I participated in GestaltIT‘s TechFieldDay which is a sort of inverse conference, where the bloggers and independent analysts go to the vendors and then discuss the information they have received. We visited the following virtualization vendors:
- vKernel where we were introduced to their Predictive Capacity Planning tools
- EMC where we discussed integration of storage into the virtualization management tools as well as other hypervisor integrations
- Cisco where CVN and CVE were discussed in detail.
At the reception at Fenway Park we also had a chance to further our discussion with all these vendors as well as Akorri with their BalancePoint software.Of these vendors what I found interesting is that all have noticed that Hyper-V is now of interest to their customer base so all either have products ready for Hyper-V or are working on products for Hyper-V. Akorri and vKernel have Hyper-V ready products. Cisco and EMC are working with Hyper-V at some level I suspect.
Akorri was not visited by GestaltIT but was well represented at the Sponsored meet-n-greet held for the delegates at Fenway Park. The talk with Akori was about Hyper-V, and their performance and capacity management tool. Akori will be showing their BalancePoint software at Microsoft Management Summit 2010 as well as speaking at the Cloud Expo in New York City. Hyper-V is a very important aspect of Akorri’s strategy. While VMware may own quite a bit of the Enterprise space, there is room for Hyper-V within the Small and Medium businesses, which has been recognized as fertile ground by Akorri and others.
vKernel has quite a few capacity management tools, many of which in their most basic forms are free. Their capacity view program will work with Hyper-V today, yet their Optimization Pack will not yet. That will change hopefully sooner than later, but since they are displaying their wares are the Microsoft Management Summit 2010, I would expect it to be much sooner than later.
vKernel’s approach to their software is not unique, but they are very open about it. They hide all the heavy lifting math behind an interface that tries to be intuitive. For most people it would be, but I found it just a little klunky. To fix this they have the free Capacity View tool that will allow you to quickly see in a dashboard form current issues and from there launch the other tools. I would make use of the dashboard on a regular basis.
The math behind all the Capacity tools is designed to predict when you will need to purchase more resources. In order to be predictive however, vKernel must run within your environment for at least 30 days. With this data it may be able to predict 30 days in the future. If you let it run for more days you get that many days of prediction on capacity which can may use of long term alarms to let administrators know if you up to years in advance to give enough time to work the purchase through the corporate process. The example used was one company that sets the alarms for 45 days in advanced as that is how long it takes to run through the corporate process and receive hardware.
Another interesting Free Tool from vKernel is VM Stats which is a different type of dashboard than capacity view. VM Stats is designed to provide a dashboard to the other tools offered by vKernel within the Optimization Pack, specifically Capacity Right Sizer and Capacity Waste Finder products. Righ Sizer looks at the size of your VMs. Over or under allocated resources while waster finder tries to delete VMDKs, however at this time it does not map VMDKs to existing unregistered VMX files, so if a VM is unregistered but on disks Waste Finder will offer to delete the VMDK, which would then break the unregistered VM. Hopefully, this change will be fixed and Waste Finder will only look at all the files necessary for a VM instead of just the VMDK in order to find orphans.
vKernel has tagging capability similar to many of the virtualization security tools, but like those tools does not contain any method to integrate with each other. In essence causing a proliferation of tags that need to be synced between multiple and distinct systems.
vKernel’s Right Sizer and Capacity tools have the ability to apply optimization to the virtual environment to better balance resource utilization. One of the ones that is often an issue with vSphere is NFS performance numbers. vKernel is working to extract them so they can also balance NFS loads like they do for the block devices. When it comes to storage it is all about IOPs but vSphere hides the data for NFS.
EMC’s storage virtualization management story is all about better integration of their tools with the virtual environment management suite. For VMware, that means better integration with vCenter. They realize that virtualization administrators are not necessarily storage administrators but have distinct requirements on the storage environment. Chad Sakac spoke to us at some length on EMC’s Vision for the future (which I wrote about earlier).
We were given a demo of the vSphere NFS Plugin for Celerra physical and VSA (which you can find links to on Chad’s site.) I fully expect this to be the level of integration we can expect for future EMC developments. The NFS Plugin exposes quite a bit of information about the Celerra specifically previously unavailable metrics data. vSphere itself does not export NFS metrics to vCenter or anywhere, the NFS Plugin for vSphere changes this for the Celerra only. It would be very nice if this worked for all NFS servers.
We were presented two Cisco products, CVN or Secure Multi Tenancy, and CVE which is the VBlock. Future posts will deal with both of these. However, a few observations about each:
- CVN/Secure Multi-Tenancy was not initially designed with security in mind, just guaranteed QoS with respect to IO through the entire system. It does NOT currently look at all the issues with secure multi-tenancy. However, they did say the reference architecture will grow over time. Availability is only one aspect of security.
- CVE/VBlock is definitely not for everyone, the target is quite clearly the cloudprovider or extremely large enterprise that need some level of functionality that they can easily purchase just like building blocks. With VBlock comes a brand new user interface for vSphere and management of all the layers called Unified Infrastructure Manager (UIM).
Both of these products have some interesting points and limitations. The first is that they are designed for LARGE environments and not necessarily inexpensive. The second is that they have their limitations. The goal of SMT has yet to be reached as they do not cover Integrity and Confidentiality. VBlock on the other hand is not for the small user who could benefit the most form such a deliverable.
The VBlock goal is to provide a system that will run out of resources all at the same time. I.e. when you run out of Memory you also run out of IOPs to the array. So that growth is just adding another vBlock or maxing out to the vBlock limits. These limits are VERY important. Once you exceed or go below the limits specified you no longer have a VBlock and loose access to the specialized support team.
The fact that there is not a smaller VBlock is most likely solely due to requirements that all resources run out at the same time. There is most likely some very fancy mathematical equations that fail with lower numbers of virtual machines given that they are really looking at averages while leaving enough room for the peaks.
Akorri and vKernel tools make very powerful adjuncts to your Virtualization Life Cycle process. To allow you to properly judge Capacity, and balance it out across all your nodes. EMC is providing more metrics to the virtual environment management tools to allow administrators to better balance their workloads with the Celerra NFS Plugin for vSphere. For VBlock there is now the Ionix Universal Infrastructure Management which joins together all the tools necessary to manage a VBlock (vCenter, Storage, etc.) into one interface.
The average VBlock changes capacity planning as all resources are designed to expire at the same time, so the needs change to answer the question of ‘when do I need a new VBlock’ instead of individual nodes. This may actually simplify capacity planning for the extremely large enterprise, but the VBlock does not help the smaller businesses.
SMT is a completely different issue and Cisco’s discussion made no bones that its about IO Quality of Server and not necessarily security, which I found disturbing given the name of the product. This will be a discussion for a future article however.
Outside of Akorri and vKernel other tools are looking to work with Hyper-V as well. This is becoming a necessity as the small and medium businesses are really looking at Hyper-V. VMware needs a better line up of products for the smaller businesses. Not everyone has 1000 or more VMs. There needs to be a better product line-up from VMware.
Major Take Aways
- Create a Life Cycle procedure that involves capacity planning is a must while involving all the components of your environment. If you are using EMC products, this could be even easier.
- Hyper-V Utilities are on the rise and becoming more important.
- VBlock is for the large Enterprise
- SMT currently only guarantees IO Quality of Service not all aspects of Security.