Storage Field Day 6 covered presentations from 8 vendors spread across 3 days. On day 3 it was Nimble Storage and NEC’s turn to tell us about their products. The previous two days were pretty heavy on easy to use (and shiny!) graphical user interfaces and centered around simplicity to provide storage to the application / app owner. Day three would continue on this theme with amazement and… a bit of a disappointment.
First presenter of the day was Nimble Storage. Nimble builds hybrid storage arrays which use the Nimble CASL (Cache Accelerated Sequential Layout) architecture: reads are cached on flash drives and writes are coalesced, compressed (in-line with a variable block size) and written sequentially to NL-SAS disks.
The systems can either scale up or scale-out. Scaling up is done by increasing controller performance (e.g. move from a CS300 to a CS500 or CS700); adding capacity is done by adding expansion shelves to the controller. Scaling out is simple: add more nodes/controllers to the cluster. The current qualified maximum configuration is 4 nodes which can deliver a total of 500k IOps and about 1PB of data. On the front-end the systems can currently only be connected to servers over iSCSI but FC support is promised to be added in the 2.2.0 release in the near future. And finally the data on the Nimble systems can be protected with either snapshots in the box itself or with asynchronous, snapshot based replication to a different Nimble storage array.
On a hardware and software perspective that sounds all good and well, with one possible caveat. I’ve blogged before that not all data access patterns are cache friendly. Think very random reads or infrequently used data. Flash is used as cache in a Nimble array and cache is finite and usually only a small percentage in your array. If your data is not in cache it will have to come from NL-SAS drives and those are slooooowwwww…… So while I can see the benefit of mixing NL-SAS with Flash in one array (combining ultra cheap + super fast), I’d like to POC this array for my workload to see if my applications would be nice and cache friendly and the users wouldn’t start complaining about sudden slowdowns if I accidentally hit NL-SAS.
Something where Nimble storage excels is their InfoSight cloud based management platform. InfoSight gathers statistics about the data patterns at roughly 94% of the Nimble customers; the rest of the customers opted out. It uses these statistics to proactively respond to failures. 80% of the total service requests are detected using Infosight and the majority of that percentage is fixed using the same VPN connection that InfoSight uses to send the data to Nimble HQ. Fun detail for us storage partners: the Nimble channel partners can view the data for the customers they are handling in the InfoSight portal as well. NEAT!
Another gem: Nimble engineering can blacklist upgrade paths on an array using the analytics in InfoSight. Say you have an underscore in your LUN names and some freaky bug in a new software release will generate problems with LUNs that contain underscores. Nimble Engineering runs a couple of queries, singles out the affected arrays and blocks the new software release. You, the customer, will see that a particular release is blocked and an alternative upgrade path. This results in rock solid upgrade performance and the current Nimble customers trust that enough to…. upgrade their storages in the middle of the day. Aaarghh! While I applaud the Nimble Q&A team for their stellar performance, I still wonder about the change proces at the customers. Risk & impact analysis anyone?
The last presentation of Storage Field Day 6 was given by NEC and spanned two products: the M-series storage arrays and the HydraStor deduplicated storage system. And for once I’m going to skip technical specs entirely.
NEC is everywhere; most of us have probably used one of their devices (e.g. a projector) at least once in our lives. It’s a big Japanese company with an annual revenue of $30 billion, a total of 101k people working for them and a R&D budget of $1.4 billion spread over 5 R&D centers worldwide. So definitely no start-up!
And it looks like none of that R&D money goes to the UI. Behold, the M-series GUI:
Appalling. I work with Unisphere Analyzer and that GUI is pretty ancient compared to some of the things we’ve seen during this weeks presentations. This is beyond ancient and gives me flashbacks to Windows 95. And the rest of the GUI for the M-series isn’t much better as you can see in the background with for example the alerts.
I have to admit that I didn’t even know NEC did anything with storage. I knew they made projectors and other appliances but that was about it. While the presenters were introducing the company to us and with NEC being the behemoth it is, I assumed (and I will admit my imagination was unchecked for a second) they would make storage for banks or oil rigs or factories. The kind of “install once, constant workload, keep alive for 5 years, replace and repeat” deployments.
It turns out the M-series is aimed at the SMB market. The market that usually doesn’t have a dedicated storage admin and that wants to spend as little time as possible on managing storage so that it can put out other fires. Most of the other companies presenting this week leverage cloud based analytics or intuitive GUIs to help this market spend as little time as possible on managing storage. Think dashboards that tell you in one instant what’s wrong with the system and what to do: EMC has been going in that direction for several years now with their VNXe range and all the Unisphere improvements. And the NEC M-series GUI is nothing like that. At all.
I do not doubt that the new M-series systems will do the job technically. You can put flash in them, auto tiering happens in relatively small chunks of 256MB (same as the EMC VNX2), there’s VAAI support, QoS… okay, cool, but that’s pretty much standard by now. Everyone does that. And in the meantime companies are differentiating themselves on manageability and simplicity. I really hope NEC uses a teeny, tiny bit of budget to revamp this GUI and bring it to the 21st century…
The HydraStor GUI was a lot better as Nigel pointed out during the HydraStor demo, but it’s still an enormous difference compared to the rest of the week, especially if you compare it to the Nimble GUI we viewed a few hours earlier…
My thoughts on day 3 and SFD6…
So that was it. The last SFD6 presentation. After lunch and saying goodbye to many of the delegates we headed back to the hotel. My own flight home was scheduled on Saturday afternoon so I still had an afternoon and evening to kill. I tagged along with Nigel and Stephen to Fry’s and several other electronics stores to hunt for a Nexus 10 for Nigel. Later that night Nigel, Tom and I watched Interstellar in the local IMAX after grabbing dinner at the nearby In-n-Out. Animal style double-double.. I like it!
The trip home was pretty uneventful! Get on the plane at SFO, chat with the lady next to me and watch a couple of movies, land in AMS, get on the plane to DUS, land and drive home. I’m actually enjoying all this travel, who would have thought!
SFD6 was a blast. Exciting, exhausting and inspiring! Thanks to GestaltIT for inviting me and my fellow SFD6 delegates for being just plain awesome and an inspiration to continue blogging and being curious about the entire storage market. And of course thanks to the vendors for telling us about all the new and cool things in and around their products. To be continued with some deeper dives into the various products that I found especially interesting during the Storage Field Day 6 week…
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.