Software Defined Storage

In order to be fully buzzword compliant, I'm going to talk a bit around software defined storage.

In fairness, I think this should be looked at as more of a software defined persistence.  

I've spent some time running HP's Storevirtual VSA.  There is a LOT that is compelling about this platform.  I have been able to use their documented APIs and automate deploying a 8 node cluster with full storage auto tiering between SSDs and HDDs.  I can automatically deploy my ESX host with PXEboot, deploy the VSA cluster and license it, configure and present luns and format them as VMware datastores all as part of a scripted process and have a VMware cluster up and ready for VMs.

This whole process takes less than 1 hour to deploy an 8 node cluster from bare metal to fully redundant but the best part is that I really don't have to do anything manual other than kick off the VSA deployment script once the 8 nodes are finished PXEbooting.

I think the potential for this kind of technology is awesome.  At the same time, I don't want to limit this category to just iscsi luns, NAS, or traditional storage kind of technologies.  While SCSI is a protocol to access block storage, I feel like I could make an argument that your SQL database is just as much software defined storage as an iscsi block device, NFS datastores or an HDFS filesystem.  It just has a different access protocol and different services that are exposed and different service levels.   

What I like about the idea of software defined persistence is that my same physical cluster of servers with local disk that has a VSA, also runs Hadoop, and also has a Cassandra cluster.  

I didn't have to buy separate hardware for one versus the other.  I still have a capacity planning effort for each storage layer but the conversation has shifted from separate infrastructures that must be scaled and managed independently at purcase time to more of how do we divide the resources we have.  If things change, we can always readjust the ratios of resources. 

So in essence, software defined storage gives me storage flexibility.  The cost models are more linear as compared to large storage arrays.  I understand this scares some of the traditional storage vendors.  If it scares you as a vendor, that's probably a good thing.  It means you are paying attention.  I would just say be open to changing your business model because if you don't you run the risk of being disrupted completely.   As a consumer of IT, I have my own business priorities and if you can help me with those that is great.  If you start holding me back from accomplishing my business priorities because you are protecting your revenue stream then you move from someone helping me succeed to  someone getting in the way.  This is what opens you up to disruption more than being attacked on a pure technology front.

This neglected blog

So it has been a while since I have posted anything.  I actually have several things as drafts that I have never gotten around to finishing because I want them to be useful and meaningful (and contain enough information that I can use them instead of having to recall everything in my brain).

So this post is as much about getting into the habit of posting regularly as much as anything.

A few hints at what has been going on in the lab over the last year.

Software Defined Storage -  by this I mean software implemented storage.  I almost want to call this software defined persistence layer instead but that is not so buzz word compliant.  I've got several experiments here to share.  This has been fun.

OpenStack -  I'm not sure I'm ready to talk about this yet.  There are things I like, things I don't like and a fair amount I either don't have an opinion on or haven't really tested.

Big Data -  I've been playing with VMware's BDE.  I Found a few things I would consider bugs but overall it is working well and was very easy to setup.

Flash Cache - OK, so I have been playing around with using software in a VMware environment as a flash cache.  I've tested several vendors in this space all with differing results.  Overall this is a very interesting space to watch develop.

Hybrid Cloud - so last year involved two trips to Asia for me all centered around how to transform to a hybrid cloud kind of model.  It is really amazing what all I have learned in this area in terms of the reality of a hybrid cloud versus the marketing material.

So those are the things I intend to write about.  I am also hoping to finish some of the draft post which are much more about performance tuning and troubleshooting.