Switches - Someone please explain difference Managed / Smart / Unmanaged

Here's what I have in my network:

7 pc's plus one server plus router/modem.  The server has an application, a database and processes reports.

The server application is run by a shortcut on the client users' desktop which is through a mapped drive.  I understand that the processing takes place on the server, then.  The client systems are thin clients then?

My question is...
I need a new switch because some PCs are needlessly hogging all the resources.  I would prefer if 2 clients plus the server have more bandwidth.  It would be even better if I could view these connections like a router and set priority when I want to.  Better yet would be if it could sense which connections are important and reduce bandwidth for others.

What I really hate is that a vital process is shared 50/50 with something low bandwidth.

What switch is ideal for me?  What does each type do?  My budget is 80 - 200 british pounds.

Let me know if you need any further information.  I'm fairly knowledgeable in IT generally and don't mind researching things myself to get the right system.

Some links I've found which might help you?
Who is Participating?
I wear a lot of hats...

"The solutions and answers provided on Experts Exchange have been extremely helpful to me over the last few years. I wear a lot of hats - Developer, Database Administrator, Help Desk, etc., so I know a lot of things but not a lot about one thing. Experts Exchange gives me answers from people who do know a lot about one thing, in a easy to use platform." -Todd S.

What you describe does not look like a thin client setup to me.  If the apps are started from a shortcut mapped to a network  share  then are running locally on the client. You would need to use Remote Dektop or Citrix sessions to run the app on the server.

As for switches you cannot usually manage the bandwith on a per port basis.  A switch is designed to provide full bandwidth to a connection as opposed to a hub which shared the bandwidth between all connections.  This provides best use of the available bandwidth as each client gets a full bandwidth link to the server.
A managed switch provides other facilities such as VLANS and Diagnostics.  
Please let us know what your current hardware is.  Make and Model of your switch for instance.
Do You Have a Trusted Wireless Environment?

A Trusted Wireless Environment is a framework for building a complete Wi-Fi network that is fast, easy to manage, and secure.

To be quite honest, with that switch and just 7 users I don't thnk your network infrastructure is the problem.  This should be quite capable of providing a good level of performance.  It seems that either your server is being overworked or your clients are overworked.  If you are downloading the app from the server to be run locally you could try creating a local copy of the app on each client and running it from there.  This will complicate maintenance as you will have to update each machine when the app is updated but if the app is only used on a couple of machines it shouldn't be too much of a problem, and not having to download the app from the server will ease the load on the server and network.
Databases are usually heavy workloads and producing reports is one of the heaviest tasks they do, so I would have a look at your server.  Maybe additional memmory could be added if you aren't at the capaciity for your machine ?

Experts Exchange Solution brought to you by

Your issues matter to us.

Facing a tech roadblock? Get the help and guidance you need from experienced professionals who care. Ask your question anytime, anywhere, with no hassle.

Start your 7-day free trial
chriscounter07Author Commented:
I thought I'd take a moment to run though your answer :)

DCMBS:             To be quite honest, with that switch and just 7 users I don't thnk your network infrastructure is the problem.  This should be quite capable of providing a good level of performance.  It seems that either your server is being overworked or your clients are overworked.  
Each client is accessing the database all the time.  However sometimes one client will be very slow while the others are faster than normal.
Conversely I find that after heavy periods with all users, followed by activity of only one user the system is still slow.  Lagging if you will.

If you are downloading the app from the server to be run locally you could try creating a local copy of the app on each client and running it from there.  
Traditionally when processing anything on the server it is very fast.

This will complicate maintenance as you will have to update each machine when the app is updated but if the app is only used on a couple of machines it shouldn't be too much of a problem, and not having to download the app from the server will ease the load on the server and network.
The app is used on all machines, though some people e.g. managers 'require' a better services.

Databases are usually heavy workloads and producing reports is one of the heaviest tasks they do, so I would have a look at your server.  
So if a client connects through a mapped drive to the server AND they dont have anything in their program files Your opinion is that it is still the client that runs the program and Not the server?
Maybe additional memory could be added if you aren't at the capaciity for your machine ?
Like I said, server is Very fast.  I completely understand your reasoning though.
Heres the server FYI:
If you do want to upgrade your switch I would reccomend the following HP procurve


The important figure when looking at swtiches is the data through put or switching rate

On the Netgear you have this is 100 mbps (100 megabits per second).  The HP procure has a Switching rate of 2.6 Gbps (Gigabits per second) which is vastly faster than the Netgear.

It is still an unmanaged switch but with 7 users I really don't think you need a managed switch
Unless you are using a remote desktop session or Citrix then the app is almost certainly executing on the client.  There is not always a need to have anything in program files, especially if the app is just a standalone exe file which doesn't need to load dlls etc, this is common with custom apps although rarer these days.  I have often run into issues with performance when generating reports in a client server scenarion.  The performance is degraded if the reporting app is running on the client and getting data from a server based database as each query has to be transferred across the network and the results sent back to the client.  When generating reports this can be an intensive process.  Database programs (You haven't said which one you have i.e SQL server) can usually make good use of all available memory for processing queries so if you can add memory it will usually be beneficial as it will improve the performance of the database in processing the queries.  You could try using a remote desktop to the server and running the app in the Session on the server, but to do this properly you would need to set up terminal services and buy terminal services licences.
Another switch for you to consider


This is a managed switch with two Gigabit Ports.  You can therefore connect your server to one of the gigabit ports which will provide a gigbit backbone and ensure that your server is able to service all the clients with a good level of service.  It also has a data throughput rate of 9.6 Gbps
Just going through your sever specs I see it has Dual gigabits NICs so potentially you could team them and connect to the two gigabit ports on the 1724 switch which would be good for your scenario
Apologies the link in my previuos post seem to have got broken.


Also I am not recommending any particualr vendor.  I have linked to the model just to get product data.
Just managed to work my way through your list and I see the 1724 is included in your list so from your list this is the one I would recommend.
chriscounter07Author Commented:
will reply in a moment DCMBS ;-)
Just a bit more information for you chris, im suprised no one mentioned this what your after is called QOS (Quality Of Service).

QOS allows you to specifiy which network protocols should get more bandwidth than others and vica verca. im not sure if you can get a switch for under £200 but i had the same issue at home and bought the below switch as it does something simular to QOS.

I would also google QOS and learn what it does.

Powerconnec 2708 is an 8 port gigabit switch that
does traffic queuing for £59+vat

Kind Regards
I'm not sure QOS is relevant here.  QOS can prioritize traffic according to type and is usually used to improve Voice over IP traffic  on a LAN.  I am not sure you can prioritze general TCP traffic from workstation over another or by application.
Several red flags are here are not mentioned...  

The server might sound fast, but the listed model of PowerEdge is not really very fast at all.  It is really just a Workstation class machine with extra drive bays.  If it was ordered as a dual CPU, Quad Core 3GHz server with 16GB of RAM and 64-bit Windows server, on 6 or 8 RAID 10 450G, 15K RPM SAS drives with a battery backed RAID cache....  Now that would indeed be fast, but >12K$ (US) that is not likely.

If the applications are all sharing a single 100Mbps link to the server, and running their application FROM the servers shared disk, that could be a big problem.  A local SATA drive in a PC is the same as a server SATA drive and you are asking the ONE (mirrored RAID1 drives are really just one) disk in the server to do it's own work, plus let SEVEN workstations use it for the applications as well.  So the server configuration matters a tremendous amount and the more details provided the better.  For so few systems, an unmanaged GigE switch will work just fine and will provide much more total bandwidth for the server and keep you on budget.  Managed is better of course, but costs more $$$

The server disk subsystem is a key component here in performance.  You might be able to install the application on each workstation and that would move the application disk loads (except for the data that must be shared for the applicaiton to work), across 8 machine SATA disk drives instead of the one server disk.  As mentioned that creates updating issues with the application.

In this case a GigE unmanaged switch is better than a 100Mbps switch (managed or not) since you can only use about 10% of the servers disk performance with your current configuration.  If the application can run from the server console and run well, WHILE the users are sharing it and having performance problems, then the network is the most likely culprit.  If the server runs the application slowly as well then most likely the network is not the bottle neck.  I would still change the switch out for a GigE switch and there will be some improvement for sure.  (SATA drives can push about 1Gbps of data and you are throttling the seven machines sharing the server down to 1/10th of that with a 100Mbps switch.)

Connecting both GigE ports on the server to the sme network may cause more trouble that you fix.  If you can't fill one GigE link, then two won't help.  Getting the clients all on GigE is more important in this case than just the server.  The clients probably cant fill the first GigE server link anyway with their application.  The combination of the apps and disks are likely just too slow.

However...  The right way to fix this and be sure you really fix the cause is to gather more data on the server while the users are connected to it and running the applications.

Windows Performance Monitor will tell you what is going on if you can spend a few hours watching, learning and testing.  Then you will have accurate information about what the correct fix really is and be sure you can really address the problem.

Really...  :-)

If you want to learn more about tuning Windows Servers, Microsoft has some very useful information.



Also there are quite a few very useful links here that will assist you in understanding Windows Performance...

chriscounter07Author Commented:
thanks for your post CoreyMac and to the others.

Guess what, I've just learnt that of the 7 machines one of these also has a database to which 2 other machines connect.  So now I have the server hosting a db and one of the clients hosting a separate db.  
This is too much for one switch so will review your posts and stick a 4 port mini switch for the second db machine so it's not interfering with the rest of the switch.

will award points soon.
While I agree that keeping the traffic separate sounds helpful, it is hard to see how that could be a root cause of your problems.

Putting everyone on the same 16-port GigE switch should work just as well as splitting them between two switches.  Potentially better since the only bottleneck would be each server/workstation individual port and no user can really affect another...

Modern switch designs are such that most of the time, switch bandwidth and/or congestion should not be an issue with a small number of users (<16-25).

Just my opinion though...  In generally, one 16-port GigE switch is better than two 8-ports connected together and GigE is much better than 100Mbps in almost any configuration (assuming the workstations and/or servers can connect at GigE rates)...  Latency is simply better even if the applications do not regularly use more than the 12.5MB/sec that 100Mbps Ethernet offers...

Really though I am not trying to rain on your parade here...  I am just suspicious that you are not going to be much better off as I agree with DCMBS that you are not likely switch constrained at this point...
chriscounter07Author Commented:
At the moment I presume the workstations are only running 100Mbps.  Each workstation is about 4 years old.  The switch is 100Mbps.
The server is configured to be 100Mbps but can be 1gig.  The cabling linking everything up is cat5e.

What will I need to do to get the network up to 1gig standard.  Perhaps the relative newness of some of these machine means I wont have to buy new PCI network cards?  Also will cat5e support 1gig?  
Cat5e is good enough for GigE.

The workstations are a ?.  You will just have to check each one.  Some might have it and others might not, as it just depends.  Updating the drivers for the workstations and server are a good idea as well.

Intel and Broadcom are among the best Ethernet Chipsets for NICs if you have to buy new.  Intel is easier to find in most markets, but not all...

PCI Express is better for GigE than PCI.

Otherwise, it should be pretty straightforward, and for the Broadcom/Intel NICs, newer is generally better.
chriscounter07Author Commented:
I've installed a gigbit switch the other day and I personally see a difference although I haven't run any benchmarking software as yet.
Annoyingly the chief user I wanted to improve the QoS actually says it has got worse for her though. :-(     And I made a point of not telling anyone that I had made the change.

Will look at updating drivers on one workstation and see if I can make that gigabit compatible by the end of the day, run a performance test on it compared to 'normal' and hopefully try and isolate where the problem is.

It's looking like either the server or the enormity of the database on the server which is the issue.
chriscounter07Author Commented:
I believe now that it is a case of the Server not being sufficient for the job.  Perhaps it is the way the database is read - or the type of database.

Either way, a gigabit switch proved much faster initially though with all users accessing the Server DB this quickly slowed
If you really are disk bound on the server and SATA is your connection link for drives, then  there is a simple fix if you are still interested.

First thing to do is to look at the Performance Monitor stats on the server when the users are making it slow down.  If the disk where the database resides is getting swamped and really is a single SATA drive, then the queue depth should start to climb way up once the users start hitting it.

Just add an Intel 80GB G2 SATA SSD to the machine, move the database files to it and your problem will likely imrove massivley if not disappear completely.  The Perfomance Monitor numbers should show exactly how hard the disk is getting hit in I/Os per second and MB/sec.  Watching these and the disk time per transaction would be key to moving the bottlenecks somewhere else.

If you want to discusss this more, let me know.
It's more than this solution.Get answers and train to solve all your tech problems - anytime, anywhere.Try it for free Edge Out The Competitionfor your dream job with proven skills and certifications.Get started today Stand Outas the employee with proven skills.Start learning today for free Move Your Career Forwardwith certification training in the latest technologies.Start your trial today
Network Management

From novice to tech pro — start learning today.