ModernStorage

IMHO -- winners are those that give customers storage that changes... more flexible. To be honest, doesn't matter if it's hardware-defined or software-defined... as long as storage keeps pace with changing apps, users, and data needs.

i think that's a healthy attitude Rob - solutions that work how we need them to vs. form factor per se.

are people doing more scale up or scale out today?

Scale up is what most people are used to. There's a shift to scale out happening in the industry.

Customers win when they don't have to make expensive exclusive choices. They shouldn't have to buy different solutions every time they need to access their storage in a different way.

. @pdvanhelden #opensource commodity is driving more scale out for sure but old school enterprise are used to scale up general purpose storage..

given we're on the front end of SDS - probably more scale up installed - but we hear plenty of interest and desire to scale out (and we'll see this phasing in - already are).

Oracle is all about engineered systems which is classic scale up how do you compare with that?

We also see a lot "modernization efforts" where companies are doing new architectures alongside their legacy... often with OpenStack and other environments. Here scale-out is preference from the beginning.

Also, scale-out and scale-up can be a red herring. OPEX is the silent killer in storage. Solving the human latency issue is the driver for most companies... the fact that it's scale-out is a bonus.

EMC Oracle IBM HP NetApp all big players trying to transform with software defined. How are they doing compared to you guys? How will you compete?

engineered systems meaning "pre-integrated"? We see that as folks buying the discipline they lack.

@rwhiteley0 That's a good point - issues revolving around human latency are some of the biggest complaints we hear when it comes to enterprise storage.

and buying simple is OK - we think you can get there with software that works with commodity servers (which are not that complex...!)

Management is a huge issue on scale out. How are you guys handling that piece of the value prop?

Engineered systems lack the flexibility and scalability required by modern apps.

Re: Competition. There's more than enough space for everyone. There's plenty of opportunity to target the next generation of applications where you need an elastic approach, as well as solve the provisioning nightmare of traditional apps.

Today we have our own management interface that tracks everything that is going on in the cluster. Single pane of glass to get to everything within the cluster.

These artifacts will of course get better and better over time.

Re: Management challenges... it's about making the software scale intelligently. Just roll in new racks of hardware and the cluster grows. If Amazon and Facebook have solved the problem at the scale they run, then why can't enterprises?

@rwhiteley0 Can you explain what you mean by "human latency?"

For eg. with our platform one could provision storage storage assets from your iPhone. Monitor real time metrics via iPhone. Disk repairs and the associated statistics can be observed from a mobile device.

We are not completely there but the building blocks do exist for us to build upon

Traditional RAID-based storage was very good at solving the problems of 10 years ago, limited disk capacity and limited cpu. Today's challenges are much more excessive capacity and scaling.

Scale-out is a means to make this manageable for human beings with minimal effort and pain.

@KennethHeal1 scale out is very good ROI too but the mgt layer has to be baked out otherwise hidden costs go through the roof. We have seen that with Hadoop

Re: Human latency.... that's the amount of time it takes for manual provisioning of infrastructure. Multiple trouble tickets... multiple consoles... most companies measure storage provisioning in days and weeks, when it should be minutes, secs

modernapps require #modernstorage

Will RAID work with the T-byte scale databases predicted for Big Data and particularly Internet-of-Things?

@pdvanhelden can you share some examples of modern apps?

btw @pdvanhelden copyright that phrase "Modern Apps for Modern Storage" - gr8 tagline

Apps for big data, analytics, every app that assumes no limit in underlying infrastructure

@BertLatamore this is where RAID as we know it can be a challenge - distributed systems (distributed RAID?) methods help ensure the type of protection RAID intends to provide, but in a different way. Rebuilds on a large cluster can actually happen quickly

@BertLatamore Traditional RAID as we know it is broken. What we have is a version of "Distributed RAID" which embraces the mantra "Bigger the cluster faster the repair"

Well, given my background in Distributed Systems I thought I had accumulated a unique set of skills to disrupt a big space.

Tell us how you came up with DynamoDB we use that on this product? Do you think AWS is doing well with it?

congrats @HedvigEng . tell us more about the similar features of Cassandra and dynamo that we expect to see in hedvig

Clarification - I didn't have anything to do with DynamoDB. I was part of the Dynamo team

@HedvigEng thanks for clarifying - what are the large scale challenges today - the hard problems that are most relevant to customers?

@mungamoori Well, Hedvig Storage Platform has nothing to do with either Dynamo or Cassandra since they are more like databases. Hedvig Storage Platform is more a storage system on top of which systems like Dynamo etc could be deployed.

Well that is loaded question. But I think Storage today is a large scale distributed systems problem. What we are fundamentally trying to do is provide complete storage protocol consolidation on one platform and make provisioning super simple.

When I say simplicity in provisioning one can push the envelope further and make it such that the platform lends itself to different workloads by change of policy and or media.

@HedvigEng Traditional enterprises consider moving to distributed storage as anti-pattern...how can they start that journey.

@HedvigEng When was @hedvig founded?

@arianagradow Founded in 2012 by Avinash Lakshman
Avinash worked on Amazon Dynamo and Apache Cassandra $12.5M in Series A funding

@mungamoori great question..certainly not a rip n replace but migration is a huge issue; where do they start ?? what's the playbook?

start in a tier 2 app - why not? i think that's a great entree for #SDS

that is great for @rwhiteley0 to address

I could give you a technical answers (granularity of provisioning, breadth of protocol support, hybrid cloud), but the real answer is flexibility. It's powered by our distributed systems approach.

for storage I would say to @arianagradow that Petabyte scale is table stakes today and zetta byte scale is coming fast

@hedviginc there is a conscious decision to provide the table stakes features - and combine with the best of a hyperscale distributed system.

@rwhiteley0 So do you aim to make the choice of storage vendors irrelevant?

The important difference at @HedvigInc is that we aren't wedded to old traditional ways of doing things and a scaled systems approach is very much in our DNA.

so think - combining things like dedupe, compression, snaps, clones, replication, etc. but in software and with resliency/scale

@pgillin No, not per se. We aim to make the storage platforms for different application needs irrelevant. We believe you should have one platform configured as necessary, rather than islands of storage. Object storage is the latest "island" for many.

.@pgillin good question.

@ercarter And all of those features (along with replication factor, if it should be flash-only, etc) can be dealt with on a granular basis, so you don't have to set them all cluster-wide for all your storage.