Recently EMC announced that they were upgrading the storage software that drives their XtremIO to be able to compress and deduplicate data. This is an important addition, since not all data benefits from deduplication, databases being a good example. But often environments that don’t benefit from deduplication do derive a benefit from compression. As a result, having both is very beneficial. The downside to this announcement was that current XtremIO customers would have to experience a disruptive upgrade.
By disruptive, we mean very disruptive. Data will have to be completely backed up, restored to another system, and the former system would need to have an upgrade. Considering that XtremIO is a scale-out architecture, there is a natural assumption that upgrades should be non-disruptive. When the news broke of this disruptive upgrade, the XtremIO competitors jumped all over it.
To be fair, a disruptive upgrade is a big deal and the effort that an XtremIO customer has to expend to install this upgrade is not insignificant, but I do think we are seeing a bit of a piling on in the industry as there have been a high number of vendor attack blogs about the situation. This reminds me of the saying, “Those who live in glass houses shouldn’t throw stones”. Instead of jumping on the pile along with everyone else, we went to EMC to get a real understanding of what was going on.
First, we found EMC to be very upfront that they knew that this was not an ideal situation but felt is was necessary to get the important features in the latest software update to their customers. They could have played a marketing game and came out with a whole new “generation” of products instead of enabling current products to be upgraded. But EMC chose to give legacy XtremIO customers the option to leverage the capabilities of the new software.
It is also important to point out that this was not an emergency upgrade. Current XtremIO customers could implement this upgrade at a time of their choosing, the current version of the software works just fine (except it can’t compress data) and EMC is committed to supporting the older version of the software for the foreseeable future. There is an important lesson here, whether you are an EMC customer or not. Highly redundant systems are still vulnerable, it could be a planned event like this upgrade, but we have seen software bugs creep into an upgrade and corrupt all stored data. This is simply another reminder of the importance of backups.
Second, when the customer does need to upgrade, EMC has stated that they will be on hand to help out as well as providing a loaner system to transfer production to while the customer’s system is being upgraded. Of course this could still mean an outage of some type when the cut over occurs, but with proper planning (and I would expect the EMC engineers to be able to provide that planning) the outage should be relatively short.
Third, this upgrade in software only impacts legacy XtremIO customers. New customers should be starting with the new code. The bigger question: is this a sign of things to come? Will EMC customers have to face a disruptive upgrade in the future? EMC assures us that this was a one time occurrence. The reality is that nobody knows for sure. But that is not unique to EMC, all vendors of storage systems may face some sort of disruptive upgrade in the future as new technology or ways of managing data come to market.
The purpose of this column is not to defend EMC but to add a little balance to the discussion. Proposing an upgrade process like this is never a fun experience. Maybe EMC should have seen compression being a requirement from day one, maybe not. The fact is that customers bought it without the capability and if they now want it, on their current hardware, they need to go through this process. But it seems to us that EMC has taken appropriate measures to make that process go as smoothly as it can under the circumstances. Should EMC have some “points” taken away? Sure, but keep in mind that all vendors may have to face a similar situation in the future and it may not be a schedulable event. For this reason it is critical to make sure you’re All-Flash array is backed up as any other storage system in the environment.
The architecture of an All-Flash array is critical to it becoming a strategic component of the storage infrastructure. Read Storage Switzerland’s latest report “Understanding All-Flash Architectures”.