Storage Field Day 6, Day 1! The day before was spent acclimatizing to the 9 hours of time travel but on Wednesday it was finally about to begin. I had a general idea of what was waiting for me: first of all it’s going to be an exchausting, challenging but also fun week, and also: 99% of Americans likes stroopwafels. So armed with a bag of them I headed to the SFD6 breakfast briefing. Stephen laid out the ground rules, explained the newbies what was about to happen and off we went to the first vendors of the week: Avere, StorMagic and Tegile.

Avere Systems

Avere Systems has a crystal clear mission statement: creating hybrid cloud NAS solutions that integrate public and private object storage with on premise NAS systems into a cost effective pool and with scalable performance for users everywhere. Originally they started shipping FXT Edge filer systems to smooth out performance bursts on on-premise NAS systems:  adding a layer between the client and the NAS system and caching I/O. If you add a global namespace and the ability to move data into the cloud this enables customers to gracefully move (part) of their data into the cloud and back again. For the end-user nothing will change: their data will come either from (cheap) cloud object storage or (faster) on premise NAS storage.

The kicker is that compute is much more expensive than (archive) storage. Hence if your compute demands change substantially over time (i.e. you need temporary resources for batch jobs) you might not want to invest in all those resources and then not use them half of the time. In this case the cloud will be ideal: rent cloud compute resources when you need them and use them to crunch data stored on your own NAS systems. Or the other way around: store data in the cloud and crunch the data with spare compute power on your own site (e.g. during the night).

The Avere vFXT solution runs in the cloud and creates a Hybrid Cloud: local and cloud compute and storage are connected with each other and can be used in a crisscross manner. This enables you for example to burst into the cloud when you temporarily need additional cloud compute resources that still need to access local storage, or move data to the cloud for long term, cheap storage but which still needs to be accessible by local computing systems.

In the Netherlands cloud adoption is still pretty low. In some use cases the Avere solutions might make it a bit easier for customers to start moving some jobs or data to the cloud. Check out what other #SFD6 delegates thought about the Avere solutions: Eric Shanks, Dan Frith, or watch the recorded presentations.

StorMagic

We all know the typical customers that have lots of branch offices but no IT crew on-site. Think customers in the oil and gas industry or retail: a sales office has some equipment in a closet somewhere in the building but no on-site IT staff. You still want to virtualize your servers to gain all the advantages in hardware consolidation and resilience, but installing a storage system is way too expensive. In fact: installing one storage system on-site doesn’t make sense since that will become a single point of failure, so you’ll have to install two of them.

StorMagic claims to currently have 1226 customers worldwide with over 30.000 licenses and a grand total of 57PB of storage. One of their customers (a major US retail chain) even spans more than 2200 sites! The average site usually needs about 2TB of capacity, does not require much performance (several hundred IOps) and doesn’t have a proper equipment room. So to host the (on average) 7-8 apps they want a shared storage solution that they can “set and forget”. SvSAN does this by leveraging the disk capacity in the virtualization hosts (Hyper-V or VMware, any commodity hardware will do) and present this as a virtual iSCSI disk. The big advantage of the SvSAN solution is that it only needs two virtualization hosts instead of the mandatory three hosts for the VMware vSAN solution. This saves you 50% on CAPEX and OPEX. Add an incredibly simple licensing model based on capacity that roughly equates to $500 per TB (an unlimited capacity license for $10.000) and this might be a perfect solution if you have small branch offices.

On the technical side data is synchronously mirrored between the pairs of virtualization hosts and can instantaneously failover in case a host decides to go up in flames. Compared to VMware vSAN, StorMagic SvSAN offers you some more control over which pairs of servers replicate the data: useful if you want to split your servers over two rooms in your branch office. On a monitoring side, performance statistics are logged every minute and kept for 24 hours. For another month you can view the stats in an hourly granularity, after which everything is rolled up into daily stats which are kept for a year. This should be sufficient to determine whether there is a storage bottleneck, although there were some requests from other delegates to reduce the granularity of real-time monitoring to less than 1 minute.

The fact the German Army/Bundeswehr chose SvMagic for their in-field deployments should be a good indicator that it is a military grade solution! Watch the presentations or read the blog from John Obeto II or Dan Frith for additional information.

Tegile

Tegile (pronounced Tai-gile, the name coming from “technology and agility”) produces hybrid flash and all flash arrays. They support a wide variety of applications across pretty much all the vertical markets. The systems offer both block and file storage, all carved out of the same pool of storage for which you can dial the performance and capacity up and down: more performance, add more flash.

One of the problems with (all-)flash arrays is the garbage collect that reclaims capacity on the flash portion of the array. If your applications are squeezing the absolute maximum out of your all-flash array (AFA), this garbage collect will have a noticeable impact on the application response time.  Tegile claimed that by segmenting their flash capacity in metadata and a high- and low I/O range they’ve reduced the impact of this garbage collect. Other vendors (like EMC² with their XtremIO or the PureStorage systems) just simply throw out the garbage collect and handle data placement completely different. A Proof of Concept with the Tegile systems will have to point out if their approach actually works.

One of the major discussions during the presentation and especially afterwards on Twitter was the fact that the Tegile arrays first compress data and afterwards dedupe it. This is the complete opposite of the COHO Data and PureStorage approaches which first dedupe and then compress. To be honest I was very skeptical myself since in pretty much every deduplicating back-up solution they advise against compressing back-ups before sending it to the back-up software. The reason Tegile chose to first compress and then dedupe seems to be a design approach of ZFS (on which the Tegile arrays are based) and, according to the Twitters, shouldn’t have too much impact on the eventual space savings and CPU overhead.

@shaun_gee @vStewed @robcommins @nigelpoulton @JonKlaus @Tegile don't see why, with LZ4 compress cycles are minimal and stupidly efficient

— Craig Morgan (@craigmorgan) November 8, 2014

John wrote a nice summary about the Tegile solutions over here, Scott wrote a book and the videos are live as well.

My thoughts on SFD6 Day 1

I didn’t know about any of the vendors that we visited on day 1. Of course I did my homework, visited their websites and read about their solutions but it’s shocking how far off first impressions can be compared to post-presentation impressions. The StorMagic product for example didn’t really make me swoon after the first website crawl (“oh, just another VSAN product!”). The presentation itself though really turned that around: the company has a crystal clear vision, and it’s hard to argue with such a simple pricing model.

From an IT trend perspective day 1 was centered round simplicity. Simplicity to exploit the cloud (Avere), simplicity to run a bare necessities, low cost virtual shared storage solution (StorMagic) and simplicity to provide flash performance to applications (Tegile). This has been the theme for some years now and was bound to return for the remaining days during Storage Field Days 6.

I was very impressed by the SFD delegates: incredibly knowledgeable people from across the world and with various backgrounds. The GestaltIT crew (Claire, Tom and Stephen) also did a kick-ass job in making sure everything was taken care of: hotels, dinners, transportation between vendors (in an awesome limo; the things we discussed in there I will not disclose ;))… everything was top notch. Later in the evening we watched a movie at a rented, private SFD6-only cinema: Twelve Monkeys w/ Brad Pitt and Bruce Willis. Weren’t there twelve SFD6 delegates… hmmm, I’m curious if there is a relation… 😉 Great movie though and I certainly enjoyed seeing it again!

Overall a great first day! Lets continue with day 2…

Disclaimer: Even though GestaltIT paid for the flight, hotel and various other expenses, I was not compensated for my time and there is no requirement to blog or tweet about any of the presentations. Everything I post is of my own accord.