Slashdot Mirror

I'd have to go with Street Fighter II and its sequels, up to Super Street Fighter II: Turbo. That series just got better, between those two releases, if you ask me, with the sole exception being the SF2: Hyper Fighting, which was Turbo done wrong.

It seems like I made half my friends playing this game. One of my best friends and I met playing that game, and more than 10 years laters, we're still very close. I met a roommate in college (we went on to pay for school running a business selling magic cards).

It's funny how games bring people together. I was an avid player of the Carrion Fields MUD, also, and while it didn't necessarily change ME all that much, it did change several people I came in contact with, because they met me there. Two people I met playing that MUD who later went on to start Avendar with me I later recruited to work with me at Exodus, long before it was devoured by C&W. (One of them subsequently recruited a fourth mudder, too) Since neither was part of the industry to begin with (one was doing post-graduate math, the other a substitute teacher in Alaska), and yet both remain gainfully employed in IT to this day (it's been around 4-6 years now since I got them to sign up).

Go see Exodus

They are a data center company that provides co-location to companies with numerous locations across the US. I used to be at the one in Irvine, CA but moved to SBC (Pacific Bell). SBC makes Exodus look like a joke.

Exodus is an upstream ISP. They own uber pipes on the coasts.

Blame Cisco

Times have changed,
Our Slashdot's getting worse,
There's no more "stuff that matters,"
Just a hit on VA's purse!

Should we blame the government?
Or blame our ISP?
Or should we blame the h4x0rs at DirecTV?

No!
Blame Cisco! Blame Cisco!
With their blinking LEDs,
And inflated techsupport fees,
Blame Cisco! Blame Cisco!
We need to form a full assault!
It's Cisco's fault!

Don't blame me for old JonKatz,
He lost his damn connection,
Now he's shooting at little brats!
And poor Roblimo once had
pictures of Heidi Wall,
But now, when I see him,
He tells me to suck his balls!

Well,
Blame Cisco! Blame Cisco!
It seems like everything's gone down
Since Cisco came to town.
Blame Cisco! Blame Cisco!
They're not even a real company, anyway.

Slashdot could've been the place to get our fix of daily news,
Instead we just get jpegs of the results of anal screws!

Should we blame the Editors?
Should we blame the Trolls?
Or the moderators who let them take their toll?

Heck, no!
Blame Cisco! Blame Cisco!
With all their worthless stock options,
And that bitch Anne Tomlinson,
Blame Cisco! Shame on Cisco, for...

The crap that we flood,
The news that's a dud,
The MPAA,
Your Rights gone away:
We must blame Cisco! Shout and cuss -
Before somebody thinks of blaming us!

They rent rackspace from Exodus (who according to messages (index of week's messages) on INCIDENTS). Exodus is doing nothing it seems and condones their activities. They don't seem to be doing anything more than getting some REALLY paranoid sysadmins underwear in a knit, but I really don't like being batch scanned for no real reason. So here's my info I've scoped on them.
whois -h whois.networksolutions.com quova.net ...

Registrant:
David Naffziger (QUOVA2-DOM)
333 W Evelyn
Mountain View, CA 94043
US

Domain Name: QUOVA.NET

Administrative Contact, Technical Contact, Zone Contact:
hostmaster (HO8675-ORG) hostmaster@QUOVA.COM
Quova, Inc.
333 W. Evelyn Ave.
Mountain View , CA 94043
US
(650) 962-2933
Fax- (650) 962-2025
Billing Contact:
billing (BI4691-ORG) billing@QUOVA.COM
Quova, Inc.
333 W. Evelyn Ave.
Mountain View , CA 94043
US
(650) 962-2933
Fax- (650) 962-2025

Record last updated on 23-May-2000.
Record expires on 16-Nov-2001.
Record created on 16-Nov-1999.
Database last updated on 6-Jul-2000 18:55:18 EDT.
Domain servers in listed order:

NS1.QUOVA.COM 208.37.145.35
AUTH50.NS.UU.NET 198.6.1.161


www.quova.net is running Apache/1.3.12 (Unix) PHP/4.0.0 FrontPage/4.0.4.3 on Solaris netcraft
AND SINCE THEY shouldn't mind!!!

cherrycoke:~$ sudo nmap -sX -vv -O www.quova.net
Starting nmap V. 2.54BETA1 by fyodor@insecure.org ( www.insecure.org/nmap/ )
Host (205.177.226.233) appears to be up ... good.
Initiating FIN,NULL, UDP, or Xmas stealth scan against (205.177.226.233)
The UDP or stealth FIN/NULL/XMAS scan took 69 seconds to scan 1525 ports.
For OSScan assuming that port 23 is open and port 1 is closed and neither are firewalled
Interesting ports on (205.177.226.233):
(The 1520 ports scanned but not shown below are in state: closed)
Port State Service
23/tcp open telnet
80/tcp open http
111/tcp open sunrpc
514/tcp open shell
2049/tcp open nfs

TCP Sequence Prediction: Class=random positive increments
Difficulty=132682 (Good luck!)

Sequence numbers: 6A1BA7D9 6A255F59 6A2A5515 6A2F4624 6A37B2F6 6A3CE0D6
Remote OS guesses: Solaris 2.6 - 2.7, Solaris 7
OS Fingerprint:
TSeq(Class=RI%gcd=1%SI=2064A)
T1(Resp=Y%DF=Y%W=2297%ACK=S++%Flags=AS%Ops=NNTNWME )
T2(Resp=N)
T3(Resp=N)
T4(Resp=Y%DF=Y%W=0%ACK=O%Flags=R%Ops=)
T5(Resp=Y%DF=Y%W=0%ACK=S++%Flags=AR%Ops=)
T6(Resp=Y%DF=Y%W=0%ACK=O%Flags=R%Ops=)
T7(Resp=Y%DF=Y%W=0%ACK=S%Flags=AR%Ops=)
PU(Resp=N)


Nmap run completed -- 1 IP address (1 host up) scanned in 83 seconds


Some "security company," with all those notoriously insecure services running on their webserver (NFS, telnet, shell, RPC). Oh well. It looks like their webserver is colocated with some company.
cherrycoke:~$ traceroute www.quova.net
traceroute to www.quova.net (205.177.226.233), 30 hops max, 40 byte packets
1 orangecrush (192.168.0.1) 2.638 ms 2.239 ms 2.238 ms
2 quincy-asx-2.ziplink.net (206.15.185.18) 509.732 ms 203.12 ms 219.374 ms
3 206.15.185.17 (206.15.185.17) 209.86 ms 215.767 ms 199.762 ms
4 * zl-qnz-cisco2bcn.ziplink.net (206.15.158.150) 205.427 ms 214.611 ms
5 zl-pru-h20-1z172h209.ziplink.net (206.15.172.209) 219.845 ms 214.564 ms 219.459 ms
6 206.15.185.217 (206.15.185.217) 219.572 ms 216.462 ms 199.567 ms
7 bay4-322.quincy.ziplink.net (208.196.109.82) 279.498 ms 274.794 ms 259.6 ms
8 zl-sf-e20-2sf7k.ziplink.net (206.15.172.6) 279.477 ms 265.691 ms 279.473 ms
9 pacbell-1.globalcenter.net (198.32.128.32) 279.597 ms 272.632 ms 279.56 ms
10 pos4-2-155M.cr1.SNV.gblx.net (206.132.150.25) 269.622 ms 272.892 ms 299.483 ms
11 pos2-0-622M.cr1.IAD3.gblx.net (206.132.113.102) 337.01 ms 333.853 ms 339.512 ms
12 pos0-0-0-155M.br2.IAD3.gblx.net (206.132.253.26) 339.529 ms 343.903 ms 349.513 ms
13 digiweb.s2-1-1.br2.IAD.gblx.net (204.152.166.190) 349.878 ms 273.863 ms 299.393 ms
14 209.143.145.194 (209.143.145.194) 309.769 ms 277.821 ms 299.558 ms
15 ucla.digiweb.com (206.161.225.11) 299.497 ms 292.234 ms *

I've worked a number of years in the colocation business. Choosing a good one depends on your definition of 'good'.

Many people are solely concerned with price, and what that really comes back to is the old saying you get what you pay for.

That being said, many people have pointed out some of the more obvious (atleast to us) questions to ask.

Bandwidth is usually the first concern. People mentioned peering and how big the pipes are, but a very important factor to find out is how utilized they are. Some company might say they have 20Gb of available bandwidth (they may say 40Gb if they double their number because its full duplex). Then they say they are using only 10Gb of that 20.

Wow great, that's plenty of space left over. Unfortunately usage of the available bandwidth is not distributed that evenly. There are usually a handfull of peers which take the brunt of the traffic. If those links are at 95% all the time, you're going to see problems. Colo's always have underutilized DS-3's and OC-3's to ISP's which they may only do a few megs of traffic to.

It is also important where the traffic goes. If the colo is peered with a Tier 1 backbone, you will want to make sure the peers are to different locations on that backbone. It's much less likely to get congested, and much less likely to affect you in event of a backbone failure. How much of their traffic goes to a public NAP? These places can often be a black hole from which packets never return.

And depending on who you are, find out where the majority of your customer base is. If you are a B2B business, and you're getting or sending feeds to yahoo news, then maybe you should be on GlobalCenter's network where those servers live? Or lets say all your clients are coming from AOL, maybe a colo facility with lots of underused AOL bandwidth is right for you. Obviously these conditions change, but it's good to look at them.

A number of people mentioned physical security. It is definately good to have, but it's by no means any type of guarentee. Having things in a nice locked cage or cabinet is often more percieved security and showcasing than anything else.

Colo facilities do not have Elite Special Ops guys posted at all doors. These are underpaid security guards. So your servers are as secure as any of the items in Radio Shack in the mall. Not to say physical security is a waste, it IS good to have things locked and protected. And it's not like there are people just wandering around the datacenters looking to swipe things. But many times security is boasted upon, but behind the scenes it's kinda hokey.

Equinix is a datacenter which goes all out on security and facility. So far they have done a good job at it, though their network is second on the priority list as opposed to their facilities. There have been cases when physical security has been an issue for a site, but there are usually far easier ways to harm a site than bashing the server through a cage with the end of a broomstick.

The infrastructure the Colo run's on is important if you are looking for certain things. Network wise a switched 10/100 port is pretty standard, but remember, while it is switched, the uplink bandwidth from that switch to the core routers is shared. You are competing for bandwidth at some points. So make sure you have options. Can you upgrade to a router port directly? Can you multi home to different switches? Can you get HSRP or VRRP for redundant uplinks? Can you use IBGP or BGP for load balancing uplinks? Does the colo have their own backbone? Or is it leased from someone else?

The rest of these many people don't think about until it's far too late to change. Can you bring in an outside circuit of your own? Can you manage that connection yourself? Can you get a phone line into the cage?

One thing I have noticed more and more is that Colo's have increased the number of core, and border routers to provide additional redundancy for the customer. Unfortunately this has also vastly increased the complexity of the layer2 and layer3 fabrics as a whole. So little problems often times have a ripple effect on the whole thing with results that may degrade or bring everything down.

More facility related ones are: How much power do I get? Is that included in the price? Can I get additional power later (or do I need to order it now?) Is there additional rack/cage space near me that I can expand to? Do I have first right of refusal on that space?

Is there adequate power for the whole facility? Is there adequate UPS? Is there adequate generators for when UPS fails? Are there extra generators incase one/all of the primary generators fail? How much AC is there? Can it support the whole facility (this is important because of the popularity of 1 and 2U servers which suck up more power and put out more heat per rack).

What are the plans for current local expansion of data center space? Where do you have other large datacenters? How are they connected? How much peering does each datacenter have (remember no datacenter is an island)?

The next major piece is managed services. What this means is what will they do for you that you could do yourself but don't want to. Some colo's offer this (Exodus, Globalcenter), and some do not (Abovenet, Equinix). Bandwidth and space are becoming easier to come by, and may not always be the limiting factor in a sites growth.

Most all of them do offer some basic services such as a NOC, remote hands capability, and some basic troubleshooting. There are some people in the various operation centers who are VERY sharp, the majority of them are people fresh out of HS, or doing it as a second job to get by.

Someone mentioned earlier the little fish, big pond case. Unfortunately this seems to ring true of the colo's that offer managed services. If you're the little fish, it's very likely all of the smart people who do managed services for a large colo are busy with the bigger fish (say ebay, hotmail, geocities, yahoo, etc) than with you. You're likely to get basic services, but nothing special or out of the ordinary.

You can look back in the trade mags about 10 months and see the slamming Exodus got for customer service because they had more business than they knew what to do with. Fortunately this has gotten better, but there is way more business in the market than colo's know what to do with.

AboveNet and Equinix just plain don't have an offering in those areas. They usually partner and recommend other people who are experts in that space. Companies like SiteSmith, Emanage, MimEcom or Lante offer outsourced managed services for colo customers. Designing a scalable fault tolerant site, implementing it, and running it so the customer doesn't have to. Unless these types of colo providers develop their own managed services, the options are to outsource, or do it yourself.

No site is an island, which is why managed services become more and more important. The tier 1 colo's are all experiencing a bit of commoditization of the space and bandwidth. As a few people have already noted, many of the colo's are all starting to seem the same. Some people have personal preferences one way or another, and there are also usually niche type reasons why one is better suited for a certain customer.

Good managed services provide a way to scale a site to more users, higher levels of fault tolerance, and all over reliability. It's often hard to find people with these skills in house (or its hard to make them divide their time from whatever their primary job focus is). This is why people look for managed services from their colo, or from another managed services provider.



-Alex

Seriously, the claims are going to be very similar from co-lo companies. Previous posts really covered the technical things you want to look for. But I think the real value is in first hand experience. For my $.02 I have always liked Pair Networks, but I don't know if they offer co-location with your own servers any more.

Ultimately, I think it would come down to:

1. Support. This includes the overall competence of the company.
2. Location, so you can physically administer your servers when needed.
3. Good routing and bandwidth capabilities.

Or maybe reverse the order of these. Like I said I have always gotten a good feeling from Pair.

Who/what exactly is Exodus?

Exodus is one of the world's biggest (in terms of service capacity available) Internet Service Providers.

"We're going to need bandwidth. Lots of bandwidth."

Exodus specializes in having more bandwidth then most of the third world. They've got NAPs (Network Access Points, i.e., backbone connections) all over the continental United States, and a few outside the US as well. They link this all together using both external and internal networks. The end result is, most anywhere on the net that has a good connection, has a good connection to Exodus.

They provide servers. Do you need to host downloads for ten million users? Exodus can give you servers to do so.

They provide co-location space. If their standard server packages just won't cut it -- bring your own. They'll give you a rack, a dedicated co-loc cage, or a dedicated high security vault.

Their web page has a lot of graphics because they have a lot of pictures of their equipment and graphs of their capacity. It is actually justified. You may want to make a return trip.