What you will learn from this tip: This tip can help you maximize uptime by limiting bandwidth, distance, latency and congestion issues in your cable plant.
Cable management is usually an afterthought when implementing a new SAN infrastructure, but can have a dramatic impact on the overall availability of your data.
The issues that affect SAN functions include bandwidth, distance and connections, latency, congestion and cable issues.
Bandwidth
This is the speed of your storage network. SAN equipment can be purchased today using either 100 MBps (1 Gb), 200 MBps (2 Gb) or 400 MBps (4 Gb) parts, with 10 Gb on the horizon. SAN equipment vendors try to be nice by making all the faster stuff backward compatible with the slower stuff. The problem is that everything in the path always slows down to the speed of the slowest part in the path.
In order to reduce issues with speed negotiation between components, it is sometimes beneficial to hard set all the components at a fixed speed. If everything is 4 Gb capable, then manually set each component to 4 Gb to reduce speed negotiation overhead and errors.
In other words, use all the same speed parts and use the speed your applications need.
Distance and Connections
Distance issues can be the cause of all kinds of flaky intermittent gremlin-type problems. In a 1 Gb SAN, there can be no more than 500 meters between devices. Two Gb limits you to 300 meters; 4 Gb is even shorter. When upgrading from 2 Gb to 4 Gb, you need to make sure the distance between your hosts and SAN storage is still within the appropriate range.
When using a patch panel, signal loss is a concern. Each patch can lose about .5 decibels (db) per connection, Try and keep total loss under 4.0 db for all connections per link. Limiting signal loss is key, so try not to use more than two patch panel connections between your server and the storage in a SAN.
Latency
Latency is caused by too much distance or too many hops between your servers and storage. A hop is defined as a connection through a switch. If you connect a server through one switch, that equals one hop. Hops can add milliseconds of delay through the fabric. One or two hops is normal in a well-designed SAN. The more hops the data takes, the higher the latency.
Congestion
Congestion occurs when there are too many things happening at once over the same connection. Congestion in a SAN is usually called "over subscription." Avoid congestion by using a separate backup fabric. If your data needs to traverse an ISL to get to storage, then make sure you are using enough connections to satisfy the bandwidth's requirements. Switch "trunking" (interconnecting switches in local area networks to form larger networks and to interconnect to a wide area network) can help in this regard.
Cable Issues
Some companies try to save money by reusing 62.5 micron cables used for optical Ethernet. The two cables have different core sizes. Using mismatched core sizes will cause signal loss. Stay with the 50u cables built for SAN.
Fibre cables use light to transmit data. Dispersion and absorption of light due to micro or macro bending of the cables can cause signal loss. Avoid excessive crimping or bending.
When implementing the cable plant for a SAN, proper design and standards should be followed to help avoid those intermittent problems that can keep you awake at night.
Do you know...
How to safely relocate your SAN?
The reasons for zoning your SAN?
About the author: Christopher Poelker is the co-author of SAN for Dummies.
