Q
Problem solve Get help with specific problems with your technologies, process and projects.

# The problem, believe it or not, is the speed of light

We are doing some testing on two SAN islands linking by a CWDM link. However, we have observed performance hits as distance gets longer. What is the possible reason?

Performance measurement:

- 0km (3meters) : OK
- 12.5km : 10 - 15% lower
- 25km : 20-30% lower

Possible trouble spot:

1. I thought the FC 10km limitation only applies to the FC optics itself.

2. For the CWDM system, each lambda is 2.5G so each Channel can actually takes two 1G FC and "mux" it to a 2.5G lambda for the CWDM filter. However, the CWDM system "should" be a transparent box that takes FC lambda as the vendor claims.

3. The HBA/RAID controllers and disk enclosures are all 2G but use multi-mode optics.

The problem, believe it or not, is the speed of light.

Fibre Channel uses optical fibers for data transmission. Data is transmitted over the cables using serial SCSI over FC protocol. This means that every bit of that block of data you just transmitted gets converted to an optical pulse of light and gets sent over the cables through your CWDM multiplexer and over to the other SAN island. (WDM or Wave Division Multiplexer, is a technology that "multiplexes" multiple sessions over a single fibre cable by slicing up the light frequency into many frequencies that can then be used to transmit multiple data streams at the same time.)

The speed of light in a vacuum travels at around 300,000KM/second The speed of light over a single mode fibre cable travels at around 200,000KM/second

That gives us a "latency" (the time it takes for the light pulse to travel across the cable) of around 50 microseconds per 10KM. Therefore, latency over long haul transmission cables (9u SM fibre or "dark fibre" as it is called) is approximately 1 millisecond for every 200KM traveled.

As your distance increases, so does the latency of getting your data to the other side. We have not found a way to speed up light yet. (By that time, we would all be carrying around "phasers," and saying things like "beam me up Scotty!)

Data transmission over long distance affects performance of production applications, especially when that transmission is done "synchronously". A better method would be to use asynchronous transmission methods where you "spoof" the application into thinking the I/O is complete but the actual data transmission happens behind the scenes through the hardware.

A number of storage vendors have this capability, mostly built into "replication" firmware.

Connecting two SAN islands together using wave division multiplexers is usually only viable up to 100KM. Make sure the switches you are using includes support for "extended fabric". Brocade has an optional license for this that is used to increase the total amount of buffer credits for your data. McData includes this capability with their switches. You need enough buffer credits in order to "fill up the pipe" if you will. Sixty-four buffer credits should be a minimum when connecting islands together.

Chris

Editor's note: Do you agree with this expert's response? If you have more to share, post it in one of our .bphAaR2qhqA^0@/searchstorage>discussion forums.

#### Start the conversation

Send me notifications when other members comment.

## SearchDisasterRecovery

• ### Ensure your DR strategy meets the challenges of a new decade

Traditional disasters and new cyberthreats have organizations reexamining their DR plans to ensure they maintain high levels of ...

• ### Zerto CEO: DR has become something real

DR is no longer just about compliance, as enterprises find maintaining infrastructure uptime is critical to their business. This ...

## SearchDataBackup

• ### Cloud key to top 2019 data backup and disaster recovery products

Commvault, Cohesity and Actifio make the podium in our Products of the Year competition for backup and DR. Cloud Daddy wins the ...

• ### Ensure your backup infrastructure has sufficient networking

How can you effectively modify your backup process to fit your network speed? Because bandwidth constantly varies, employ best ...

• ### New Infrascale CEO outlines multi-cloud backup and recovery vision

Infrascale's new CEO has his eyes on the 'multi-cloud' sky. The vendor plans to roll out updated backup and disaster recovery ...

## SearchConvergedInfrastructure

• ### Best hyper-converged systems win Products of the Year 2019

Datrium, Western Digital and HiveIO win praise for ground-breaking products that led the way in a year of hyper-converged and ...

• ### What is Intel Rack Scale Design and how does it work?

Find out what Intel's Rack Scale Design architecture is and how its reference code and APIs enable Intel-partner products to ...

• ### Nutanix Karbon upgrade focuses on non-Kubernetes experts

Nutanix aims Karbon at 'mere mortals' who want to run Kubernetes in on-premises data centers, but it faces plenty of competitors ...

Close