Infrastructure Systems Team Lead at a manufacturing company with 10,001+ employees
Real User
2021-12-22T19:08:00Z
Dec 22, 2021
The evergreen factor, which is the ability to upgrade in-place upgrades to new hardware could be improved. I would like to see the ease of deployment and built-in Metro clustering.
The replication of Dell EMC XtremIO could improve. In the newer versions they have improved, however, the replication can be improved further where we can include concurrent or cascaded methodologies. In the next release, the solution could have better integration and if we can host assets on the cloud, such as NetApp has the NetApp volumes, which we can host on the cloud directly called NetApp CVO (cloud volume ONTAP). Dell EMC should come up with something purely on the cloud rather than manage services.
G. Manager- Technical Services with 51-200 employees
Real User
2020-12-28T16:33:11Z
Dec 28, 2020
The cost of this solution could be reduced. Scalability is something that can be improved because there is an issue when it comes to mixing versions. We cannot mix version one and version two. This is something that may have been improved but earlier, it is something that was a challenge in terms of scalability.
Vice President of product at a tech services company with 51-200 employees
Real User
Top 10
2019-09-24T05:43:00Z
Sep 24, 2019
The management should be improved and the GUI interface could be better and easier. In the next release, they should improve the replication. There should be high availability. You can't do replication from one EMC to another, you would need to use another tool with the way it is now.
Manager of Customer Services with 1,001-5,000 employees
Real User
2019-09-23T06:33:00Z
Sep 23, 2019
The most important thing for the system engineer is to check if there is latency in the IOPS for any run. You cannot measure the number of IOPS or whether or not it is overloaded. You cannot measure anything in EMC about this. Most solutions, especially HP, improved our fall-over performance, with our database and servers. Most servers are HP, but we use EMC now only for backup. One thing that should be improved is the reporting and monitoring tools. It should use real-time monitoring for storage, IOPS, latency, etc.
Snr. Build Engineer at a computer software company with 1,001-5,000 employees
Real User
2019-09-15T16:44:00Z
Sep 15, 2019
It's not an improvement so much as a query. What I'm curious about is whether or not the feature has the ability to do inline data reduction between two physical storage solutions. If we have one storage solution in the UK and the other in Australia, will the system be able to tell us what we have already, so we aren't sending 95% of our data across the world unnecessarily? I know that the feature works locally on the source solution. What I'm curious about is, would that data reduction feature work between two instances of XtremIO in two different places.
IT Architect at a healthcare company with 5,001-10,000 employees
Real User
2019-09-15T16:44:00Z
Sep 15, 2019
What needs to improve is that the end compression uses a lot of CPU, which makes it difficult or impossible to upgrade. I believe that it has improved in the upgraded versions, which we don't have yet. Also, the price and interface should be simplified too.
Bring all-flash, scale-out storage to your enterprise applications with EMC XtremIO. Purpose-built for flash, XtremIO storage arrays are amazingly fast. Delivering high IOPS at less than 1 millisecond latency is just the start. EMC XtremIO helps you harness the power of flash storage by building in innovations like content-based data placement and dual-stage metadata.
Dell XtremIO needs to provide better performance to keep up with new products.
The evergreen factor, which is the ability to upgrade in-place upgrades to new hardware could be improved. I would like to see the ease of deployment and built-in Metro clustering.
The replication of Dell EMC XtremIO could improve. In the newer versions they have improved, however, the replication can be improved further where we can include concurrent or cascaded methodologies. In the next release, the solution could have better integration and if we can host assets on the cloud, such as NetApp has the NetApp volumes, which we can host on the cloud directly called NetApp CVO (cloud volume ONTAP). Dell EMC should come up with something purely on the cloud rather than manage services.
The cost of this solution could be reduced. Scalability is something that can be improved because there is an issue when it comes to mixing versions. We cannot mix version one and version two. This is something that may have been improved but earlier, it is something that was a challenge in terms of scalability.
The management should be improved and the GUI interface could be better and easier. In the next release, they should improve the replication. There should be high availability. You can't do replication from one EMC to another, you would need to use another tool with the way it is now.
The most important thing for the system engineer is to check if there is latency in the IOPS for any run. You cannot measure the number of IOPS or whether or not it is overloaded. You cannot measure anything in EMC about this. Most solutions, especially HP, improved our fall-over performance, with our database and servers. Most servers are HP, but we use EMC now only for backup. One thing that should be improved is the reporting and monitoring tools. It should use real-time monitoring for storage, IOPS, latency, etc.
The solution needs to be simplified. When you integrate your storage with other systems, it could use a little bit of automation.
The deployment of the solution could be simplified. The solution should be integrated into the system by default and not separately.
The product could be improved by reducing the pricing and having better organization in their technical support team.
It's not an improvement so much as a query. What I'm curious about is whether or not the feature has the ability to do inline data reduction between two physical storage solutions. If we have one storage solution in the UK and the other in Australia, will the system be able to tell us what we have already, so we aren't sending 95% of our data across the world unnecessarily? I know that the feature works locally on the source solution. What I'm curious about is, would that data reduction feature work between two instances of XtremIO in two different places.
What needs to improve is that the end compression uses a lot of CPU, which makes it difficult or impossible to upgrade. I believe that it has improved in the upgraded versions, which we don't have yet. Also, the price and interface should be simplified too.
The stability of the product needs improvement.