And the new storage is…

I won’t play coy and string you along. I will just expound on my reasons in my choice for our new storage then give you the answer you have waited for.

Quick summary for new readers: I was looking for a solution to replace my old and cold (well, not very old but horribly performing Nexenta H/A cluster system). (see posts one, two, and three)

Tintri and Tegile where the two to compete against each other and both performed very well on my network.

Without 10Gbps ports I had to put some limits in place and the great Justin Lauer brought up a good point (I am paraphrasing from our discussions):

when systems have SSDs and you are running 1Gbps speeds you aren’t really capable of seeing the full potential.

He’s right but those are the limits I had with since I could not bond 2 ports together on the Tintri system (the T540 has 2 10Gbps ports in LACP configuration operating in active/standby – makes perfect sense to me). I use the lowest common denominator: 1Gbps ethernet speeds on a storage network.

And try as I might, the two systems did complete very well together saturating the 1Gbps ethernet with ease and both doing very well in I/O. Justin provided me with a Linux binary called ‘dperf‘ to test I/O loading, measurement, and creation of some icky data patterns on the storage appliances. Wow! It was fun to watch with both systems hitting over 20K IOPS. Tintri consistently hit higher IOPS than Tegile’s Zebi solution over NFS (not surprising). What was also fun was allowing it to generate terabytes of this data (takes a while over standard 1Gbps ethernet) and see what happens. It was the only way I could create some pain for both systems though this is really outside of normal operating parameters for either appliance. (suggestion to Tintri: release dperf to the world so that others can see how it all stacks up)

Then, recently, I got the notice that our Compellent SAN was coming up for maintenance renewal. Now, I am a fan of Compellent in many ways but the unit I was ‘left’ with was one that didn’t perform very well (at least not like I should have wanted). 16 750 GiB 7.2K RPM disks just don’t have the IOPS needed for our customers and the upgrade costs were way too much for me to afford. (the other units I help support for customers are excellent but they aren’t burdened with SATA disks slowing them down)

Stepping back almost 2 years – this is also what caused me to build the Nexenta solution I built in March, 2010 (deployed May, 2010) in the first place. I needed disk space for our VMware cluster and I needed more I/O than the Compellent could give. I had already been handed this system I never purchased myself that was in use to service customers over fiber-channel. The added load from VMware was just too much.

Back to today (well the last few months at least)…and it is time to end the suspense!

I need a system to replace the Nexenta H/A cluster because of performance reasons for our VMware NFS environments.

I have that maintenance payment coming up in the near future on a system that has never performed like I would have wanted it to and it is pretty damn expensive to run. I can’t just skip the maintenance plan because I have real, paying customers running on this gear.

Economics kick in.

Here I am with 2 very well performing systems from two smaller storage companies. Both companies have been great to deal with, excellent in their help and answering my questions, with products that fit my needs for our VMware hosting.

I can kill two issues with one appliance.

Only one of them also handles fiber-channel (and iSCSI, if needed): Tegile’s Zebi Appliance

Pricing between them was very compatible ($2,000 difference between them), maintenance contracts were within 5% of each other year over year, interactions have been positive. And I can support some new up and coming companies.

Performance (you have looked at those pages, right?) is not only comparable but excellent in both cases.

My thanks to both companies for allowing me this time to play, test, and experiment with their flagship storage appliances. (though not as much as I would have wanted but work comes first). Huge love for both appliances and the people I met along the way.

When you are out looking for storage then please, look at these two companies.

Excellent choices for VMware (the main part of my search).

Working compression and deduplication without hassles or big performance hits.


Easy to use web interfaces for end users to configure and handle day-to-day checking.

Working statistics via web interfaces that gives you data you want to know (poke: both need external interfaces as well to tie into enterprise monitoring and measurement systems!).

Thanks for reading and come back to see how things progress as I continue to post updates from time to time.

People I worked with:

Justin Lauer (SE), Mark Gritter (co-founder, architect)
Justin Cheen (co-founder, sales), Rajesh Nair (co-founder, CTO), Alok Agrawal (co-founder, SE)