Sign in to follow this  
Brian

The EMC XtremIO 3.0 Disruptive Update Debate - Customers Don't Car

Recommended Posts

=== Disclaimer: Pure Storage Employee ===

Brian, Kevin & StorageReviews Team,

Your reporting that customer don't care stands in stark contrast to what we are experiencing. XtremIO customers have been surprised at the numerous situations which require the data to be evicted and restored after the update. Many are referring to this process as a 'destructive upgrade'. These scenarios include:

Adding an Xbrick to increase performance

Adding an Xbirck to increase capacity

Upgrading to XIOS 2.4 (encryption at rest)

Upgrading to XIOS 3.0 (data compression)

Surely in analyzing XtremIO, StroaegReviews is aware of these requirements.

Most customers do not have spare flash capacity on hand to use as 'swing space' nor do they want their staff having to plan, prepare, and work off hour to complete a migration. Even with for hire professional services there's always risk of outages due to human error.

Its fair to say that most customers incorrectly believed that a scale-out storage array would never require downtime for HW expansions or software updates and suggesting that they are OK with the reality that a scale-out cluster does not deliver this value seems like a stretch.

While EMC has stated future XIOS upgrades will non-disruptive, they have made the same commitment when increasing the number of nodes in an XtremIO cluster. Again, this process is destructive to the existing data.

While the unexpected will always occur, customers deserver more than to be promised NDU from a scale-out architecture that has delivered neither and within 12 months of being GA'd required a major architectural overhaul.

I am a competitor, so colour my comments as such, but the first year of XtremIO has raised more concern than confidence. Let's revisit this topic a year from now. it'll provide a better perspective.

Share this post


Link to post
Share on other sites

=== Disclaimer, Im not a Storage Vendor Employee, but i do own several vendors equipment === ;-)

Now i think most people can agree on that Distruptive Upgrades are a unwanted risk to a platform in general. And to me it would be close to unacceptable, even though im not running a "Enterprise Senario" (we only have 1200 employees, in EMC talk that would be SMB).

Im not exactly sure why a disruptive upgrade would be needed, but from what ive read it has something to do with changes to the Metadata in the array. I am not sure why it cant handle more than one metadata type, and just do internal migration of the data to a new Set but im sure there are technical reasons as to why its not possible. EMC should look into that, and implement it so future updates would not require them to be in the same situation. That ofcause depends on their architecture beeing in a state that allows for this.

If i were an EMC Customer, i would assume that my Storage Partner or EMC directly would provide me with some sort of Migration Path in form of either a physical array, so the data could be replicated, and the unit either be replaced or upgraded without downtime on my Production enviroment. I assume that is what EMC means by their press release/statement. If not, that is clearly unacceptable. Relying on downtime and/or risk of dataloss incase of inadiquate backup is not in any form acceptable. Even in VDI solutions.

My newest Storage Array (a 3Par 7200), will be getting inline deduplication soon (flash only unfortunaely), and i doubt HP would allow the upgrade to be disruptive. It simply isnt "Enterprise", to do so.

Regards

Christian

Edited by Darking

Share this post


Link to post
Share on other sites

=== Disclaimer: Pure Storage Employee ===

Brian, Kevin & StorageReviews Team,

Your reporting that customer don't care stands in stark contrast to what we are experiencing. XtremIO customers have been surprised at the numerous situations which require the data to be evicted and restored after the update. Many are referring to this process as a 'destructive upgrade'. These scenarios include:

Adding an Xbrick to increase performance

Adding an Xbirck to increase capacity

Upgrading to XIOS 2.4 (encryption at rest)

Upgrading to XIOS 3.0 (data compression)

Surely in analyzing XtremIO, StroaegReviews is aware of these requirements.

Most customers do not have spare flash capacity on hand to use as 'swing space' nor do they want their staff having to plan, prepare, and work off hour to complete a migration. Even with for hire professional services there's always risk of outages due to human error.

Its fair to say that most customers incorrectly believed that a scale-out storage array would never require downtime for HW expansions or software updates and suggesting that they are OK with the reality that a scale-out cluster does not deliver this value seems like a stretch.

While EMC has stated future XIOS upgrades will non-disruptive, they have made the same commitment when increasing the number of nodes in an XtremIO cluster. Again, this process is destructive to the existing data.

While the unexpected will always occur, customers deserver more than to be promised NDU from a scale-out architecture that has delivered neither and within 12 months of being GA'd required a major architectural overhaul.

I am a competitor, so colour my comments as such, but the first year of XtremIO has raised more concern than confidence. Let's revisit this topic a year from now. it'll provide a better perspective.

No doubt that you're hearing customers vent frustration to you, there is a lot of that going around. But the bulk of the people we talked to said it's not that big of a deal and while they're frustrated, they're happy that EMC is providing swing hardware and services to help the migration along. That mitigates the concern over spare internal swing hardware. Some customers are also moving their VMs to something else, VNX say, then storage vMotioning back after the upgrade. It doesn't seem to take too long to run the upgrade.

Appreciate your contribution to the conversation.

In full disclosure we have not reviewed XtremIO or Pure Storage, so we're not making any claims for or against either.

=== Disclaimer, Im not a Storage Vendor Employee, but i do own several vendors equipment === ;-)

Now i think most people can agree on that Distruptive Upgrades are a unwanted risk to a platform in general. And to me it would be close to unacceptable, even though im not running a "Enterprise Senario" (we only have 1200 employees, in EMC talk that would be SMB).

Im not exactly sure why a disruptive upgrade would be needed, but from what ive read it has something to do with changes to the Metadata in the array. I am not sure why it cant handle more than one metadata type, and just do internal migration of the data to a new Set but im sure there are technical reasons as to why its not possible. EMC should look into that, and implement it so future updates would not require them to be in the same situation. That ofcause depends on their architecture beeing in a state that allows for this.

If i were an EMC Customer, i would assume that my Storage Partner or EMC directly would provide me with some sort of Migration Path in form of either a physical array, so the data could be replicated, and the unit either be replaced or upgraded without downtime on my Production enviroment. I assume that is what EMC means by their press release/statement. If not, that is clearly unacceptable. Relying on downtime and/or risk of dataloss incase of inadiquate backup is not in any form acceptable. Even in VDI solutions.

My newest Storage Array (a 3Par 7200), will be getting inline deduplication soon (flash only unfortunaely), and i doubt HP would allow the upgrade to be disruptive. It simply isnt "Enterprise", to do so.

Regards

Christian

You're correct that EMC under-delivered here with the DU and customers are indeed unhappy with the need to move data off their arrays. As stated though, EMC is providing hardware and services to ensure there's no down time. Will any of these migrations get messed up? I don't know...maybe, there's a law of averages in there somewhere that says something will go wrong. But the mass of their customers will likely be fine. I do wonder if EMC had architected this from the ground up instead of buying the technology if things would have been any different.

Share this post


Link to post
Share on other sites

Brian

Did you talk to any customers to form the opinion that customers don't care? Were any of those customers running mission critical workload or more "transient" stuff like VDI? Of course I'm sure you're aware that EMC puts tight controls in contracts to prevent customers from discussing details on their hardware. How many customers do you feel were unwilling to speak out about this?

The key thing here is the principle of the way EMC marketed this platform. They clearly describe XtremIO as non-disruptive and it's not. Whether customers care, is only part of the story; prospective customers watching online content could feel this platform is non-disruptive and get a nasty shock in the future. It's all about honesty.

Share this post


Link to post
Share on other sites

In this case I think EMC is trying to be more transparent because they made a mistake in the marketing materials and that's what they're getting hit on the most. I think they're doing well to take their medicine and move on. Customers are indeed bothered by it, but it's not as big of a deal as enthusiasts and competitors would have us think. People we talked to were using VDI as you say, but also DB and patient care related apps.

As to what happens in the future, it's right to be cautious about the platform. I could have posted more comments about people feeling at first a bit upset, but then being thankful they bought EMC, who has the staff and product to repair the issue...but that wasn't really the point. If EMC were to have another destructive update within the next couple years, I think that would be devastating for them. The XtremIO next Gen will be a new hardware platform probably, more inline with VNX2, etc.

Share this post


Link to post
Share on other sites

Full Disclosure here.... Pure Storage Employee

"The XtremIO next Gen will be a new hardware platform probably, more inline with VNX2, etc"

This particular statement lends itself to the need to ask what will the introduction of that NextGen hardware look like. When those NextGen controllers and shelves of capacity are brought to market will they be able to be introduced into an existing cluster, allow a customer to rotate in new hardware while taking out the old, all while data remains in-place and available to hosts? NDU is MUCH more than just about revisions to the software in today's data centers and more about a complete non-disruptive approach to the entire lifecycle management of both the software AND the hardware. If the introduction of NextGen controllers and denser shelf capacities cannot be cycled through an existing cluster then the platform has done absolutely nothing to shift the paradigm in data center operations and the three year forklift migration cycle will forever remain.

Today it is understood that even though XtremIO is a scale-out "architecture" it is indeed a "fixed" one, that meaning one cannot add new bricks to an existing cluster and this once again is a disruptive operation requiring swing gear. It is most definitely fair to say that at some point EMC will harden this aspect of the platform to make cluster expansion online but in the meantime customers are stuck with the "size it right on first pass or forklift it at a later date". Are customers enjoying the benefits of free swing gear and services in these particular cases?

While there are more than likely a great number of happy XtremIO customers that despite the destructive nature of this current disruptive upgrade are pleased with the performance and functionality of the platform for their current use case it would be interesting to understand if these same customer are betting the farm on XtremIO becoming their standard platform for Tier 1 Performance oriented workloads and the foundation of an all-flash storage strategy to replace all spinning disk in the data center.

How many of those happy XtremIO customers are also saying right now they want to replace every one of their VNXs or VMAXs they have on the floor today with XtremIO? Maybe all of them, maybe none of them. The customers that are happy and accepting of the disruption that is about to come their way, would they be saying the same thing if they had a dozen or two dozen clusters on the floor and had to manage the double migration across that entire footprint, free services and swing gear or not? When NextGen Controllers and denser shelves become available will the customer be able to enjoy the benefits of a non-disruptive path to much more efficient TB per rack unit without having to do a rip-and-replace?

There is much more at play here than just a software upgrade to get some new features.

Edited by robertway

Share this post


Link to post
Share on other sites

My comment was more speculation about what could cause a fundamental change at some point in the future, that's not from EMC. Dell is going to have a similar problem as they merge EqualLogic and Compellent, eventually today's gear won't play with the next gen stuff, whatever that looks like. I'm sure Dell will try to mitigate as best as possible, but it does happen.

As to what customers expect to do with their plans, you have the same flash adoption numbers as everyone else...flash is growing, though not at the pace I'm sure you'd like ;) It's my opinion that eventually we end up with flash, large fat slow disk, and gasp...tape. Could be wrong, but that's my lean today. The enterprise isn't always quick to adjust however.

Share this post


Link to post
Share on other sites

Brian, thanks for the clarification, my comment was less directed at you but instead to the general question that may be on customer minds when mention of NextGen hardware becoming available and "how do I get there from here?" and is my current platform/investment a vehicle that allows me to take advantage of it.

Thanks again for the clarification....

Share this post


Link to post
Share on other sites

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now
Sign in to follow this