Archives for October 2013

Oct
25

Nimdesk Provides Choice – Linux, Windows, RDP or HDX or PCoIP

I first wrote about Nimdesk just about a year ago prior to joining Nutanix. Since that time Nimdesk has had some great growth including being the engine behind tuCloud, an Desktop-As-A-Service (DaaS) provider and having such customers as UCLA. I really like Nimdesk because like Nutanix, simplicity saves times, decreases deployment times and in the end saves $.

From a DaaS perspective it’s quite easy to have multiple AD domains in your roll out and even use local accounts for special use cases all being controlled through their broker.

And Domain users or local users.

And Domain users or local users.

There is even control for Admins only to have access to certain pools for administration as well.

Nimdesk is still using vSphere but the software works with the free version. When talking to Nimdesk they’re working on getting vSphere 5.5 certified which doesn’t have a maximum memory limit. With no requirement for vCenter the entry point is really low. Nimdesk provides a great platform for non-persistent. Nimdesk provides HA to non-persistent users in that there is no dependency implied on a per-Server basis that would, say, force a logged in user to be brokered to any particular VM on any particular Host.

In other words, when an end-user logs into our web portal, he or she has not yet been assigned to any particular VM in a non-persistent pool. It’s only the time when the end-user actually clicks on the pool name that the brokering action takes place. At this time, the broker identifies a “matching” VM from amongst all the Hosts in the grid/cluster that is in a state that’s ready for assignment to an end-user, and only then is the RDP (or HDX or PCoIP) connection brokered (and when appropriate – routed through the security gateway). There is an additional cost for using HDX or PCoIP but great to see there is the option to use them.

The other great implication here is they can make smart decisions about which Host to pick a VM in to broker to, based on host load averages and we can make those decisions in real time in a dynamic environment, thus removing the need for DRS. Yes I know DRS is more robust but seems like the next best thing.

Nimdesk can use both local storage and shared storage. Using local storage you can put your base images to go onto local SSD and the deltas to go on to spinning disk(or a different SSD volume if you have them). This part of the setup put the control in the hand of the user. If you mark a volume for images, it will self-replicate those Images to all nodes in the cluster. Nutanix can help to reduce the storage being used for the replication and provide auto-tering as most likely the application data folder for windows will end up spinning disk. This bad because most of the writes will end up here.

Nimdesk does offer ways to distribute the IO load over different volumes but it’s really only for windows. Since Nimdesk can broker Linux desktops, the afore mentioned issue can be fixed with Nutanix’s auto teiring. Companies that wan’t persistent desktops and especially Linux based desktops will benefit from not having to figure out data placement for their virtual desktops. At the end of the day the only way to fix Microsoft licensing is not use the product :-). With Nutanix having a lock less implementation of inline dedupe, we can provide a simple way of deploying applications for example to Linux desktops without breaking the bank from infrastructure costs.

It’s great to see Nimdesk flourish under the TuCloud brand. If you want to take it for a spin you can download a free version on sourceforge.net/.

There is also a good article on the background of TuCloud by Gabe Knuth

Oct
15

Nutanix: When Blunt Force Trauma Isn’t Enough on the vSwitch

My philosophy with most things is to keep it simple. The more bells and whistles turned and tooted usually lead to problems down the road. Each Nutanix node has over 20 Gb/s of available traffic which is a truck load in my mind. Living in the VDI land it’s common space to lots and lots of desktops running on one host. Multiple desktops, multiple vMotion streams during maintenance mode can take most of a 10 Gb link. If you want to ensure performance of your desktops you should look at using a virtual Distributed Switch(vDS). Lots of improvements have been made over the years with vDS and a great resource is the WhalNetwork.
[Read more…]