RSS
Email
Print
ShareVirtual Instruments: Imagine an MRI in every home
Posted by Frank Berry on June 23, 2009
Once in awhile we run across a company that is doing something for data center administrators that is orthogonal to what everyone else is doing. One example is Virtual Instruments. A spin-off of Finisar in 2008, Virtual Instruments is an exciting new company that has taken existing pieces of very sophisticated technology, added some more sophisticated technology, and redefined monitoring solutions for SANs in virtual operating environments.
Magnetic Resonance Imaging (MRI) equipment is expensive. 3.0 tesla scanners cost a hospital over $2 million dollars - which is why patients are charged about $3,500 for a single procedure. But if you're really, really sick, MRIs produce high quality images of the inside of your body that help a physician quickly diagnose and correctly treat your medical condition.
Now imagine if there were MRIs with hardware so affordable and software so friendly that we could all have one in our home. Instead of using the procedure only after a doctor was unable to diagnose our condition, we would step into our home MRI each morning for a preventative look deep into our internal organs. We would be so much healthier. Fitness fanatics would know if they are maintaining peak performance. And those of us that are less fit would receive alerts long before an issue becomes a serious problem.
Protocol analyzers are MRIs for a network. When a data center network is really, really sick, specially trained engineers are brought in to use the analyzers to monitor packet traffic and quickly pinpoint the root cause of a problem. A few of the largest data centers have their own analyzers, staff and home-grown processes for collecting additional information, setting acceptable thresholds, and firing off alerts to identify issues before the health of their SAN fails. But most data centers can't justify the capital expense and specialized skills needed to deploy analyzers. So they pad the capacity of their networks to accommodate potential overload or failure.
Comment by XgigGuy on June 23, 2009 4:02 AM
Hey Frank, the last thing and SAN admin will do is add an unknown piece of 3rd party hardware in the production data path. We use Finisar boxes when we go down and need to help the vendors understand what happened. In reality half the time the issues go away when we introduce another point on the line.
Nice idea for a lab but how do you see this playing out in production environments? Why not use existing SAN tools form the fabric or host guys?
Reply to this comment
Comment by sanguy32 on June 23, 2009 1:10 PM
The Virtual Instruments products operate "out-of-band", so they are not in the production data path. They are just monitoring the data flows and the info about the data flows (like latency stats). Admins can use the software-only monitoring part of VirtualWisdom or implement the "best practice" of "TAPping" the Fibre Channel network when new storage or switches are deployed. IT organizations have been tapping ('instrumenting') their IP networks for well over a decade. Now SAN managers can do the same for their SANs and accelerate trouble-shooting and optimize performance in a non-intrusive way. The products from the existing fabric and host vendors have multiple deficiencies. First, they are not looking at the Fibre Channel frames as they are really only looking at port utilization. They can't measure end to end latency across the SAN, HBA queue depths, etc. Also, when traffic is at its peak, nearly all SAN switches de-prioritize their measurement capabilities in favor of the actual traffic. Monitoring essentially turns off, unfortunately at the worst possible time. The Virtual Instruments solution, being out-of-band, doesn't have this problem and provides a much deeper level of information about the overall SAN and virtual infrastructure.
Reply to this comment
Comment by frank.berry@itbrandpulse.com on June 23, 2009 3:04 PM
Xgigguy - you might want to look at the VirtualWisdom dashboard to see other examples of how it would play out in a production environment. For example, the link utilization meters will tell you that performance thresholds are bing exceeded for specific virtual machines and it's probably a good time to execute a vMotion. - Frank
Reply to this comment
Comment by Anonymous on June 23, 2009 6:28 PM
Greater monitoring of SAN-traffic lanes with real-time data points, better utilization management of current storage network assets; delay redundant or unnecasary storage capital expense... sounds like good technology to deploy in a down economy!
Reply to this comment
Comment by Curiosityktc on June 24, 2009 12:27 PM
Hi Frank,
I know that your background is in marketing, but I was wondering if you received any compensation from Virtual Instruments and/or have any current or future financial business ties or aspirations with them. As I do not know the rules for B&S bloggers on disclosure, I just wanted to check.
Reply to this comment
Comment by ItsSpelledFibreChannel on June 24, 2009 1:56 PM
It is hard to actually accept advice from someone who doesn't know how to differentiate between Fibre and Fiber. You might want to research what you are talking about before praising it.
Also, I find this article hardly credible as the word on the streets is that Finisar sold off the product line to VI because it was an abysmal failure, not because they didn't like to make money. SAN Monitoring technologies are critical to many industries, I doubt this one is.
Reply to this comment
Comment by frank.berry@itbrandpulse.com on June 24, 2009 2:10 PM
Curiosityktc - For the record, I don't receive compensation for blogging. I should. When I write about vendors, products and initiatives I believe in -- competitors thrash me!
Reply to this comment
Comment by RyanP on July 1, 2009 2:17 AM
I've been using NetWisdom in my shop for a little over a year now. I run a medium sized infrastructure (1600 SAN ports total), and support some data hungry data warehouses (they'll run 40,000 IOPS for hours during some peak periods). Any performance issue used to ALWAYS come back to the SAN. I used to joke that Oracle's recommended read response time is officially 'faster than what you are doing today'. As a SAN guy, I hated going into 'all hands' production issue meetings with little to no data on actual SAN user experience.
I wrapped NetWisdom (specifically TAPs and Probes) around my most critical database storage ports. Instantly I was able to see some performance bottlenecks (queue-depth on some HBAs was set to 2, read block size was 1MB on my OLTP DB, etc...) I was finally armed with real metrics, and easy to understand graphs that put my Unix guys and DBAs on notice; the SAN guys were not their whipping boys anymore. :) The bottom line is that switch manufacturers give you kids toys compared to the data NetWisdom can provide. And relying on storage vendors for performance information is like asking your mortgage company if your interest rate is ???fair???. I needed detailed, non-biased information on the SAN end user experience, and got it and more from NetWisdom.
Since I brought NetWisdom in (and I wrapped it around only one database at first, but now I TAP every storage port, and monitor all my databases) I???ve been able to use it???s metrics and data for much more than simple performance monitoring. I???ve optimized backup schedules, POCed new SAN gear, troubleshot (troubleshooted?) replication issues...the list goes on. I have other teams asking for read only access to the data, and for once, my network guys are jealous of my stuff. :) I only wish I had installed it earlier, so i had more long term trend-able data.
If you don???t mind the shameless plug, you can read about my work managing a SAN infrastructure at www.sandorasbox.com (Like Pandora???s Box, but for the SAN). I try and post little gems of experience that I???ve gained over the years in this industry, and ask probing questions on some PITA things I???ve run across. It???s new, but I???m posting fairly often.
Reply to this comment
Comment by RyanP on July 1, 2009 2:20 AM
Note to self, don't use single quotes in a post here...you get question marks. :(
Reply to this comment