Since Softlayer has come up in this thread, let's use them..

 

http://fixedorbit.com/AS/36/AS36351.htm

 

This is inaccurate, as their current provider list, and portal listing shows:

 

Dallas DC: INTERNAP, SAVVIS, GLOBAL CROSSING, VERIO, ABOVENET, LEVEL 3

 

I don't see mzima on the equinix peering exchange as well. I also only see a couple of Tier1s, and I don't see any end user networks. End user networks are great to peer with, because you offset the cost of paying for transit costs. However, it's difficul to peer with anyone like that, unless you're on equinix peering exchange, or switch-and-datas peering fabric.

 

From my understanding, MZIMA strives to produce a network that is independent from exchanges, by sending data over it's own terabit providers via their own backbone, and not that of an independent peering fabric. Different strategy, but it seems to still be effective.

 

EDIT: Corey beat me to it.

Link to comment
Share on other sites

It should be noted that I am not dissing mzima, I'm just trying to point out the difference between marketing fluff and actual numbers :p

 

Mzima being a US company, if they really used "marketing fluff" on these #'s they'd of been in a heap of trouble with the FTC long ago. Fortunately, personal websites that you referenced are inaccurate at best, as is proven when checked against arin.net and our own records (and i would assume, colo cross records as well).

Link to comment
Share on other sites

I don't see them being announced on ord on the peering exchange. I don't see their routes from any equinix location.

 

NANOG uses CIDR report for a number of good reasons, such as aggregation policy. It's hardly a personal website, since phillip smith is the owner of cidr reports. It's useful, probably a little more than you would think.

 

Your blocks are being announced from within mzimas AS. That's why most aren't showing up. Your doing proxy aggregation, since you are single homed to them. Most of your prefixes aren't originating from your own AS. You can get around this by announcing them with atomic aggregation, so you don't pollute the routing tables with more specifics. We are doing proxy aggregation in one location, due to some issues with internap refusing to accept our prefixes under another AS.

 

Here's from our chicago location, full tables:

r509-aggr509#sho ip bgp | include 15003
*  66.207.167.0/24  12.123.33.249                          0 7018 3356 25973 15003 i
*  66.207.174.0/24  12.123.33.249                          0 7018 3356 25973 15003 i
*  67.201.48.0/23   12.123.33.249                          0 7018 3356 25973 15003 i

 

As you can see, I only see 1024 ips being announced from your AS.

 

sFlow is used on foundry gear as well, not just force10. I've never used force10 stuff before, I would use foundry before the former, and I would use juniper gear after cisco.

 

Anyways, Their network is not as peer-heavy as you would think.

Link to comment
Share on other sites

Are we assuming fixedorbit is 100% up-to-date? I know of far too many providers throughout the hosting industry that peer with a provider, or have tier1 transit and produce erroneous results on fixedorbit. Since Softlayer has come up in this thread, let's use them..

 

http://fixedorbit.com/AS/36/AS36351.htm

 

This is inaccurate, as their current provider list, and portal listing shows:

 

Dallas DC: INTERNAP, SAVVIS, GLOBAL CROSSING, VERIO, ABOVENET, LEVEL 3

 

 

 

From my understanding, MZIMA strives to produce a network that is independent from exchanges, by sending data over it's own terabit providers via their own backbone, and not that of an independent peering fabric. Different strategy, but it seems to still be effective.

 

EDIT: Corey beat me to it.

 

If that were true, then why are they posting they are on the equinix peering exchange? They don't own their own backbone. They lease it from another carrier. They are NOT a tier1. They rely completely on buying transit to get to the world. Same with internap. Internap is a tier2, mzima is tier2, etc.

Link to comment
Share on other sites

If that were true, then why are they posting they are on the equinix peering exchange? They don't own their own backbone. They lease it from another carrier. They are NOT a tier1. They rely completely on buying transit to get to the world. Same with internap. Internap is a tier2, mzima is tier2, etc.

 

I can't speak for MZIMA's press releases, but I can infer that that PR was from September 22, 2004, meaning their method of peering via the exchange could be indirect, or their traffic could be accessing the IBX exchange differently now, then it was then some 3+ years ago. The PR is so out of date that it implies gig-e was big news, but it was then, not now. I would presume that they're likely on different connections, etc.

 

As for MZIMA and internap, I never stated they were Tier1, but rather that they have TIER 1 transit...Yes you are right that they are, as an upstream combined, Tier 2.

Link to comment
Share on other sites

Also Gary, I did some digging and this is worth looking at. They're definitely live on the exchange. As for their peering relationship, and exact numbers, I can't speak for.

 

route-views.oregon-ix.net>sho ip bgp 206.223.119.62

BGP routing table entry for 206.223.119.0/25, version 3553693

Paths: (4 available, best #4, table Default-IP-Routing-Table)

Not advertised to any peer

2914

129.250.0.11 from 129.250.0.11 (129.250.0.51)

Origin IGP, metric 46, localpref 100, valid, external

Community: 2914:420 2914:2000 2914:3000

2914

129.250.0.171 from 129.250.0.171 (129.250.0.12)

Origin IGP, metric 56, localpref 100, valid, external

Community: 2914:420 2914:2000 2914:3000

7500 2518 2914

202.249.2.86 from 202.249.2.86 (203.178.133.115)

Origin IGP, localpref 100, valid, external

293

134.55.200.1 from 134.55.200.1 (134.55.200.1)

Origin IGP, localpref 100, valid, external, best

Community: 293:16 293:42

 

[root@eagle~]# whois 206.223.119.62

[Querying whois.arin.net]

[whois.arin.net]

 

OrgName: Equinix, Inc.

OrgID: EQUINIX

Address: 301 Velocity Way

Address: 5th Floor

City: Foster City

StateProv: CA

PostalCode: 94404-4803

Country: US

 

NetRange: 206.223.119.0 - 206.223.119.255

CIDR: 206.223.119.0/24

NetName: EQUINIX-IX-CHI

NetHandle: NET-206-223-119-0-1

Parent: NET-206-0-0-0-0

NetType: Direct Assignment

NameServer: NS.EQUINIX.COM

NameServer: NS2.EQUINIX.COM

Comment:

RegDate: 2002-03-04

Updated: 2004-04-16

 

OrgAbuseHandle: EQ127-ARIN

OrgAbuseName: Equinix Response Center

OrgAbusePhone: +1-888-892-0607

OrgAbuseEmail: support@equinix.com

 

OrgNOCHandle: EQ12-ARIN

OrgNOCName: Equinix NOC

OrgNOCPhone: +1-650-513-7000

OrgNOCEmail: noc@equinix.com

 

OrgTechHandle: GDE27-ARIN

OrgTechName: Dendy, Greg

OrgTechPhone: +1-650-513-7000

OrgTechEmail: gdendy@equinix.com

 

OrgTechHandle: LL249-ARIN

OrgTechName: Lee, Louis

OrgTechPhone: +1-408-360-5253

OrgTechEmail: louie@equinix.com

 

OrgTechHandle: SM125-ARIN

OrgTechName: McQuarry, Steve

OrgTechPhone: +1-650-513-7254

OrgTechEmail: steve@equinix.com

 

# ARIN WHOIS database, last updated 2008-01-06 19:10

# Enter ? for additional hints on searching ARIN's WHOIS database.

 

And it appears that this IBX-CHI connection appears to be MZIMA:

 

[root@eagle ~]# nslookup 206.223.119.62

Server: 72.37.224.5

Address: 72.37.224.5#53

 

Non-authoritative answer:

62.119.223.206.in-addr.arpa name = ge5-0.cr01.ord01.mzima.net.

 

Authoritative answers can be found from:

 

[root@eagle ~]#

Link to comment
Share on other sites

Well, nearly... :)

 

Here's from our chicago location, full tables:

 

As you can see, I only see 1024 ips being announced from your AS.

 

I'm not a networking guy, but I would say it looks like you have big problems with your Chicago switch. That's an extremely tiny excerpt of what we're announcing in Dallas alone; I've confirmed that nearly all of our IP's (Mzima and ARIN) are currently announced via our AS15003. Jake was kind enough to pull this off of one of our Dallas switches for me..

 

Sw1ra1.Dal1.ubiquityservers.com>show bgp
BGP table version is 28, local router ID is ***.*.***.*
  Network          Next Hop            Metric LocPrf Weight Path
*> 66.207.167.0/24  0.0.0.0                  0         32768 i
*> 66.207.174.0/24  0.0.0.0                  0         32768 i
*> 67.201.48.0/23   0.0.0.0                  0         32768 i
*> 216.6.226.0/24   0.0.0.0                  0         32768 i
*> 216.6.231.0/24   0.0.0.0                  0         32768 i
.....

 

As you can see that's 1,500 just in one POP, which goes on quite a lot from there. If these tools were correct they really should be showing roughly 20,000 IP addresses from AS15003 as ARIN.net displays in allocations.

 

If that were true, then why are they posting they are on the equinix peering exchange? They don't own their own backbone. They lease it from another carrier. They are NOT a tier1. They rely completely on buying transit to get to the world. Same with internap. Internap is a tier2, mzima is tier2, etc.

 

Again I've never seen any evidence either way, and I really don't see where you could get reliable information to make such a claim (much like telling the community that Mzima is lying about their private peering) but maintaining private fiber, leased or owned, is not the same as just using other providers for upstream. And there are actually only 9 recognized tier 1's...

 

http://en.wikipedia.org/wiki/Tier_1_carrier

 

 

 

Unless colo cross has more mud to sling at our network, I'm done.

Link to comment
Share on other sites

Well, nearly... :)

 

 

 

I'm not a networking guy, but I would say it looks like you have big problems with your Chicago switch. That's an extremely tiny excerpt of what we're announcing in Dallas alone; I've confirmed that nearly all of our IP's (Mzima and ARIN) are currently announced via our AS15003. Jake was kind enough to pull this off of one of our Dallas switches for me..

 

Sw1ra1.Dal1.ubiquityservers.com>show bgp
BGP table version is 28, local router ID is ***.*.***.*
  Network          Next Hop            Metric LocPrf Weight Path
*> 66.207.167.0/24  0.0.0.0                  0         32768 i
*> 66.207.174.0/24  0.0.0.0                  0         32768 i
*> 67.201.48.0/23   0.0.0.0                  0         32768 i
*> 216.6.226.0/24   0.0.0.0                  0         32768 i
*> 216.6.231.0/24   0.0.0.0                  0         32768 i
.....

 

As you can see that's 1,500 just in one POP, which goes on quite a lot from there. If these tools were correct they really should be showing roughly 20,000 IP addresses from AS15003 as ARIN.net displays in allocations.

 

 

 

Again I've never seen any evidence either way, and I really don't see where you could get reliable information to make such a claim (much like telling the community that Mzima is lying about their private peering) but maintaining private fiber, leased or owned, is not the same as just using other providers for upstream. And there are actually only 9 recognized tier 1's...

 

http://en.wikipedia.org/wiki/Tier_1_carrier

 

 

 

Unless colo cross has more mud to sling at our network, I'm done.

 

Nothing is wrong with our gear. We're taking full routes from ATT and partials from another. I think you're just accepting default routes from mzima, judging by what you pasted. We're filtering /24's out to reduce the amount of routes installed. That is probably why we don't see all those /24's youre announcing. Since you're not a network person, half of the internet's routing table are /24's. Filtering those removes the pressure on our CEF tables, since countries like china like to announce piles /24's without aggregation.

 

I wasn't slinging mud at your network. I could really care less what you say, or do. I was just trying to tell you that mzimas network isn't as expansive as say, internap, in terms of usable peers. Regardless, most of mzimas peers are useless when it comes to gaming. I mean, if you're with a carrier who has nothing but colocation/hosting companies as peers, and you're hosting games, those peers don't do jack for you.

 

Internap's network is signifigantly better for gaming than mzima.

 

-M

Link to comment
Share on other sites

Well I thank you for yours and jon's un-motivated public service announcement that Mzima is worthless for gaming. The traces I get back on a weekly basis from new sales contacts that do quite frequently show lower pings to Mzima than Internap are going to have to start coming with a disclaimer warning gamers of this.

Link to comment
Share on other sites

Archived

This topic is now archived and is closed to further replies.

  • Who's Online   0 Members, 0 Anonymous, 26 Guests (See full list)

    • There are no registered users currently online
×
×
  • Create New...

Important Information

We have placed cookies on your device to help make this website better. You can adjust your cookie settings, otherwise we'll assume you're okay to continue. Terms of Use