Hybrid cloud is a neat concept – but we need to be able to move data

I remember going to an HP(E) event back in 2011 (Discover Vegas, I think) when the idea of hybrid cloud and converged infrastructure was just getting going. HPE talked about the idea of “cloud bursting” – moving data and/or applications on demand into cloud infrastructure to cope with increased demand. At the time I was …

[Chris Harden]

RackSpace

It's probably worth mentioning OpenStack at this point, while AWS tends to be the go-to nowadays for cloud we have a rather neat solution with them.

We have physical database servers, connected to a disk array, with an ESX cluster hosting our site in their intensive hosting environment. We then burst a rather chunky data processing job out to the RackSpace Cloud and because it's all run in the same data center we get near LAN speeds connecting to it (plus it rather neatly cleans up any privacy issues someone could have not knowing where our data is geographically)

[Lusty]

Completely overlooking the obvious

This article assumes from start to finish that we're keeping the same application architecture. That's why you're having problems with "Hybrid cloud". It's because hybrid cloud doesn't exist, it's a marketing term. If you take a few steps back and work out what you're trying to achieve then hybrid cloud is simple.

Take, for instance, a retailer with a website. They have 4 visitors an hour for the whole year until black Friday when they have 30,000. It's entirely possible to host this internally and use "hybrid cloud" to precache results once a year in the cloud, or to spin up new web farms in the cloud. They could shard the site so that heavy lifting is done in the cloud and realtime data operations on prem.

Here's the clever bit though, and listen carefully because I'll only say this once...they could just architect for the cloud and not bother with hybrid. Since you're suggesting moving back and forth anyway we can assume that regs and security are fine with cloud, so just go cloud and reap the benefits.

Without exception, every vendor I have seen pushing hybrid cloud where data or VMs move more than once has no real cloud strategy. They are trying to delay the inevitable day when someone who actually understands this stuff re-architects the applications to work in a cloud friendly way. In many industries, the apps will never be re-architected. Not because they can't be, but because the incumbent vendor will be sidelined by a better one (Concur, O365, Salesforce, Xero, the list gets bigger by the day). Once that happens, it's too late and the old vendor is in Novel/VMware land with a great product whose time has passed.

[Wobble1]

Re: Completely overlooking the obvious

Hi Lusty,

Designing an app for the cloud can absolutely replace the requirement for running on-prem with outsourced IaaS. I'm a huge advocate of cloud native where its actually possible. However, operating multiple clouds is a real thing, not marketing - and I cringe when people think its that easy to change an entire business or line of business to adopt cloud in the way its meant to be used.

The problem is that your example uses probably one of the easiest possible workloads (ecommerce website) to make the point. Cloud is not the answer to everything and its businesses like AWS that forget this (even though i still believe they will release some type of on-prem version ala Azure Stack or Oracle Cloud on prem). Also, its not as easy as turning up to work and saying "lets get native". The process is long, arduous, requires strict governance and cultural change. Its easier for start ups or small to mid-business, but for large global enterprises with overlapping areas of concern its not.

Large enterprises are still barely even scrapping the barrel with cloud in relation to on-prem vs cloud ratio (for many reasons including the above) and this will be the case for years to come - this is the target market for vendors and this is why they are playing the hybrid card. There is still lots of money to be made in this arena.

Me

[Lusty]

Re: Completely overlooking the obvious

I think you may have read more than I wrote!

My point was that traditional applications don't need to move to the cloud, they certainly don't need to span between clouds and they definitely don't need to be split between on prem and cloud just because a vendor said cloud bursting was cool. My example was one of where cloud bursting is perfectly feasible right now with no changes. I also said that if you want to re-architect then cloud is great and yes you can use more than one cloud. That doesn't mean on prem and cloud though, if you're cloud architected then you've accepted the cloud so keeping some stuff on premise doesn't often add a lot to the scenario. I've done all of these scenarios in the real world, and crowbar hybrid is the absolute worst use-case where people just put bits arbitrarily in the public cloud.

[Anonymous Coward]

So it's a replication issue then?

Fair enough. Let me just Google that for you... hmm... ?q=rdbms+data+replication&year=1997...

[Anonymous Coward]

Low Barrier to entry concepts

Lets be real here, any vendor can insert a system on Equinix and use their Cloud exchange to provision compute on AWS, Azure or Softlayer.

Then create a SQL cluster across two providers and voila failover and "arbitrage" between loud providers without any data movement.

Not exactly science.

[Anonymous Coward]

Cloud Arbitrage

Speaking of cloud arbitrage, assuming I have 100s or TBs residing on some hyperscaler and can move data from one hyperscaler to another.

The question now becomes one of cost and time. Cost at least for a while will increase substantially because I will need to pay on both ends and in all likelyhood may offset any arbitrage advantages. It's also important to note, it would be expensive to use array based replication and maintain sync copies of data across hyperscalers. This practice would defeat the purpose to lower cost.

While I do believe the ability to move data is important I just don't believe in Netapp's specific use case.

[allthecoolshortnamesweretaken]

"The challenge in the future for storage administrators is not in managing the hardware; with the introduction of flash all those old issues of performance management are gone.

I guess that's what they said when transistors replaced vacuum tubes... performance management will always be an issue. Or will have issues.

[aheawood]

No mention of active-active replication. I'm surprised.

http://www.wandisco.com/product/amazon-s3-active-migrator