IT Employment

Ask something
Discussion Locked

Chassis versus Stackable

0 helpful
11 replies

by nd_it

Closed to new repliesThis discussion is closed. It stays readable, but it takes no new replies.
I have just be handed the responsibility of designing the LAN inframstructure of a new manufacturing building our company is expanding into, something that is a first for me. The estimates right now, is the building will roughly cover 250,000 to 300,000 square feet, with about 4 wiring closets to start with. The new building will start out small for the first couple years, with about 100 or less end nodes. I plan to use Gigabit in the infrastructure. Which would meet the requirements to start up this new building and allow for future expansion, chassis or stackable? All while retaining redundency. Cost, as it always is, plays a big part in this. Any suggestions would be greatly appreciated.

11 replies

by road-dog

I would suggest the stackables to start, 24 port switch in each closet will meet your ~100 port initial design.

Unless one of your closets is your "main closet" remember you will need a core switch to aggregate your Gig links. It will require a gig interface for each closet, at least to start.

Depending on what you are manufacturing, your environment may be hostile to networking hardware. Make sure that the closets are air conditioned with filtration in case of dusty environment. Your closets should also have solid reinforcement or barriers around them. I've seen too many wooden comm closets get taken out by forklifts.

Don't try to be too thrifty, run at least 12 fibers to each closet. You WILL find a use for them, even if it isfor security cameras.

by nd_it

I was thinking stackable first also. Our current facility has 6 wiring closets with one closet being the "main closet" with a stack of six 24 port 10/100 Fast Ethernet Switches. The remaining five have anywhere from one to five switches of the same configuration. Each of these closets have 2 pair of fiber running back to the main closet. We implemented this infrastructure about 3 years ago, before I started my position. Is there a new or better way of structuring this?

by oldefar

For many manufacturing environments, wireless makes a better solution than wired. There are several reasons for this -
- Electrical noise from large motors
- High traffic/heavy traffic (carts, sleds, etc.)
- Flexibility for manufacturing line changes
- Resiliancy/redundancy

Going wireless changes the stackable versus chassis discussion.

Manufacturing also lends itself well to terminal server environments since the machine operators are not involved with writing documents, building presentations, or doing spreadsheets. Plant management may be, but from their offices.

Process control is another issue. Depending on what is being manufactured, you may have to deal with insurance liability constraints as to what type network is acceptable. However, most process control systems are not traffic heavy in either direction.

by omie

In my own opinion, wireless is not a viable option. Electrical noise in wireless is also a noise in wireless. The EMF generated by the transformers and cables will greatly affect the link strength. Even the humidity greatly affect the signal specially when the building walls and floors are concrete.

In my own experience, I spend more time in responding to wireless network problem compared to hardwired LAN.

Again he is looking for a GBIC infrastructure to begin with and in wireless you have only about 11 mbps at its peak but usually dependent on how many nodes connecting to a paricular access point. Based on my knowledge the max no. of nodes per AP is 21 nodes and the bandwidth is on a crawl.

In terms of cost it is the same as whenyou hard wire the land considering the cost of AP and the card.

by oldefar

Your points are valid in a large office environment, but not in a plant environment.

EMF from the motors is outside the spectrum used by wireless, and has no impact on signal strength. Humidity has an affect over distances exceeding the space indicated in the original post. Manufacturing space tends to be open, so walls and floors are only at the boundaries. Spread spectrum systems work around the pillars and equipment when properly positioned.

A wired office complex faces environmental problems between workstation and service jack, and in the wiring closet. Most of the horizontal and vertical runs are protected within walls, ceilings, conduits, walker ducts, or floor plenums. Not so on the plant floor.

The high bandwidth being considered is overkill for most process control and production control systems used by plant workers. However, the equipment manufacturers build to the larger office use environment making the cost savings of a lower bandwidth between plant and external locations negligible. However, on the manufacturing floor the issue is resiliancy and reliability, not bandwidth.

As with any design, business objectives and requirements should drive technical requirements. My suggestion to consider wireless is based on my experience in manufacturing environments and must be validated by the poster's unique situation.

by oldefar

More, faster, better is overhyped IMHO.

Computer based tasks have several time elements. Keyboard, or user skill, is a part. Workstation or terminal processing is a part. Server or servers is a part. Communications is a part.

From experience I have found that in the LAN environment, communications is 10 percent or less of task completion time excluding keyboard time. To determine this, the keyboard element is scripted and data analyzer used to measure traffic by volume, direction, payload, overhead, and time stamp. Total wire time can be determined by total bytes times 8 over LAN bandwidth. The speed gain achieved with higher bandwidths only impacts this element of total time. In a LAN, the signal speed is less than the ethernet packet preamble time, making it a non-factor for your time analysis.

Bandwidth also impacts total number of users who can utilize the shared media before creating congestion. How much traffic the tasks performed generate determines where thisoccurs, not the number of users.

Even in office applications, the speed factor becomes negligible somewhere around 256K and bandwidths above that are only addressing congestion. With switches and VLANS instead of pure hubs, the extra bandwidth at the station adds little value unless dealing with VoIP or video to the desktop. Full duplex versus half duplex - no advantage with single threaded client server exchanges.

by nd_it

We currently do have a small wireless LAN in place on the productions floors with only about 15 nodes that connect to 4 Cisco 350 access points, (mind you that are current facility is about 500,000 square feet). The WLAN is used for mostly data terminals the warehouse uses for tracking inventory, and a couple programming laptops to go online with programmable logic controllers. Bandwidth I do believe will not be an issue on the production floors, since most of the traffice is coming from PLC's, which only have a max rate of 10Mbs. There are more PLC's talking with the network, then there are PC's

by nd_it

What I like about the wireless also, is the fact is that line changes happen all the time, so I usually have to spend a day or two, just moving ethernet cables. Just another comment to add.....

by oldefar

I am happy to see confirmation of my assumptions.

Regarding the original issue, I see only one drawback to the stackable solution in manufacturing. That drawback is with sensitivity to vibration when interconnecting the units with ribbon cable extensions. I suspect that the single board backplane of a chassis will be more reliable.

If the stackables are used to build independent LAN segments then they offer greater flexibility and lower intial cost.

by nd_it

I see your point, but is a chassis going to be more redundent compared to a stackable? If the whole backplane fails, you lose network connection from all the nodes, where as a stack, if one swtich fails, say you have 24 devices connected to that switch, you only lose 24. The rest of the stack still functions. But I see your point in our type of enviroment. Our enviroment is clean, but there is dust, and there are alot of other machines with electric motors running 16 hours straight. Thanksfor all your input, this is helping alot.

by oldefar

I haven't looked at a chassis in a couple of years. Used to be, the backplane was primarily passive in nature and so highly reliable. Failures were mechanical in nature, caused by improper card insertion. This held true from the late 1970's through the late 1990's. Failures were few, and always attributable to poor maintenance activity. The active boards, properly installed, vibrated in tandem with the backplane and so limited the problem of vibration failures. Shock mounting the rack or chassis completed the solution.

I suppose a properly mounted stackable solution may have the same characteristics. I am looking at a satellite system on my desk - stackable design, transmitter and receiver, with an external backbplane interconnect. Units are secured together and I suspect will vibrate as a single unit, limiting the potential of vibration loosening the interconnect cable. In addition, it has all screws installed to secure it, and I could add some locktight to be certain.The MTBF of a system increases with the number of components, so adding units has a negative impact on MTBF. However, these days equipment that lives for 90 days will tend to live beyond its useful life unless subjected to an external force like power surges, metallic dust, or moisture.

I would have to say that cost and flexibility rather than hardware failure and vibration sensitivity would be the key factors to look at.

Thanks for making me rethink this.

This discussion is closed to new replies.