July 2013 archive

Top 5 Tips for monitoring applications in a Virtual Environment

The focus on applications often gets lost in the shuffle when implementing virtualization which is unfortunate because if you look at any server environment its sole purpose is really to serve applications. No matter what hardware, hypervisor or operating system is used, ultimately it all comes down to being able to run applications that serve a specific function for the users in a server environment. When you think about it, without applications we wouldn’t even have a need for computing hardware, so it’s important to remember, it’s really all about the applications.

So with that I wanted to provide 5 tips for monitoring applications in a virtual environment that will help you ensure that your applications run smoothly once they are virtualized.

Tip 1 – Monitor all the layers

The computing stack consists of several different components all layered on top of each other. At the bottom is the physical hardware or bare metal as it is often referred to. On top of that you traditionally had the operating system like Microsoft Windows or Linux, but with virtualization you have a new layer that sits between the hardware and the operating system. The virtualization layer controls all access to physical hardware and the operating system layer is contained within virtual machines. Inside those virtual machines is the application layer where you install applications within the operating system. Finally you have the user layer that accesses the applications running on the virtual machines. Within each layer you have specific resource areas that need to be monitored both within and across layers. For example storage resources which are a typical bottleneck in virtual environments need storage management across layers so you get different perspectives from multiple viewpoints.

To have truly effective monitoring you need to monitor all the layers so you can get a perspective from each layer and also see the big picture interaction between layers. If you don’t monitor all the layers you are going to miss important events that are relevant at a particular layer. For example if you focus only on monitoring at the guest OS layer, how do you know your applications or performing as they should or that your hypervisor does not have a bottleneck. So don’t miss anything when you monitor your virtual environment, you should monitor the application stack from end to end all the way from the infrastructure to the applications and the users that rely on them.

Tip 2 – Pay attention to the user experience

So you monitor your applications for problems but that won’t necessarily tell you how well it’s performing from a user perspective. If you’re just looking at the application and you see it has plenty of memory and CPU resources and there are no error messages you might get a false sense of confidence that it is running OK. If you dig deeper you may uncover hidden problems, this is especially true with virtualized applications that run on shared infrastructure and multi-tier applications that span servers that rely on other servers to properly function.

The user experience is what the user experiences when using the application and is the best measure of how an application is performing. If there is a bottleneck somewhere between shared resources or in one tier of an application it’s going to negatively impact the user experience which is based on everything working smoothly. So it’s important to have a monitoring tool that can simulate a user accessing an application so you can monitor from that perspective. If you detect that the user experience has degraded many tools will help you pinpoint where the bottleneck or problem is occurring.

Tip 3 – Understand application and virtualization dependencies

There are many dependencies that can occur with applications and in virtual environments. With applications you may have a multi-tier application that depends on other services running on other VMs such as a web tier, app tier and database tier. Multi-tier applications are typically all or nothing, if any one tier is unavailable or has a problem the application fails. Clustering can be leveraged within applications to provide higher availability but you need to take special precautions to ensure a failure doesn’t impact the entire clustered application all at once. This may also extend beyond applications into other areas, for example if Active Directory or DNS is unavailable it may also affect your applications. In addition there are also many dependencies inside a virtual environment. One big one is shared storage, VMs can survive a host failure with HA features that bring them up on another host, but if your primary shared storage fails it can take down the whole environment.

The bottom line is that you have to know your dependencies ahead of time, you can’t afford to find them out when problems happen. You should clearly document what your applications need to be able to function and ensure you take that into account in our design considerations for your virtual environment.  Something as simple as DNS being unavailable can take down a whole datacenter as everything relies on it. You also need to go beyond understanding your dependencies and configure your virtual environment and virtualization management around them. Doing things like setting affinity settings so when VMs are moved around they are either kept together or spread across hosts will help minimize application downtime and balance performance.

Tip 4 – Leverage VMware HA OS & application heartbeat monitoring

One of the little known functions of VMware’s High Availability (HA) feature is the ability to monitor both operating systems and applications to ensure that they are responding. HA was originally designed to detect host failures in a cluster and automatically restart VMs on failed hosts on other hosts in the cluster. It was further enhanced to detect VM failures such as a Windows blue screen by monitoring a heartbeat inside the VM guest OS through the VMware Tools utility. This feature is known as Virtual Machine (VM) Monitoring and will automatically restart a VM if it detects the loss of the heartbeat. To help avoid false positives it was further enhanced to detect I/O occurring by the VM to ensure that it was truly down before restarting it.

VMware HA Application Monitoring was introduced as a further enhancement to HA in vSphere 5 that took HA another level deeper, to the application. Again leveraging VMware Tools and using a special API that VMware developed for this you can now monitor the heartbeat of individual applications. VMware’s API allows application developers for any type of application, even custom ones developed in-house to hook in to VMware HA Application Monitoring to provide an extra level of protection by automatically restarting a VM if an application fails. Both features are disabled by default and need to be enabled to function, in addition with application monitoring you need to be running an application that supports it.

Tip 5 – Use the right tool for the job

You really need a comprehensive monitoring package that will monitor all aspects and layers of your virtual environment. Many tools today focus on specific areas such as the physical hardware or guest OS or the hypervisor. What you need are monitoring tools that can cover all your bases and also focus on your applications which are really the most critical part of your whole environment. Because of the interactions and dependencies with applications and virtual environments you also need tools that can understand them and properly monitor them so you can troubleshoot them more easily and spot bottlenecks that may choke your applications. Having a tool that can also simulate the user experience is especially important in a virtualized environment that has so many moving parts so you can monitor the application from end-to-end.

SolarWinds can provide you with the tools you need to monitor every part of your virtual environment including the applications. SolarWinds Virtualization Manager coupled with Server & Application Monitor can help ensure that you do not miss anything and that you have all the computing layers covered. SolarWinds Virtualization Manager delivers integrated VMware and Microsoft Hyper-V capacity planning, performance monitoring, VM sprawl control, configuration management, and chargeback automation to provide complete monitoring of your hypervisor.

SolarWinds Server & Application Monitor delivers agentless application and server monitoring software that provides monitoring, alerting, reporting, and server management. It only takes minutes to create monitors for custom applications and to deploy new application monitors with Server & Application Monitor’s built-in support for more than 150 applications. Server management capabilities allow you to natively start and stop services, reboot servers, and kill rogue processes. It also enables you to measure application performance from an end user’s perspective so you can monitor the user experience.

With SolarWinds Virtualization Manager and SolarWinds Server & Application Monitor you have complete coverage of your entire virtual environment from the bare metal all the way to the end user.

Share This:

The new HP ProLiant MicroServer Gen8 – a great virtualization home lab server

I’ve always liked the small size of the HP MicroServer which makes it perfect for use in a virtualization home lab, but one area that I felt it was lacking was with the CPU. The original MicroServer came with a AMD Athlon II NEO N36L 1.3 Ghz dual-core processor which was pretty weak to use with virtualization and was more suited for small ultra-notebook PC’s that require small form factors and low power consumption. HP came out with enhanced N40L model a bit later but it was just a small bump up to the AMD Turion II Neo N40L/1.5GHz and later on to the N54L/2.2 Ghz dual-core processor. I have both the N36L & N40L MicroServers in my home lab and really like them except for the under-powered CPU.

Well HP just announced a new MicroServer model that they are calling Gen8 which is in line with their current Proliant server model generation 8. The new model not only looks way cool, it also comes in 2 different CPU configurations which give it a big boost in CPU power. Unfortunately while the GHz of the CPU makes a big jump, it’s still only available in dual-core. The two new processor options are:

Having a Pentium processor is a big jump over the Celeron and it comes with more L3 cache. Unfortunately though neither processor supports hyper-threading which would show as more cores to a vSphere host. Despite this its still a nice bump that makes the MicroServer even better for a virtualization home lab server. Note they switched from AMD to Intel processors with the Gen 8 models.

Lets start with a the cosmetic stuff on the new MicroServer, it has a radical new look (shown on the right below) which is consistent with it’s big brother Proliant servers, I personally think it looks pretty bad-ass cool.

2micros

Note the new slim DVD drive  opening on the Gen8 instead of the full size opening on the previous model. One thing to note is that while the new new Gen8 models are all depicted with a DVD drive, it does not come standard and must be purchased separately and installed. The Gen8 model is also a bit shorter and less deep and slightly wider than the old model. On the old model the HP logo would light up blue when powered on to serve as the health light and it looks like on the new model there is a blue bar at the bottom that lights up instead. There are also only 2 USB ports on the front now instead of 4. The old model also had keys (which I always misplace) to open the front to gain access to the drives and components, it looks like on the new model they did away with that and have a little handle to open it. On the back side they have moved the power supply and fan up a bit, removed one of the PCIe ports (only 1 now), removed the eSATA port, added 2 USB ports (2 of them 3.0) and added a 1GB NIC port. This is a nice change especially the addition of the second NIC which makes for more vSwitch options with vSphere. I have always added a 2-port NIC to my MicroServers as they only had 1 previously so it’s nice that it comes with 2.

Inside the unit still has 4 non-hot plug drive trays and supports up to 12TB of SATA disk (4 x 3TB). The storage controller is the HP Dynamic Smart Array B120i Controller which only supports RAID levels 0, 1 and 10. Also only 2 of the bays support 6.0Gb/s SATA drives, the other 2 support 3.0Gb/s SATA drives. There are still only 2 memory slots that support a maximum of 16GB (DDR3), this is another big enhancement as the previous model only supported 8GB maximum memory which limited it to how many VMs you could run on it. The Gen8 model also comes with a new internal microSD slot so you could boot ESXi from it if you wanted to, both the old & new still have an internal USB port as well. The server comes with the HP iLO Management Engine which is standard on all Proliant servers and is accessed through one of the NICs that does split-duty, but you have to license it to use many of the advanced features like the remote console. To license it costs a minimum of $129 for the iLO Essentials with 1 yr support which is a bit much for a home lab server that is under $500.

Speaking of cost  which has always made the MicroServer attractive for home labs, the G1610T model starts at $449 and the G2020T starts at $529, the two models are identical besides the processor and they both come standard with 2GB of memory and no hard drives. I wish they would not include memory in it and make it optional as well and lower the price. If you want to go to 8G or 16Gb of memory (who doesn’t) you have to take out the 2GB DIMM that comes with it and toss it and put in 4GB or 8GB DIMMs. Here are some of the add-on options and pricing on the HP SMB Store website:

  • HP 8GB (1x8GB) Dual Rank x8 PC3- 12800E (DDR3-1600) Unbuffered CAS-11 Memory Kit  [Add $139.00]
  • HP 4GB (1x4GB) Dual Rank x8 PC3-12800E (DDR3-1600) Unbuffered CAS-11 Memory Kit  [Add $75.00]
  • HP 500GB 6G Non-Hot Plug 3.5 SATA 7200rpm MDL Hard Drive  [Add $239.00]
  • HP 1TB 6G Non-Hot Plug 3.5 SATA 7200rpm MDL Hard Drive  [Add $269.00]
  • HP 2TB 6G Non-Hot Plug 3.5 SATA 7200rpm MDL Hard Drive  [Add $459.00]
  • HP 3TB 6G Non-Hot Plug 3.5 SATA 7200rpm MDL Hard Drive  [Add $615.00]
  • HP SATA DVD-RW drive [Add $129.00]
  • HP NC112T PCI Express Gigabit Server Adapter [Add $59.00]
  • HP 4GB microSD [Add $79.00]
  • HP 32GB microSD [Add $219.00]
  • HP iLO Essentials including 1yr 24×7 support [$129.00]

With all the add-ons the server cost can quickly grow to over $1000, not ideal for a home lab server. I’d recommend heading to New Egg & Micro Center and getting parts to upgrade the server. You can get a Kingston HyperX Blu 8GB DDR3-1600 Memory Kit for $69 or a Kingston HyperX Red 16GB DDR3-1600 Memory Kit for $119.00 which is half the cost.

All in all I really like the improvements they have made with the new model and it makes an ideal virtualization home lab server that you are typically building on a tight budget. HP if you want to send me one I’d love to do a full review on it. Listed below are some links for more information and a comparison of the old MicroServer G7 N54L and the new Gen8 G2020T model so you can see the differences and what has changed.

Comparison of the old MicroServer G7 N54L and the new Gen8 G2020T model:

FeatureOld MicroServer G7 (N54L)New MicroServer Gen8 (G2020T)
ProcessorAMD Turion II Model Neo N54L (2.20 GHz, 15W, 2MB)Intel Pentium G2020T (2.5GHz/2-core/3MB/35W)
Cache2x 1MB Level 2 cache3MB (1 x 3MB) L3 Cache
ChipsetAMD RS785E/SB820MIntel C204 Chipset
Memory2 slots - 8GB max - DDR3 1333MHz2 slots - 16GB max - DDR3 1600MHz/1333MHz
NetworkHP Ethernet 1Gb 1-port NC107i AdapterHP Ethernet 1Gb 2-port 332i Adapter
Expansion Slot2 - PCIe 2.0 x16 & x1 Low Profile1 - PCIe 2.0 x16 Low Profile
Storage ControllerIntegrated SATA controller with embedded RAID (0, 1)HP Dynamic Smart Array B120i Controller (RAID 0, 1, 10)
Storage Capacity (Internal)8.0TB (4 x 2TB) SATA12.0TB (4 x 3TB) SATA
Power SupplyOne (1) 150 Watts Non-Hot PlugOne (1) 150 Watts Non-Hot Plug
USB ports2.0 Ports - 2 rear, 4 front panel, 1 internal2.0 Ports - 2 front, 2 rear, 1 internal) 3.0 Ports - 2 rear
microSDNoneOne - internal
Dimensions (H x W x D) 10.5 x 8.3 x 10.2 in9.15 x 9.05 x 9.65 in
Weight (Min/Max)13.18 lb/21.16 lb15.13 lb/21.60 lb
Acoustic Noise24.4 dBA (Fully Loaded/Operating)21.0 dBA (Fully Loaded/Operating)

Share This:

Hell must of frozen over…

…because I actually got a VMworld session approved!  I last had a VMworld session approved (2 of them) in 2010 for my Deep Dive on Virtualization and Building a Home Lab sessions. I’ve really given up hope in recent years of getting another one due to the sheer number of submissions and because VMware has so many sessions because there product offerings are so broad now. This year I had one of the sessions that I submitted through HP “The Top 10 Things you MUST know about Storage for vSphere” approved, it is based on a presentation that I built for our VMUG user conference sponsorship’s this year. As part of HP’s sponsorship we were guaranteed a few session slots but my session made it on its own outside of those slots. So thank you to all who voted for my session and I look forward to seeing you at VMworld.

Share This:

VMware & Microsoft positioning in Gartners Magic Quadrant over the years

Gartner just released their annual “Magic Quadrant for x86 Server Virtualization Infrastructure” report for 2013 and Microsoft continues to draw closer to VMware’s top position. If you look at the differences over the years VMware pretty much hasn’t moved and remains near the highest position that you can be in the quadrant as they lead in both their ability to execute and completeness of vision. That should really come to no surprise as VMware still is and always has been the market leader and have set the bar high in the x86 virtualization field. But Microsoft is slowly but surely closing the gap, I’d be surprised if they ever pass VMware in either of the directions given VMware’s lightning focus on virtualization, but I think they are close enough for most companies to take Microsoft seriously. Also to note the niche players are still mostly niche players and will most likely remain there for the foreseeable future as VMware & Microsoft dominate the x86 virtualization playing field. Citrix isn’t really doing much on the server virtualization side anymore and a large percentage of their client virtualization installations are on non-Citrix server hypervisor platforms like vSphere.

mq-square-small

Share This:

Come see me speak at the Phoenix VMUG User Conference on July 11th

Last year the VMUG leaders in Phoenix invited me to present their keynote at the annual Phoenix VMUG User Conference. I love VMUG’s but had to turn it down unfortunately because of conflicts and having to travel there from Denver. Well I felt bad about having to do that so I ended up just recently moving from Denver, CO to Phoenix, AZ so I could be available to do it if they asked again this year. So as a result I’ll be speaking at the Phoenix VMUG this year which takes place on Thursday, July 11th at the Desert Willow Conference Center in Phoenix. I believe I’ll be doing the morning keynote which will be based on a deck that I will be presenting at VMworld this year for a session that was approved titled “Top 10 things you MUST know about storage for vSphere”. So if you’re in the Phoenix area make sure you head on over to the VMUG website and register. I look forward to meeting my fellow VMware geeks in the valley and participating in the VMUG community here.

Share This: