Infinite I/O

Storage performance: Traffic jam at the speed of light

Posted by Matt Brender on May 22, 2014
Find me on:

We just started a webinar series with well-known vExpert, blogger, and speaker Eric Siebert which he titles “Top 7 Storage Performance Gotchas.”

In this first round, Eric broke down some fundamental facts that often get abstracted away as we discuss performance analysis. I want to review just one that really struck me and then answer a webinar audience question that came in too late to answer live.

Observation: You can’t always make storage performance faster with more bandwidth.

Cue Eric’s metaphor of where the network fits architecturally:

  • Bandwidth is the roadway
  • Protocol is a car
  • Speed limit is the speed of light

dump-truck

Here’s how I internalize it: you can’t change the speed limit by adding more lanes to the roadway. You can prevent congestion, which may or may not be the cause of storage performance concerns.

A viewer asks...

One question that came in at the end of the webinar that we didn’t get a chance to answer was this:

Any considerations for impact on operations and array-based data devices, backups, snapshots, replication, etc?

I love this follow-up question because it highlights, yet again, how storage can become a bottleneck. Any advanced services that are array-based require further operational overhead for the storage system. As we’ve seen with customers (which I blogged about previously), Infinio can have the benefit of reducing CPU utilization on already taxed centralized systems. It brings to light the distributed nature of our solution — allowing the offload of a significant portion of data management that reduces resource contention throughout the storage stack, CPU included.

These observations are just two of many to make from Eric’s work. I highly recommend setting aside the time to listen to this and the next two talks. You can sign up for all three of them here.

____
Matt is a Sales Engineer at Infinio

Topics: Talking Tech, Events