This product speeds up VM IO speeds dramatically. It doesn't work with ESX yet.
My question is what could they be doing that would be able to generate such large gains in IO (up to 30x)? It talks about caching? I would like to know how this would work with a VM load that I would really be concerned with regarding IO: Databases?
I haven't seen the web site but if that is how it works than it's not a technology specific to vm's ...... it would be a generic IO speeder by means of RAM caching .....
My pure speculation ......
DXtreme doesn't actually work the way a cache does. It works by predicting IOs before they happen, not storing IOs and waiting for them to reoccur. It also isn't a read ahead based cache, but that is probably the easiest way to think about it. DXtreme gains performance from the prediction works even across databases with entirely random read/write IOs occuring. This works because the prediction engine usually exceeds a 90% accuracy. The prediction engine operates in and with a pool space, which does live in RAM.
I hope this helps.
CTO - InovaWave
Helps, but I'd really like to see some numbers of what it can do for Exchange 2003 IOs.
I am trying to figure out if this is a revolutionary or just a niche product that speeds up certain types of IOs.
You will see big improvements on Exchange 2003. I've run a few of MS's Exchange SIM tests, the old Jetstress as well as the new. We aren't VM or application specific, but independent.
If you want, I can work with you on running some tests with our Enterprise or Datacenter Products.
Thanks for the reply.
I will keep my eye on your product as it is definitely interesting.
When do you plan on having the ESX version out? Also, where will DXTreme get installed for ESX, on the ESX host or the VMs? Will it be certified by VMWare?
DXtreme has a small component that installs through the console (it is tiny) and there is nothing installed inside of the VMs. DXtreme is VM agnostic, so Windows 2K3, 2K, XP, RHEL, SUSE, etc., whatever ESX supports. There will be a management interface and we expect in the future to plug into other management products/platforms based on customer demand.
As for VMware certification, there isn't a direct certification available by VMware for a product like ours (there could be, one day). We will support the product on ESX and have an agreement with VMware around this. We are also in the TAP and the community source program, so we are being very strict in adhering to VMware's best practices.
DXtreme has a small component that installs through the console
Just to clarify the matter .... do you mean the "ESX Service Console" ?
Actually, it isn't installed through the console, but you can view it through the console. It is installed as a shim via. a CD/DVD following VMware's driver loading certification process (at least this appears to be the only way to conform to VMware's requirements for doing this). We then interface to the shim to do our prediction.
Does this make it a bit clearer or have I created more confusion?
So basically, it's a kernel driver that is getting loaded that interfaces with VMFS? Can this also affect speeds on the service console itself? If so I would LOVE to get a demo of this stuff in my lab.
Kix @ Vizioncore
I am also interested. Please send me a PM
If you guys are interested, you might want to also sign up to request becoming a Beta participant or Beta site for the product.
You can sign up to find out more information or apply for the Beta program.
The component that is loaded via the CD/DVD is a kernel mode driver. There can be an increase in efficiency on the service console, but I'm not sure how pronounced it would be. I will take a look and see.
If you sign up for the Beta program as David mentioned, we can work with you based on your specific needs/scenario.
So lets get this right...the Hypervisor adds about 25% overhead typically to a VM, with this bad boy are we talking about removing this overhead?
INterested to know how this works on SAP and high level DB's.
Not to pedantically derail the thread, but the hypervisor does not typically add that much overhead to most workloads. As with all things, it depends on the workload (how much memory, I/O, context switches, etc.) and the resources available. Overhead is also probably the wrong way to look at it, since your end user just cares that they can use their app just like they did before.
See our performance blog, VROOM! (http://blogs.vmware.com/performance/) for more information about performance and benchmarking.
High level DB's are one such application that shines when using this software. Intensive database workloads are usually found to be less than ideal candidates for virtualization and consolidation. The reason for that is usually a simple one, the disk I/O bottlenecks and then performance suffers. We've all been there. The DXtreme software attacks this problem head on. This is what InfoWorld's Test Center took a look at when it tested the performance of its virtual machines with IOMeter and MySQL sql-bench benchmarking.
To see it in action for yourself, you might want to convert a backup copy of your virtual machine image to a VMware Server environment, download InovaWave's DXtreme Standard edition trial version and give it a shot. Obviously, on a large production server you would want to use something like the Enterprise edition... but if you want to see it first hand for yourself, if you have a test server, migrate your ESX image to a Windows box running VMware Server and try it out. Shoot me a message if you need anything.
TechTarget recently did an interview article with three users of the software. Maybe it will help answer a few questions people have as well.