h1 Project Description

The {link:Engineering Science|https://www.esc.auckland.ac.nz} Local Area Network (LAN) is spread over three floors and connects offices, computing labs and multiple servers as well as providing a link to the rest of the university (via ITS). Some sections of this LAN are also shared by other institutions in Uniservices House, 70 Symonds St (e.g., the Bioengineering Institute). We are interested in optimising the performance of this LAN.

The first step in improving network performance is often to simulate the network. Using the network simulation, network engineers can identify performance issues and experiment with different configurations, technologies, etc.

The aim of this Honours Year Engineering Science project is to accurately simulate the Engineering Science LAN using OPNET Modeler. OPNET Modeler is a state-of-the-art computer network simulation package. The student will be expected to familiarize themselves with OPNET Modeler, complete the model of the Engineering Science LAN and accurately simulate the traffic across that LAN.

h1 Project Diary

This is (an edited version of) the project diary of Varun Prasad, an Honours Year Engineering Science undergraduate student.

Last edited by Mike O'Sullivan and Cameron Walker on Jun 14, 2007.

h2 {anchor:|contents}{anchor}Contents
* {link:21 March 2007 - Opening Entry|#21march}
* {link:27 March 2007 - Posting on the OPNET Forums|#27march}
* {link:28 March 2007 - Meeting #2|#28march}
* {link:2 April 2007 - End of a Busy Week|#2april}
* {link:4 April 2007 - Meeting #3|#4april}
* {link:5 April 2007 - Postponement|#5april}
* {link:12 April 2007 - Meeting with Ilze and Selena|#12april}
* {link:16 April 2007 - Lots of reading..|#16april}
* {link:27 April 2007 - More looking around|27april}
* {link:30 April 2007 - Extending Bill's model|#30april}
* {link:2 May 2007 - Quick Update|#2may}
* {link:3 May 2007 - Research of Manufacturers|#3may}
* {link:7 May 2007 - Querying Suppliers, and SATA disks|#7may}
* {link:8 May 2007 - Line Cards or Hubs!|#8may}
* {link:11 May 2007 - Quick Update|#11may}
* {link:23 May 2007 - Applications|#23may}
* {link:2 June 2007 - Beginning of Exams|#2jun}
* {link:27 June 2007 - End of Exams|#27jun}
* {link:3 July 2007 - A Few More Delays|#03jul}
* {link:10 July 2007 - Presentation Slides|#10jul}
* {link:17 July 2007 - Presentation Day|#17jul}
* {link:24 July 2007 - Report Meeting|#24jul}
* {link:31 July 2007 - OPNET Returns|#10jul}
* {link:03 August 2007 - Applications and Disks|#03aug}
* {link:06 August 2007 - Profiles and Tasks|#06aug}
* {link:16 August 2007 - Background Loading|#16aug}

h2 {anchor:|21march}{anchor}21 March 2007 - Opening Entry

This is my first official entry for my (Varun Prasad's) Part 4 Project Diary.

I actually made a start last week however due to a few problems with permissions, I couldn't get underway on this wiki!
I will be following on the good work done by Xinsheng (Bill) Qu that he conducted over his summer studentship, and will be working with Cameron Walker and Michael O'Sullivan over the duration of my project. For more info on Bill's work, {link:here is a link to his project diary|Summer_0607_OPNET}.

Over the last week or so I have been familiarising myself with the project scope and getting the OPNET Modeler software version 12.0 installed on a handful of project room machines thanks to Sajy Augusty, our lab technician. I have a number of tutorials as well as Bill's work to have a good look at over the following week, as my priority for now is to get up to scratch with everything. Its all looking pretty greek to me at the moment!

Had a meeting with Mike and Cameron today, at the moment my job would be to investigate the desired switch models that are currently unavailable to us in the software package. My first target is the OPNET website {link:discussion forums|https://forums.opnet.com/cgi-bin/ubb/forumdisplay.cgi?action=topics&forum=University+Program+User+Forum&number=7&DaysPrune=30&LastLogin=}, hopefully they can shed some light in the near future.

Its going to be a busy week with a number of other uni deadlines on the way as well!

{link:Return to Contents|#contents}

h2 {anchor:|27march}{anchor}27 March 2007 - Posting on the OPNET Forums

Not an awful lot has happened so far, I have just been familiarising myself with the OPNET Topology as well as Bill's Wiki some more. We want to build the ~~ethernet switch~~ WS-C4506 but it is not available to us, and Bill alternatively has used the WS-C5000 switch. Likewise the ~~Ethernet Lane Switch~~ WS-C2950 is not available so Bill implemented the WS-C3200 instead. The best case scenario will be that the alternative switches are already available for me to import into the OPNET database, however the worst case situation is that I may have to build it from scratch !!

I have made a post to the discussion forums {link:here|https://forums.opnet.com/cgi-bin/ubb/forumdisplay.cgi?action=displayprivate&number=7&topic=000051} regarding this problem, fingers crossed I will get a response. For good measure I sent an e-mail to the support staff (support@opnet.com) tackling the same issues. Hopefully I can get an answer or a push in the right direction at the least.

I gave the OPNET FAQ's and Google a good workover for anything related to what we are after, but so far nothing. I will keep persevering...

{link:Return to Contents|#contents}

h2 {anchor:|28march}{anchor}28 March 2007 - Meeting #2

Today I had my weekly Wednesday meeting with Michael and Cam, again I am more at the stage where I am getting to know what is going on, so nothing particularly tangible has been produced at the moment. Cameron kindly lent me a text book for me to read over in terms of background reading for ethernet switches and the like (I will give a full reference for this book later). Likewise Michael has given me a few links on the net for me to look over, which I will do in due course.

Michael has been attempting to set up a meeting with Nevil Brownlee for us to discuss our issues with the missing Cisco switches, however he is out of the country at the moment. This meeting has been pushed until later in April, and until then the best I can do is keep on reading !!

{link:Return to Contents|#contents}

h2 {anchor:|2april}{anchor}2 April 2007 - End of a Busy Week

The end of a very hectic university week, in fact I haven't had much time to get cracking with the project much at all. I will catch up today and tomorrow no doubt. Just a quick word on the OPNET forums, no replies so far and it doesn't look all that likely I will get one at all. Bill apparently had a crack at putting up a post in those forums with no success either, so I think we have hit a dead end there.

I have had a look around Cisco's main website as well as other various places over the net, however details on the internal workings of the switches aren't particularly common it seems. Also thanks to a bit of kind wisdom from Bill, I am a bit more familiar with his network model that he developed over the summer, in terms of how the Engineering Science LAN works.

The semester break is just around the corner, which will give me a great opportunity to really make some headway!

{link:Return to Contents|#contents}

h2 {anchor:|4april}{anchor}4 April 2007 - Meeting #3

Michael has a few more acquaintances which he can get hold of, hopefully somebody can shed some light on our mystery switches! In the papers that Michael sent over for me to read it appeared other groups have similar models to us, however they seem to have all the right switches there for them already built into OPNET. Its amazing that our one isn't there... We had a chat about our situation, should we not find our switch it would probably be better to just go with what we have and model the traffic flow using that. At the very least we would have traffic flow data and should we stumble across the right switches at a later date, so we could easily make the modifications.

Nevil is expected to be back around the 20^^th of April, we will no doubt have a chat then. There has been a meeting set up tomorrow with Ilze Ziedins (a Senior Lecturer in Statistics), Percy Barboza (one of our Engineering Science IT technicians), Selena Hsiao (a Masters student of Ilze's who may be doing her project based on our data), Michael, Cameron and myself. This will be a good time to touch base and see where everyone is at. As for me, I will keep reading.

{link:Return to Contents|#contents}

h2 {anchor:|5april}{anchor}5 April 2007 - Postponement

Meeting with Ilze has been pushed until first week of the holidays, most likely the Thursday afternoon.

{link:Return to Contents|#contents}

h2 {anchor:|12april}{anchor}12 April 2007 - Meeting with Ilze and Selena

The moot discussion of our meeting here was to basically catch up on what we are all doing with our network simulation, and discuss what steps we are to take with regards to modelling real traffic data with OPNET. Percy has had a look around and there is hardware available to measure traffic flow across the network, we would typically want to pick a lab where there is a class running to do the modelling. The most meaningful unit of data for us would be a packet count in time increments, with the increments being as small as possible. We would also need some sort of a packet length measurement as this does vary with different packets, we think this is stored in the packet head, but gathering this piece of information could be tricky. Michael has a few contacts also to get hold of, which I'm sure will help clear things up some more.

The next step now is to have a solid background read on the theory behind packets, and more importantly how I can take our raw data and implement it into OPNET. I will be collaborating with Selena as we work through this, she has already kindly forwarded a few useful papers for me to read (papers that cover previous cases which look similar to ours). There is also the issue of the applications we want to be looking at when we do our traffic modelling, to be honest I could do with a bit more research on this before I have a chat with Percy and Bill about them.

And as for our famous switch issue, we all agreed that its definitely better to get some sort of a traffic model on our existing system even though we may not have the right switches. It is more important that we probably get cracking with modelling the traffic in OPNET, the switches can be dealt with later if a solution to our little problem is found.

{link:Return to Contents|#contents}

h2 {anchor:|16april}{anchor}16 April 2007 - Lots of reading..

Today I have been digging up some pdf's from the OPNET built-in help files about how to implement traffic flow into OPNET, and it looks like I have found a very handy guide as to how this is done. However this looks pretty out of it at the moment, and instead I've been having a close look at those papers that Selena forwarded to me as well as Bill's wiki. Something of particular note was that traffic data can be imported as a .txt or a .csv (comma separated variables) which could be very handy for us as these are very easy to manage and deal with.

Here is the location (although this isn't very useful if you don't have OPNET Modeler installed of course) \:

C:\\\Program Files\\\OPNET\\\12.0.A\\\doc\\\modeler\\\pdfs\\\01_29_Traf.pdf

I should upload this, I will when I figure out how.

Bill's wiki is clearing up on how these applications work as well, and the various types we will be dealing with. He did some great work in his student diary!

I came across this pdf in the OPNET documentation as well, it gives a nice blurb about applications, what it is and what types there are\:

C:\\\Program Files\\\OPNET\\\12.0.A\\\doc\\\modeler\\\pdfs\\\Applications_Model_desc.pdf

The main ones we will be dealing with (I presume) are FTP (file transfer), E-mail, HTTP(web surfing) and print. I would imagine that Remote Login would also be significant, but I'm not exactly sure if it is dealing with the network user login system. This applications pdf looks really useful!

{link:Return to Contents|#contents}

h2 {anchor:|27april}{anchor}27 April 2007 - More looking around

Today there was another meeting with Ilze and Selena, so we could chat about what happens next. Thanks to some excellent research by Selena, we have discovered that there is a built in package that deals with traffic flows and its statistical modelling under the name ACE. Our first instinct was that our license does not have this package as we are not under the full one, but Mike will investigate this.

Also Mike has sent e-mails around to Cisco who aren't being particularly useful saying that we don't have the right license to be given information about our missing switch, etc. However we have found a source who looks to have encountered similar problems to us in a university in New Jersey. Michael has sent out some e-mails to him as well so fingers crossed we may get some answers.

We also talked about the modelling of the disks in our network, and in fact the traffic flows to and from the disks are probably more important than the flows in other parts of the LAN. Meanwhile my uni workload has eased up immensely, so the forthcoming week should hopefully lead to some good progress.

{link:Return to Contents|#contents}

h2 {anchor:|30april}{anchor}30 April 2007 - Extending Bill's model

After quick chat with Michael and Percy I can now take Bill's model, consider three of the servers and make an alteration to it. So I will be adding four edge switches to the servers andromeda 209_108, engsci_linux 209_103 and leopard 209_98, as well as a couple of core switches and two disks. A number of servers currently present in Bill's topology have been deemed redundant and will be removed. Time to get cracking!

One more thing, the source found earlier (Mike \: Vasil Hnatyshin from Rowan University, NJ) that had the same switch problems as us said that they used a switch that was similar as the alternative and made some adjustments to it to meet their requirements. He also said he was willing to give us a hand building our desired switches, which is great news.

{link:Return to Contents|#contents}

h2 {anchor:|2may}{anchor}2 May 2007 - Quick Update

Had a quick meeting with Mike and Cam today. I have made a start towards implementing the edge switches in OPNET, but much to my frustration, D-Link is not one of the listed manufacturers in the OPNET database!! My best bet would be to custom make a switch that comes as a close match to the DGS-108 that we are after, there are a number of custom switches available in OPNET, the problem was choosing the right one. I settled on building an advanced ethernet 8-port switch, but at the moment I do no understand most of the specifics that govern this switch, which means more research for me. Also I should check if edge switches differ from core switches.

As for the disks, at the moment I haven't found a database for hard-drives in the OPNET directories, but there is bound to be something if I look harder.

{link:Return to Contents|#contents}

h2 {anchor:|3may}{anchor}3 May 2007 - Research of Manufacturers

My first job was to get a bit more familiar with some of the parameters that govern an edge switch, some of which include the packet service rates, the bridge parameters, the VLAN settings and the QoS system. All of these I have to define based on how the DGS-108 works. For a list of the specifications on this switch that I found, click {link:here|https://www.dlink.ca/products/resource.php?pid=449&rid=1721&sec=0}. I also read up details on various aspects such as ATM transfer, IEEE and MAC addresses to help broaden my understanding of this subject matter, the book that Cameron passed onto me earlier was particularly useful as well. The problem however is that some of these do not match up with the OPNET parameters, I guess I will have to do the best I can.

At the moment I am using a standard ethernet8 switch under the shared>ethernet>advanced menu.

Now onto the disks, we are after the Dell PowerEdge SC440 Server to model onto our network, but surprise surprise, it is not listed in OPNET again!! However there is a range of Dell Servers present anyway, and at the moment I think the best thing is to find a close match and tweak it a little to suit our needs. I have settled with the Dell PowerEdge 2050_2400 2.4Ghz Xeon processer for now. I will have to look into how I am going to implement the two 300Gb SATA disks onto these servers, but at least this is a start.

For a list of the specs for the SC440 Server, click {link:here|https://www.dell.com/content/products/productdetails.aspx/pedge_sc440?c=us&l=en&s=bsd&cs=04}. This is best viewed in Internet Explorer.

Another useful discovery is that the D-Link edge switch only operates with 1000BASE-T link supports, as opposed to the 1000BASE-X. Finding this out could have avoided some wasted time with the wrong links in OPNET! Also VLAN is not supported and the QoS operates under the IEEE 802.1p standard. I have implemented all this into my custom switch. And finally, edge switches and core switches are the same.

{link:Return to Contents|#contents}

h2 {anchor:|7may}{anchor}7 May 2007 - Querying Suppliers, and SATA disks

Perhaps I was looking at the wrong approach, the processors that I plonked in previously may not be the way to go, and perhaps I should be looking to put in a standard server and look for SATA disks on the inside configurations. It turns out there is something in there, and I have configured two disks, but the question is, are these disks simply internal to the server or separate, as I think we are looking for independent ones. I need to look further into that.

Some useful info, SATA II in OPNET represents SATA/300 disks. It seems that we do not have built-in 200Gb devices, so again I wonder if I am implementing the right thing. It was located in advanced_local storage system_interface config table.

What of the I/O Controllers? More food for thought.

I think I may have hit the jackpot for some useful assistance though, I have e-mailed a range of Cisco switch suppliers (as opposed to the manufacterers) and I do expect that someone will respond. I asked them what the major differences are between the 4506 and the 5000 switches, hopefully this can provide a bit of the legwork if we are to modify our OPNET 5000 switch.

{link:Return to Contents|#contents}

h2 {anchor:|8may}{anchor}8 May 2007 - Line Cards or Hubs!

Another discovery, the hubs that are in Bill's network shouldn't be hubs, they are in fact supposed to be line cards. Of course I barely know what these are, so I have pulled up a few manufacturer websites on those as well over {link:here|https://www.pcmall.com/pcmall/shop/detail.asp?dpno=365591&Redir=1&description=Cisco-Catalyst%204500%2048-Port%2010/100/1000%20Module%20RJ45-Switches%20/%20Hubs}.

I don't know if OPNET has line cards in there, that's something I will have to investigate. One of the suppliers did get back to me, and he tells me that quote "Some of the main differences between the 4500 and 5000 are that the 4500 supports POE (power over ethernet) and some additional protocols". Interesting. I have replied back to him asking for a little further elaboration, and I have also queried on whether the 2948 is a good match to the 2950 switch.

Another response posted a few more links on the products, which I will read further into and post up here in due course.

{link:Return to Contents|#contents}


h2 {anchor:|11may}{anchor}11 May 2007 - Quick Update

After Wednesday's meeting it was established that Bill's implementation of Hubs instead of line cards is fine , so that is something I do not have to worry about in the future. Annoyingly though, my manufacturer contacts have suddenly gone silent on me much like the CISCO support and the OPNET forums. Finding good help really is a mission it seems.

According to a few responses to Michael's contacts, the best way to approach the missing D-Link switches would be to build one from scratch and match up the missing parameters.

{link:Return to Contents|#contents}

h2 {anchor:|23may}{anchor}23 May 2007 - Applications

Great we now have our applications list, I decided its a good idea to type it up so that its a bit more secure than on a loose piece of paper! Here is the list Percy has provided (as exactly on the sheet given):

List of Applications:

Linux Server: engsci-linux

Applications: LDAP – user authentication
Samba – File sharing
Print Server – Lab and Staff Printing
Quota Management App – Printing only for Labs
Users Home Directory – All Lab Users
Kerberos – Authentication Mechanism
nfs – license server for cplex and ampl
(there is an app for passwords too I believe)

Windows Servers

Staff Doman

Alpha/Anaconda – Two domain controllers that authenticate users (staff and grads). It also provides Print Services and act as print servers to staff and grads. Active directory – Kerberros
Andromeda: File and print server (all users)
Home directories and shared workgroup folders that only staff share.

Lab:

Bobcat/Puma – These are the 2 domain controllers that only authenticate users.
Leopard – File and print sharing.
All users home directories are used on this servers (only for labs)

{link:Return to Contents|#contents}


h2 {anchor:|2jun}{anchor}2 June 2007 - Beginning of Exams

Mid-semester exams are about to begin and much of my focus will be going into that for the time being. I will continue work after the last exam (which is stretched all the way out to the 25th!)

{link:Return to Contents|#contents}


h2 {anchor:|27jun}{anchor}27 June 2007 - End of Exams

Exams are over finally, however it seems that the OPNET software license has expired! This is of particular annoyance as I was hoping to make use of the couple of weeks break I have. The best I can do is read up on the implementation of applications and link them to the apps list that Percy had kindly provided before the study break. I have a number of other random priorities outside the University I have to take care of in the next week regardless.

{link:Return to Contents|#contents}

h2 {anchor:|03jul}{anchor}03 July 2007 - A Few More Delays

I tried to set up a meeting with Michael however webmails ongoing crashes led to a bit of misunderstanding and Mike has flown out before I could get the chance. He will return on the 14th and until then OPNET will most likely be out of action. So my priority now is to work on my interim presentation that will be due first week back. I have started making a few outline slides, the main thing for me is to make sure I have this project crystal clear in my head before I go out to speak !

{link:Return to Contents|#contents}


h2 {anchor:|10jul}{anchor}10 July 2007 - Presentation Slides

Most of my slides are done by now, well the drafts anyway. There is definately enough work done to give a five minute talk, mostly I will be covering what my project scope is, how much have I done with a few screenshots of the OPNET topology, and then finally what I am looking to do later on down the track. This template seems pretty standard for most presentations, but it does provide a nice warm-up for the real deal later on in the year.

{link:Return to Contents|#contents}


h2 {anchor:|17jul}{anchor}17 July 2007 - Presentation Day

Today I delivered the presentation and for the most part it went well. I had a chat with Mike later on that day and in particular about the OPNET license, things are moving a tad slowly and its beyond our control, hopefully OPNET is rolling early next week and until then I will be doing some catch up work on Applications and how to process them in OPNET. That is definately the next step in the process, getting those up and running based on the list that Percy has provided us earlier.

Hopefully we can get some random traffic data uploaded somewhere in the near future as well. The topology isn't quite fully ready as I am not 100% sure as to how the disks are going to work out, or whether I have put on the right disks at the moment or not. There are also the D-link switches, as mentioned before I have made a rough custom-made switch and matched up the parameters. Is this good enough? Until OPNET is back, we can't really tell !

{link:Return to Contents|#contents}


h2 {anchor:|24jul}{anchor}24 July 2007 - Report Meeting

Today there was a fairly big talk about how I should be structuring my final report, and in particular the how much will I be able to cover before the reports are due in, which by the way is roughly a month and a bit ! Which probably means some serious hustle for me. I have been kindly given a few past reports by Michael for me to read over, which I will do in due course when I start putting together the report.

And OPNET is still delayed..

{link:Return to Contents|#contents}


h2 {anchor:|31jul}{anchor}24 July 2007 - OPNET Returns

And we have OPNET back in action as of today! I got straight back to it and was familiarising myself with all the previous work I have done, my main priority for now is to finalise the topology and then get the applications uploaded and working. With reference to Bill's wiki, I had a look at some of the Applications he has already put on there. I'm not to sure what profiles are yet but I will get into that.

{link:Return to Contents|#contents}

h2 {anchor:|03aug}{anchor}03 August 2007 - Applications and Disks

Well its been the most productive week in a while. I have put through the applications in the "Application Definitions" in the upper left hand corner. I'm not to sure what level of workload I should be putting through, so at the moment it is under medium. This can be easily fixed later. Also I have assumed the print services are running and black/white printers.

I have assumed that all the authentication services will come under TELNET, at the moment though I have nothing for e-mail, web and database, I will talk to Percy about it in the near future.

Ok, now the disks, I managed to neglect a disk definition section sitting there at the top until now, and its good to know that there is a section for making custom disks ! This is great news, and with a bit more web research I had managed to match up most of the parameters for the 300GB Sata disks. I used the standard disks that are compatible with the PowerEdge SC440. The only question mark is over the maximum seek time (which I put down as ten seconds as a calculated guess) and also the number of disk cylinders which frankly could be just about anything. Is it crucial? I dont know, at the moment I have left it as unknown.

In the interface details for the SC430 Servers put up there (remembering that 440 didnt exist) I have now uploaded the right disks, which is definately a step in the right direction! The main concern now is how exactly do we monitor what is going on in terms of traffic flowing in and out of the disks, but thats something for the future.

{link:Return to Contents|#contents}


h2 {anchor:|06aug}{anchor}06 August 2007 - Profiles and Tasks

First of all just a quick word on the SATA disks I had manually implemented into the OPNET system, it turns out that the number of disk cylinders will not have an effect on output, however the maximum seek time WILL. I will have to hunt around and find some sort of a match for this.

Over the last week I have been doing a fair bit of reading through the OPNET help files and tutorials, and there is a section that governs storage partition and disk statistics, and these are produced as a default among all the other stats. This is good news as I won't have to dig around finding information about the disks once some sort of traffic has been implemented. We are interested in Server Jobs in terms of the type of study we will run under OPNET.

Now that the applications have been loaded on, the next job is to have a look at profiles. Profiles are essentially a definition of apps that will run for each subnetwork. In other words each subnet will have its own designated set of applications that run for specific time lengths or frequencies. It is up to me to find this data.

After that we then have to define tasks, but this is somewhat trickier as I am not 100% sure what they are. At the moment it looks like they are there as specific data transfers between servers at certain times, which we define. Are these necessary? Not sure..

{link:Return to Contents|#contents}


h2 {anchor:|16aug}{anchor}16 August 2007 - Background Loading

The main question heading into this week was that of tasks, do we need these at all? I read through the help pdf's extensively and I found a useful section on Background Loading. (This is found by selecting a series of links and choosing edit attributes > background load file > intensity). Here we can upload a background loading on the links based on user input or an external text file. This could be the sort of analysis we are after, and at the moment it looks like only Apps and Profiles are required here.

The link capacities vary from 1000BaseX to 10BaseT. I will need to investigate what these capacities are in terms of bps, and then it would be an idea to come up with an arbitary text file with a range of load values in bps over time. What kind of stats will it produce?

{link:Return to Contents|#contents}



-- MichaelOSullivan - 16 Dec 2010