Successful examples of VMwareVirtual Volumes (VVOLs) deployment were unheard of in France, but perfume and cosmetics chain Beauty Success put them at the centre of their virtualisation storage upgrade.
Using VVOLs and Pure Storage it has gained significant performance increases – up to 3x, and dodged potentially fatal bottlenecks – by moving virtual machine storage away from physical array volumes.
VVOLs was launched in 2015 by VMware to address virtual machine storage issues. But it has not been particularly prominent as it contrasts starkly with the ways most array makers do things. “We were at an impasse,” said Edouard Jugie, technical lead at Beauty Success, which has more than 300 outlets across France connected to core applications by VPN at its Dordogne datacentre.
“According to established storage practice, we had divided our disk array into several volumes of 1.5TB, which was about 20 volumes between 140 VMs. The problem with this kind of schema is that you run the risk that a VM will saturate the volume it is using to the point it becomes unusable for all the others that access it. When that happens no more applications from the same set can write data.”
“Avoiding failures was becoming a nightmare. We had to keep a constant eye on performance and at the least sign of danger we would move a VM to a volume with more space and more bandwidth” said Jugie. “These are the tools – business intelligence applications, shopping portals, another to track orders, an internal logistics messaging app – that run on the VMs. They’re all as critical as each other and maintained by just nine IT staff.”
“Our applications are virtualised with VMware ESXi, which has functionality that can address our challenge: VVOLs. With VVOLs it’s the virtualisation system alone that optimises storage resources to the VMs and which, above all, avoids any VM saturating the storage of others.”
The problem was, however, that in 2019, nearly four years after VVOLs had launched, no enterprise in France had borne witness to the system functioning.
Pure Storage: More modern, more “sexy”
At the end of 2019, the Beauty Success was forced to make changes because its servers and storage had come to end-of-life. Since 2010 the company had regularly bought its storage from NetApp.
“Beyond VVOLs, our priority was to move to 100% SSD because that’s the only type that wouldn’t create a bottleneck on our arrays.”
A request for proposals was publicised and NetApp responded as usual. Pure Storage did too. Huawei responded too late.
“Pure Storage’s offer appealed to us, because they had a modern approach. Firstly, in commercial terms, their Evergreen programme allowed us to evolve the storage array as we went along. It’s possible to replace the SSDs with more powerful modules while leaving the controller in place and vice versa.”
“Also, at the level of functionality, Pure’s admin console is very sexy. For example, it shows statistics for bottlenecks that allow performance problems to be anticipated.”
NetApp promised equivalent functionality would soon be delivered and both vendors were committed to supporting VVOLs.
“A year later and NetApp still haven’t brought out the admin console they promised. So, I’m very happy to have taken the decision to opt for Pure Storage,” said Jugie.
An array delivered, and pre-optimised
Deployment of a FlashArray//X20 took place in January 2020, timed for after the peak of the retail year. It comprised dual controllers, disk shelves half-filled and providing 20TB of capacity.
“In total our virtual machines were 24TB but we played with the data reduction functions – deduplication and compression – which allowed us to store 4.6TB in 1TB of real capacity. Currently, less than 6TB of capacity is taken up,” said Jugie.
Connectivity is via 8Gbps Fibre Channel to servers. “We have kept the infrastructure that we used with NetApp. It wasn’t the fabric that gave us problems so replacing it with better-performing Fibre Channel can wait.”
Beauty Success’s integrator partner Solution Data helped Jugie and his team with the migration. “Frankly, the deployment was less complex than previous occasions. Pure Storage delivered an array pretty much ready-to-use with optimal settings already in place. We hardly had to touch it. With NetApp our experience was that we had to set things up ourselves.”
Migration of data from the old array to the new one took just a week. VMs were transferred in batches of 10 using VMware vMotion.
VVOLs do away with manual maintenance
Until that point, each physical volume of 1.5TB had looked after three VM files: its configuration, its memory image and its disk image. From here onwards, each of these would be stored on a virtual volume.
“That resulted in a large amount of very important virtual volumes,” said Jugie. “But that’s not a problem because the ESXi hypervisor manages it all in a transparent way with the disk array, including dynamic sizing of virtual volumes according to their needs.”
Read more about VMware and storage
- How resource contention affects VM storage performance. VM performance issues often occur when resource contention is present. Admins can check the number of IOPS requests and storage array connections to address any bottlenecks.
- Manage large volumes of VM data with these storage tactics. Admins can effectively virtualise large servers with storage tiering and disk allocation, but they must be wary of potential IOPS and redundancy issues.
Jugie is especially keen to underline the ease of use that his team have gained. “Until now, to reclaim space, I needed to run commands on each of my 20 volumes of 1.5TB. Now there are no manual operations: it’s all automated.”
One area that has needed a tweak, however, has been in backup operations. “We use Veeam Backup, which has plugins that automatically reclaim snapshots created by disk arrays, whether NetApp or Pure Storage. But that doesn’t function since we deployed VVOLs. We had to reconfigure Veeam Backup so that it now interfaces directly with ESXi.”
Backups now run as smoothly as previously. “Since the deployment we’ve not known any snags. Backups are even twice as quick because we’ve modernised the NAS that looks after them, an ExaGrid. That’s the proof that we don’t have an urgent need to modernise our Fibre Channel infrastructure and that we have good reason to think VVOLs poses us no problem at all.”
Proactive maintenance from the supplier
Regarding performance, Jugie’s team have measured a 3x improvement compared to before the VVOLs/Pure deployment,” he said. “Notably in the running of batch processes in business intelligence and logistics applications as a result of the presence of flash storage. “In practice, that means that our teams get orders quicker than previously and that’s a real business benefit.”
As he had done with NetApp, Jugie contracted for 24/7 maintenance with Pure Storage.
“I noted with some happiness that Pure Storage does proactive maintenance. They contact us when they discover a problem in their firmware and tell us they’ll be providing an update. It’s something that never happened with NetApp.”
“In fact, for updates we had to put in a request to our integrator who passed that on to NetApp. We had to plan the operation, which wasn’t very practical and we could only consent to it once a year,” explained Jugie, of a process that ended up looking more like it obeyed the needs of the calendar rather than technical imperatives.
“The thing to understand is that in an organisation our size we have only one storage array,” he said. “So, we can’t take the risk of losing it while trying to install an update ourselves. What Pure Storage offers is much more adapted to our needs.”