Posted by Brian on Nov 20, 2011 in Performance, VDI, View, VMware | 0 comments
The world is just now starting to get a glimpse of what vCenter Operations for View could be like. I really hope that this product comes out of the gate with a strong feature set and delivers a big win for VMware. This would really strengthen their VDI offering.
The ability to monitor performance of the connections between the endpoints and the VMs running in the data center is a vital metric that needs to be tackled by VMware. This is something that Citrix is already delivering with XenDesktop and I like what they are doing so far. You can see the latency measurement between the connection and also information like client version, IP addresses and broker that it connected through. All very helpful information for troubleshooting performance and connection issues.
I like the fact that VMware has added counters for Windows that can be monitored using Perfmon inside the OS, and you can always fire this up to look at something. But I think this needs to continue to develop further to make these connection and performance issues easy to continually monitor and troubleshoot. In the preview videos that VMware posted on this blog post are mostly centered around monitoring the infrastructure, this is what vCOPs already does. The last video did show some tasty nugguets about PCoIP monitoring which looks promising. But some type of a client summary page would be very helpful so Admins do not have to drill down into 10 screens to get the picture unless they want that level of detail. It also needs to provide performance monitoring for client connections and end user experience. Below is a list of things that I think would be very helpful in a VMware View deployment to monitor.
- PCoIP connection latency
- VM login times
- Client version
- Connection server client is connected through
- Connection type (PCoIP vs. RDP)
Below is a sample of what Citrix is offering today with their XenDesktop product. From this session screen you can shadow session which I wish VMware would add into View Manager. Then there are details about latency, connection type, endpoint details, which items are enabled within the HDX/ICA protocol. Overall a pretty good looking and helpful console from Citrix.
The lower part of the same screen shows you some hypervisor and broker health status. There is a simple graph that shows you CPU, Memory and Network usage for the VM that you are looking at.
read more
Posted by Brian on Nov 17, 2011 in Cloud, Orchestrator, vCloud, VMware | 0 comments
In working on several Cloud related projects one of the items that sticks out to me is the need for deeper automation within the vCloud Director product. I understand this is still just version 1.5, but with how hard VMware is pushing the “Your Cloud” journey. I think that some parts are just not ready for what some companies need to do in the way of automation.
If self-service is suppose to be such a big part of Cloud, then the need for automation is going to play a big part. Not everything can be accomplished from creating templates and using customization to change the identity of the new VM. In server virtualization this worked great and saved time for most IT shops. But there were still manual processes that some shops needed to do. This breaks the idea of self-service IT, if a user still relies on someone to execute a manual process to have a VM or application provisioned from vCloud.
I guess what this mostly deals with is private cloud. Many IT shops are trying to automate the creation of as many servers and platforms as possible, to reduce their work load in provisioning new servers. But there are still some manual processes that need to take place and I think that being able to tie vCenter Orchestrator more tightly with vCloud Director could go a long way in help this issue.
Other cloud software companies such as DynamicOps are already doing this type of thing. By making the workflow or automation part of their offerings built into the same admin console. This allows for tight integration and opens up the options for what you are allowed to automate.
If you listen to rumors and in dark alleys you might hear that this type of integration is coming from VMware in a future release. Nobody knows if it will be the next release or even when that will happen.
read more
Posted by Brian on Nov 15, 2011 in VDI | 0 comments
If you are thinking about starting or are currently working on a VDI project, you need to do some type of assessment on your current desktops. Without collecting performance and usage data from your current PCs you will only be guessing what on how to design your infrastructure for VDI.
If you are guessing or making too many assumptions about your users, the design is going to either be extremely over solution-ed or will perform badly. By doing your homework you are collecting the data that will allow you to make intelligent decisions on CPU, disk and network performance that will be required.
There are a few applications on the market that do this type of desktop assessment, Liquidware Labs and Lakeside Software are the leaders in this space. In this post I am showing the more valuable slides from the presentation that is created by the Stratusphere Fit tool after collecting data from your PCs.
To deploy the tool you only need to import a Virtual Appliance and assign it an IP address and DNS name. You then export a collection install file that needs to be pushed out or installed on any computer that you wish to collect data from. You can create user and machine groups for different use cases or any logical grouping that might tie in with your design or business case.
I have pasted slides from the presentation below and have made some notes around a few of the slides. This is a sample report that I created using a test environment. There was no grouping setup so a few slides were removed that would present data based on groups. This is just to provide an idea of what type of data you can get by doing an assessment and what Liquidware Labs can help you with.
Besides the presentation below there are a number of built in reports that you can pull or schedule to run on a reoccurring basis. There are several output formats that you can export reports in such as PDF, Excel and Word, among others. There is a good report that will give you summaries of user data and types of files being stored. This is very helpful when planning for profile storage.
The slide below is the opening image that you can customize with Vendor and Customer information.
This slide shows details about the scope of the assessment, such as date range, number of desktops, users and groups.
The slide below shows some of the ranges of how decisions for the Fit ratings will be calculated.
The slide below is showing the number of computers and which group they were placed in.
The slide below is showing peak and averages for all desktops and the different metrics that are measured.
The slide below is probably my favorite one. It shows the averages for all desktops hourly for a week day. So you can see the IOPs are high first thing in the morning when users are logging in and then look for other details that will aid in your design.
The next slide is showing details about the physical desktops, such as age, CPU and Memory usage.
The next slide gives details about the different CPUs that are in your desktops and their utilization.
The next slide covers the physical memory configuration in the desktops.
The next slide gives a breakdown of what Operating System is installed on the desktops being monitored.
The next slide covers local storage on the desktops. You will get a rough idea of sizes and how much data is being used. Note: there are much more detailed reports that can be run to find out more about user data.
The following slide covers devices connected to PCs, you can see that a lot of local or built in devices show up on this report also.
The following slide covers monitors and printers.
The next slide shows a summary of the most used applications.
The next slide shows applications that are used most based on time.
The next slide is application related and covers CPU utilization.
Now an application view related to memory consumed.
And finally an application view that relates to IO consumed.
The next slide is showing applications and a graphics intensity rating.
The next slide shows a view on how your desktops are for VDI candidates.
The next slide is showing how your users will fit as VDI candidates.
The next slide is covering how different applications qualify as Virtualization candidates.
The next slide is showing a list of machines and highlighting PCs that are on the bubbles or are not good candidates for VDI.
The next slide is showing a list of Users and highlighting user accounts that are on the bubbles or are not good candidates for VDI.
read more
Posted by Brian on Nov 11, 2011 in VMUG | 0 comments
Well October 31st and Halloween have come and gone and I’m just getting around to writing my recap. It was a busy day all round between the overwhelming success of Chicago’s first VMUG user conference and sugar overload from candy. I can’t say how proud I am of the turn out that the VMUG had. We broke an attendance record for a first time user conference with 500 people showing up at the conference. I believe the previous record was 390, so we have something to be proud of Chicago VMUG members.
The conference had speaking tracks from VMware and Partners on some great topics. I was able to attend sessions on View 5 and related products and a deep dive on ThinApp. Both were really good sessions with great speakers from VMware. I know that the VMUG leaders would like to thank everyone that came out to the conference and the Vendors that sponsored for making it all possible. But a huge thanks goes to VMware and the VMUG organization for helping plan this big event and making sure things went off without any issues.
I was able to meet a bunch of new VMUG members from Chicago and look forward to staying in contact with you all and meeting more. I hope that people that attended the user conference are able to attend our quarterly VMUG meetings that we hold. Growing our membership and creating a community around VMware in Chicago is a big part of the VMUG.
The Picture below is a view from the View 5 session that was presented by Josh from VMware. He covered all the infrastructure parts of View 5 and what roles they play. There were several good questions from the audience and he covered some real world implementation stories.
The image below is a picture from the exhibit space at the VMUG conference. We had tremendous vendor interest in the meeting. There was 55 boots in the space and some really great vendors and partners were there talking with VMUG members and helping them with products and solutions.
a
read more
Posted by Brian on Nov 9, 2011 in Blade Servers, Cisco UCS, Hardware, HP | 0 comments
This is a conversation that gets asked a lot when discussing the platforms with customers. Can you explain to my why Cisco UCS is easier to manage and what tools do I need to use on a daily, weekly or monthly basis. The answer is really easy with Cisco UCS, you just need to use UCS manager. For HP the conversation becomes more splintered, sure you can do pretty much anything on HP blades that you can do on UCS. But how many tools or management points do you have to touch to accomplish these same things. Below I have attempted to list common tasks that you would need to do for setup and on going management of a blade enclosure and blades within.
Cisco UCS Manager – Does all the following in one tool
- Blade Chassis mgmt
- CIMC console connections
- Hardware Monitoring
- Firmware updates – chassis, FEX modules, blades and adapters
- QOS
- Network Mgmt
- Service profiles for blade identities
- SAN connection setup
Roll Based Access Controls – Can assign ID’s access to all or just allow server, network or SAN access
HP Blades – using Virtual Connect modules of any flavor
- Blade Chassis mgmt. – Onboard Administrators (OA)
- iLO console connections – OA
- Hardware Monitoring – HP SIM
- Firmware updates – HP SIM, FDT or currently used tool
- Network Mgmt – Virtual Connect Manager (VCM)
- Server profiles – Virtual Connect Manager (VCM)
- SAN connection setup – Virtual Connect Manager
- Roll Based Access Controls (RBAC) – need to configure logins per Tool
If you elected to use the new Cisco FEX modules for HP Blade chassis then the following would be affected.
- Network Mgmt / profiles – no profiles, would use physical address from Blades. Port configs done at 5K level. Blades auto map to ports on FEX similar to pass through connections.
- SAN connection setup – Would be done vid FCoE through the 5K switches
- Network QOS – Done on 5K switches and tagged if possible in Operating Systems (ESXi 5 does support 802.1p)
If you have anything that I left off of this list drop me a note in the comments and I will update to keep accurate.
read more
Posted by Brian on Nov 8, 2011 in Blade Servers, Hardware, HP | 0 comments
I was talking with a fellow VMware nerd today about cloud and infrastructure gear. And I can’t help but wonder why it takes HP so long to execute on integrating their products, mostly around the support tools. Look HP owns a complete stack now, network, compute, storage and software. Along with their suites of tools that you need to use to manage all these devices independently. To me this is their biggest failure and does not appear to be getting fixed anytime soon.
This is a big bonus of working Cisco UCS because there is just one place that I need to go to configure, manage, patch a blade environment. With HP I might have to use the Onboard Administrators, Virtual Connect manager and HP SIM. This seems like such a no brainer that HP could have fixed years ago, because the HP Blade systems have been around for years.
So if HP wants to rule the market in the Cloud era they need to reduce their tools into less management points. I should be able to do everything in one console for compute and if they could roll in storage to this also would be a huge win. And if they are properly motivated they can. Now a unified tool for automation of hardware that could reach into VMware with API’s would be the next big win. I know they are working on some of this but last that I saw it still sounded like there would be too many moving pieces.
So HP I challenge you to solve this issue with your massive amount of Talent, code and other resources. You need to become innovate again and make products that people want to own and desire for other reasons than you have a large install base and are cheaper.
If you want some examples you can read my post about the comparison of what tools it takes to manage HP vs Cisco blades chassis here.
read more
Posted by Brian on Nov 8, 2011 in VMware | 0 comments
Recently I had a couple of people email me saying that Google was saying that this site contained Malware. Things have been very busy and I have not been able to write much lately. Will be working on getting back to regular writing soon. I searched for what Google thought was bad code and cleaned up everything. The site is fully patched and has no issues.
I have requested a review from Google to get off of their bad behavior list and should get the clean bill of health soon. If anyone still gets warnings few days after this post let me know in the comments or an email.
Thanks,
Brian
read more
