Last week I had the pleasure to talk to Arsenio Mateos who works as a Virtualization SME and architect at Trend Micro. They have an interesting environment as you can imagine and offices around the world. Arsenio had mentioned that they are looking to use VSAN in the near future for
their VDI workloads as his experience with VSAN was great so far. That triggered a discussion and this article is a result of that.
Trend Micro is a large software development company with offices throughout the world. When we started talking about Virtual SAN my first question was around their use case. Was it deployed in their central datacenter to support their production / tier 1 apps or is Trend Micro using it for different purposes. Arsenio explained that currently they have rolled out Virtual SAN in 4 offices, which are “branch offices” in Tapei, Canada, Ireland and US.
In the “old world” these branch offices would require a complex infrastructure, Virtual SAN solves this for Trend Micro. The local staff can focus on what they need to focus on which are the applications and not the underlaying infrastructure. The staff is typically responsible for everything ranging from the physical network cables to the infrastructure, PABX phone system etc. So as you can imagine they need to know a lot about many different components, reducing the complexity is a life safer for them!
The experience so far with Virtual SAN for the branch offices has been so great that Arsenio is now developing an “office in a box” solution. This solution should be a “plug and play” type of solution which comes with all the services a new office needs (Router / Gateway / DNS / Active Directory / PABX etc). This should make it possible literally to get the IT part up in literally hours instead of weeks.
I asked Arsenio which version they are using today and how has experience has been with the previous and current version. Arsenio mentioned that he started with 5.5 during their PoC and had tested that extensively. When 6.0 was released he did a performance comparison and performance doubled! When it comes to performance he saw another aspect which had direct influence on the VSAN performance, which was the 10GbE switch used. Arsenio tested max read IOPS (4KB) with two different types of switches and consistently saw a 15% performance difference. Something I personally did not expect and will definitely start paying more attention to in the future.
Talking about testing, during the PoC, and later on during the vSphere6 try-out, Arsenio did many type of irregular things to the hosts running VSAN. Reboots, pull the cables, simulated network breakdowns, pull-out disks… No matter what Arsenio did, he could not break VSAN… He even moved all disks of a full DiskGroup from one host to another host, the DiskGroup became available on the new hosts right away and the VM data was available for use. That is one of the key reasons Arsenio felt very confident selling this solution to non-virtualization experts within the company.
All their sites run 6.0 as it stands today, besides the performance it is the operational aspect that Arsenio loves most about VSAN. Especially the Health Check Plugin stands out as it provides all the details you need to manage/monitor a VSAN environment, and he can’t wait for the Performance Plugin to be released as well. Hopefully that will be soon as the Observer does work but it is too complicated for local IT staff.
Last but not least we briefly discussed the hardware. Arsenio mentioned that they use 1u servers with single sockets with a single disk group, relatively low amount of memory and a 400GB SSD with 4 x 900GB 10K RPM SAS magnetic drives. There are additional disk slots available, so if needed it is easy to scale-up as well on each of the hosts. As they don’t have that many VMs (in the branch offices) running it allows them to cut down the licensing cost on all ends… Smart move if you ask me. Thanks Arsenio for the great conversation, awesome use case… Running VSAN in branch offices to avoid WAN hops for critical infrastructure services!