I am IIIT Network

2006-11-09T22:37:00.000+05:30

Who is to be blamed ??

Wireless LAN was down today. Girls Hostel Dlink switch was down. Now... who is responsible ?? - Is it the second grade switches given by our sponsors ? Is it the IIIT administration, who did not recruit somebody to look after the network after 6 PM or for not buying new switches as recommended by the sysadmin ? Is it the admins themselves or the Electricity board who cut off power until the UPS were drained ? Who is it ??

If someone were to be blamed, it is IIIT administration. They decided they do not need any student sysadmins for some reason which only GOD knows - how ignorant of them!!!. The management should realize the fact that, it is the student who knows the state of the network which he lives with - day in and day out, rather than the sysadmins who work 10 - 6. The profs have wired connectivity in their cabins and also at their homes. They are always given special treatment, special bandwidth and special care to their problems. So there is no chance for them to realize the problems you face in your hostels. You cant even expect them to have such a realization because they are not in the situation in which you students are.

Blaming the sysadmins and making fun of them in blogs is not the right thing or atleast the solution for solving IIITs network problems. They may be efficient or inefficient - thats a different issue, but there is a scope for you to workout things with them by helping them. Be patient and work collectively with the sysadmins to solve my problems. Good to see, some students already doing that. This is your network, you should take care of it yourselves.

2006-11-08T00:05:00.000+05:30

Few reasons why DC++ is blocked

A student wanted to share some info with all you people. So he wanted to post about DC++ on my blog. The following is what he wanted to convey:

Hello, I am ABC, a student of IIIT. I wanted to share something about DC++, it goes something like this.....

1. There is a 3Com switch in OBH which has two Gigabit ports. This switch is sort of a gateway between the entire OBH wireless network and the rest of IIIT. One port was burnt some time in the past (around 2 months ago). When sysadmins enquired about this problem at 3Com tech support, tech support people informed that the switches we are using are development switches and are not fully developed to Enterprise level standards and these switches may or may not handle heavy traffic. Indirectly they meant that the ports can be burnt or damaged at any time.

Q) Then why the hell these switches were deployed in our network ???

A) Chimple - because they came for free. Infact the whole of IIIT wireless network is free. It is sponsored by the MCIT (Ministry of Communications and Information Technology) and 3Com for Research and Development. 3Com sponsored because they needed a test bed to test their devices and they also wanted the wireless team to publish some white papers about the capabilities of their devices. MCIT wanted to test the issues in deploying and maintaining a campus wide wireless LAN in India. In this regard, the wireless team had to manage with whatever they had.

Coming back to where we left - yes, there were two gigabit ports on the switch, one got burnt and the switch was left with one. Now the wireless network is in a dangerous situation. Its network traffic is soaring through the switches which were not capable of handling that load. They had to take necessary steps to prevent the wireless network coming to a state of grinding halt if some of these switches are down. They measured the traffic at the wireless router (172.17.0.2), the traffic between the wired and wireless networks of IIIT - it was around 118 Mbps. As recomended by 3Com, the sysadmins decided to reduce traffic on the 3Com switches and made a test run. Surprisingly they found that, the entire traffic flowing through the wireless router was DC++ traffic. So they blocked DC++. After blocking DC++, the traffic between the wired and wireless networks of IIIT went down to 17 Mbps - an amazing 100 Mbps reduction in traffic. The sysadmins stuck with the idea.

2. Blocking DC++ saved the nightmares of wireless sysadmins. These nightmares were due to lots of complaints which they were receiving from students prior to blocking DC++. There were many complaints about not able to associate to an access point, not able to get an ip, able to get and ip but not able to ping, able to ping but network very very slow, putty also very very slow. After blocking DC++ the number of complaints reduced drastically. After blocking DC++, some students who were not able to connect to IIITs network for week or so suddenly got connected and were happy with the minimal connectivity they had. The admins did this as a preventive measure as there is only one port left on the switch because if that fails, they are left with no options. Until the new switches arrive, the admins thought bocking DC++ is a better alternative than bringing the entire OBH wireless network down. Good to see admins not working in damage control mode.

Q) why the hell was wireless network working reasonably well last year and the year before ??

A) It is because this year there was a drastic increase in the number of wireless users. If you are looking for numbers, the wireless users increased from 312 last year to 578 users this year - almost double. Wireless network is meant for basic connectivity and mobility (this was the basic idea behind the deployment of wireless LAN in IIIT) and is not for heavy network usage. So with heavy users like us, what happens ?? - the network chokes.

3. Students started complaining about the state of wireless network, booed the admins, labeled them inefficient, not capable of managing and handling things. Some even came to a conclusion that the entire wireless network design is wrong. When asked an explanation, the same students had no clue why they made that statement. This obviously caused a stir in sysadmins minds and Sysadmins said "ok, if you want to us to do someting, fine we will do it - no problem, we will block DC++" and they did.

Unfortunately, after blocking DC++ wireless network improved, atleast in terms of basic connectivity, mobility and minimal functinality, if not in speed. Blocking DC++ is also not good on sysadmin's part but they think it is fine to do so. why ??? Because it is our attitude towards them which is making them take certain steps. Now who is the looser, it is we students. It is not only the sysadmins who are ignorant, even some of us (the students) are also ignorant.

2006-11-05T23:37:00.000+05:30

 I have improved - says IIITs Internet

Hi all, what happned to me suddenly, the pages are being displayed even before being clicked - not exactly though. But what did they do to me, why am I running well these days...read on

Actually a big thanks to the sys admins who encouraged some students who wanted to solve my long pending problems. This should have come a long time ago, but I think it took this much time for the students to get really really frustrated and decide to solve the internet problems cooperatively with the sys admins. Well how did the problems get solved ??

0) When the ISP was changed to STPH from BEAM, I was brought to IIIT through wireless. For this connectivity I need a tower, the one which is on our main building towards basketball court who's antenna are facing towards hitech city. If you notice you will see one horizontal antenna and one vertical antenna capturing signals relayed by their main tower some where in the city. I also need a modem (a white colour flat thing just like a scanner in looks but much smaller in size, sitting at the bottom of the server rack). The tower and the modem are communicating at 2.4yy GHz frequency - Now guess what could be the problem - ofcourse my twin brother Wireless LAN. IIITs wireless LAN also functions at 2.4xx GHz. The difference between xx and yy is some zz MHz. Nobody knows what was is the value of zz MHz. One thing is well known that when to diff signals work at same frequency you expect something - yes ofcourse noise, accompanied by signal degradation and increase in bit errors.

When I was sucking badly and all you students were furious on me almost for a month. Mr. Muzafer our wireless sys admin - pointed this out (wireless LAN interference) this to STPH people and asked them to initiate necessary actions to correct their device. STPH was also fed up of IIIT sys admins calling them again and again quering the reasons for slow internet speeds. They decided to put a 7 GHz modem which would be free of any interference from my twin brother. By the time they put this modem one month had already passed. This modem is twice as fast, powerfull, and big in size than the previous modem. Then why are we having problems for almost 3 months now. good question read on....

Modem was there but there was no cable. Cable ?? what the hell - I said I was brought using wireless connectivity, where the hell did this cable come from. I come to IIIT on a wireless link until the tower. From the tower to the modem in server room I travel in a big black thick cable (dont know what it is made of). I reached server room, now I have to reach the Layer 3 switch but how ?? The modem had a special port which has some - I dont know, may be 30 or 40 big fat pins. Now I need to travel through this special port to the ethernet port of the L3 switch using a cable. It is this cable that was missing.

STPH did not have the cable, so they had to import the it from USA. It took a month for this cable to fly from USA to IIIT. Finally they fixed this cable. A test run was done on me and a yellow colour LCD on that new modem showed 0 (ZERO) bit errors. Now the link is fixed. The next thing to be fixed was the proxy server.

The proxy had a lot on misconfigured options and also it was not installed properly. Its installation was not tuned according to IIITs needs. Thanks to some students who volunteered to take things in hand and decided to solve the following problems:

1) Proxy server's poor hardware configuration (it is bad as described in the last blog)

2) Squid (The proxy server) is installed with default options. so it doesn not handle more than 1024 connections at an instant.

3) No limit on the number of user connections. Any user using a Download manager can initiate any number of connections resulting in Denial Of Service to others.

4) Proxy is handling internal traffic - traffic to 200, 208, 150. 150 ??????? yes 150. If you are downloading a file from 150 say a FedoraCore.iso (which is around 500 MB) through a browser with no "No Proxy for" setting for 150, all your traffic would flow through the proxy. The proxy is already loaded with heavy internet traffic from students, now it is unnecessarily handling such huge local traffic.

5) Limited bandwidth from ISP (a tight 4 Mbps cap).

6) Tunning some connection parameters and cache settings - this is not yet looked in much detail - so no comments.

How were the problems solved ???
-----------------------------------------------

0 - solved as explained above

2 - Squid is now installed from source rather than using a rpm (rpm is not recommended). It is compiled to handle 4096 connections in parallel and many other necessary options which would not have been enabled if installed from an rpm.

3 - Users are limited to some x number of connections at any instant.

4 - It is proposed and necessary actions are being initiated.

7 - I call solution 7 as "Trick - Use Brain" - "Trick - Use Brain" is temporarily used to seperate wired and wireless traffic to internet. Actually this is the main reason why your internet is rocking. It may not last long, be ready to accept slow internet speeds after a few days. This will be explained in detail when time permits.

About problem 1 and 5, dont ask me, ask IIIT management. 6 is postponed to a later date (after exams).

2006-11-02T02:11:00.000+05:30

I am IIITs internet.... why do I SUCK ???

Well all you guys might be wondering why am I sucking for the past 4 months. well here goes the story. This is based solely on my personal analysis, I could be wrong, I could be partially right or I could be 100% right.

Long long ago once upon a time - since my birth, I rocked IIIT with my services. All the students were of me. They used to boast about my speed in front of their Non-IIIT friends. I was looked after very well until I was 4 years old, I was 1 Mbps on my first birthday and by four I was 2 Mbps. As I was getting older and as the time passed I was upgraded gracefully.

Then from my 5th birthday I was being neglected, I was not upgraded enough, The number of internet users was doubling and I was being increased only linearly. 2004, Enter wireless LAN - drastic increase in internet users - almost by 50%. OOOK fine, increased bandwidth which was just "OK". Then came 2005, number of BTechs increased, number of PGs increased, MSIT adds more students (almost doubles infact), Agriculture lab was founded, Digital library of India (DLI lab) and Regional Mega scanning center (RMSC lab) followed. Because of increase in student strength there were more and more wireless internet users. People started to browse more and more from their rooms where they spent lot of time on internet. If they were in labs, they would not have spent the same time on the internet as they did in there rooms. So internet usage per person doubled (if not doubled it increased atleast by 50% - 75%. Internet culture was also changing, blogs, orkut, yahoo 360, skype, gtalk, google earth, yum installs, kodak gallery, yahoo photos, new downloads and bla bla bla (nothing wrong with the students though - they are just doing what they enjoy). Because of all this load students started experiencing slow internet.

Hostel users increased, labs increased, usage per user increased, resulting in an overall exponential increase and I was taken care only linearly. Add to above, our ISP problems, their link faliures, fiber cut downs and bad service. So change of ISP (from Beam to STPH), now IIIT is served internet wirelessly from sattillite. The advantage of beam ISP was, he was a little flexible. He raised our internet cap when we crossed our alloted bandwidth - he had no problems in doing so. With STPH - no such flexibility - very strict limitations, 4Mbps means 4Mbps - not a byte more nor a bit less, ofcourse STPH service and link is good.

Add to all above, our proxy server problems. At any given time it has only 1024 descriptors. Number of HTTP connection requests from wireless network to proxy is around 800 - 1300. It is buffering many requests and if the buffer is full, it is dropping remaining requests, almost 380 - 480 dropped connections per minute. Think what would happen to proxy if we were to add wired internet requests as that is the real scenario. This is a software problem, this can be solved by compiling squid properly with increased file descriptor capability. But why is it not being done. Because its got one more problem in connection to it. It goes as follows...

To service 1024 incomming requests in parallel, the squid process is using 89% - 96% of CPU. Imagine what would happen if we were to configure it to handle higher number of incomming connections (say 2048 or higher). I dont know what would happen. What would be CPU usage then. Would there be any place to run iptraf (tool to check incoming/outgoing traffic), how would screening of requests and logs be done, would ntop (a very good network moinitoring tool) and mrtg (system information collector and monitoring tool) run in such a tight environment which themselves are CPU intensive applications. All these CPU usage problems because our proxy has now grown old. Its configuration is as follows :

Dual processor IBM server
each is a P III 1 GHz processor (I thought P III was history)
1 GB RAM
80 GH Hardrive

Now, One more misconfiguration or intentional configuration of proxy server, I dont know. All the web/http traffic originating from our systems to our local servers (192.168 series servers), for example 200, 150, 208, research server are all being honoured by our proxy. I feel, there is no requirement for it to honour locally generated requests to our local servers. It can just drop/ignore them. If anybody knows a reason for this kindly let everybody know. If it drops these packets we have some scope for some free CPU cycles which can be used by the additional file descriptors, provided squid is configured with increased number of file descriptors. First of all you may get a doubt that why our local web/http traffic is going through proxy, well this story is all ours and it is as follows....

It is all because of you students. In your browser settings you do not add your local servers ips (192.168.36.*) in "No Proxy for" option. when you dont add these ips all your requests will go to the proxy and proxy is unfortunately configured to handle these requests. Strangely the same does not happen with web/traffic destined your local 172.16. or 172.17. subnets. If you do not specify servers running with ips in these subnets in "No proxy for" setting your packets are denied access at proxy. when they can do it for these subnets why cant they do it for 192.168.36. subnets. If anybody things there is a reason for treating web/traffic to 192.168.36. subnets in a different manner kindly let us know. If proxy drops these packets, we will be forced to add all the servers ips in the "No Proxy for" option (which is a one time job) and all our traffic will directly go to these servers decreasing the load on proxy. You just need to add "172.16.0.0/16, 172.17.0.0/16, 192.168.36.0/24" for firefox/mozilla and "172.16.*.*, 172.17.*.*, 192.168.36.*" for IE in the "No proxy for" text box. Now as packets go directly to our servers, the proxy does not even have to do that limited processing required to drop the packets.

Add to above, all the following: my mouse ball lost, my keyboard pin broken, projector in room 305 not working, Prof X's laptop not getting ip, hardisk missing from one system, switch port burned, D-link gigabit switch woking at 10 Mbps speed, switches got reset, somebody switched off access points, ethernet cable not working, linux wireless drivers not working, windows not working, virus broadcasting, new lab establishing so need new ips, network not documented.............la la la - la la la - la la la.

As far as I can see, you have a NightMare to face in the future. I warn all students not to depend on me in the near future. You have your exams coming your way, If you think that you can depend on me - TRY ME. One note especially for the CAT aspirants - If you want to risk your CAT exam then depend on me.

I am IIIT network

2006-10-27T01:45:00.000+05:30

Hello, Disciples of IIIT Hyderabad. I am IIIT network, no, not one more social network where u try to impress your opposite sex, I am a little different - I am a computer network. Well I am a combination of a variety of devices called computers, routers, switches, cables, fibers, power supplies, avis, mpgs, mp3s, packets, sniffers, tools and ofcourse my dear DC++. To put it shortly - I Suck and I Suck Big-Time. I desperately wanted to find out - why i suck ??

sooo lately, after scanning some jeevan-vidhya packets flowing through my network i decided to "explore myself". I wanted to see whether i am in harmony or not and also wanted to analyze what is naturally acceptable to me. Surprisingly or not surprisingly I found out that, I am much more complicated than i thought. It seems these days I am not naturally acceptable by most of the people using me. There can be many reasons...

My reasons being: I dont know whats going inside me and what happens to me every day, every hour, every second, I am up sometimes, down sometimes, if my wired ROCKS and my wireless SUCKS, if both ROCK then the ISP SUCKS, if ISP is fine then 200 hangs, when 200 works then proxy crashes, if proxy works then mirage needs an update, when all these are fine I am flooded with crap viruses and bacteria and finally if everything is fine, then its an holiday and everybody is at home watching TV. I am lost - i dont know whats happening??

I dont know how many cables are stuck in my Ass, first of all I dont know which cables go inside my Ass and which ones come out, I dont know which is a network cable and which is a power cable, I dont know which cable is going where, which switch is forwarding what, which router is dropping what, which user is blocked for what and why am I still not documented even after 8 years of my existence . I dont know how many dumb-heads keep pumping AIDS and Cancer virusues into my blood stream, I dont know whether the sys admins can handle me properly, I dont know whether IIIT administration understands me and the current state of my being and will consider to spend some pennies on me to keep me alive, I dont know this, I dont know that and the dont-know list goes on and on and on....... Finally I came to a conclusion that I dont know anything except the fact that I am running on some life support system which can go down anytime (phusssss......).

In this dissapointment and in the current existence of my being, I have decided to start a blog of my own through which I hope I could sufficiently convey my problems to you (my users) and hope to collectively solve these issues in the future.