Tuesday, December 20, 2011

Network Management in VDI

Virtual Desktops is poised for prime time in 2012. After several years of feeling the pain of supporting multiple platforms and dealing with all the headaches - IT departments of most of major corporations will adapt VDI in a big way in 2012.


While IT administrators may have tested various aspects of VDI: implementation, integration into legacy applications and data management, there are several network issues that are hidden till VDI is deployed and these issues will have to resolved for successful VDI deployments.



VDI technology essentially changes the network traffic in organizations. As VDI’s are installed in a Datacenters, the WAN traffic to/from the data centers explodes. Since the network and datacenters were designed for PC clients, the network will be inadequate for VDI deployments. Moreover the network problems will not revealed immediately - rather it will be revealed in bits and pieces and that will force an ad-hoc update/upgrade to the networks.


The network issues will crop up and the entire deployment process will not be smooth sailing for enterprise wide VDI deployment.


Managing VDI deployments


From the IT managers perspective, having only one VDI platform will be an ideal solution, but in reality companies will have a heterogeneous (VMWare, Citrix, Microsoft) environment. This adds to the complexity of managing the VDI environments. Having a heterogeneous VDI environments will create a need for a unified Infrastructure manager.
All this means a big need for network Infrastructure management software - such as Ionix ITOI.


Major Network Challenges in VDI deployments


WAN Issues
As VDI would change the data routing within an enterprise in a big way. Currently the individual PC’s connect to the data center servers over the LAN - but with VDI, all end devices will connect to the servers over WAN - as all the virtual desktops are now running on the servers and users are accessing the servers over the WAN gateway. .



As the number of VDI users increase WAN traffic could increase exponentially. The only way to manage the WAN traffic will be opt for WAN optimization technology such as Cisco WAAS, Silver Peak, Blue Coat, F5, Expand Networks, Exinda etc



QoS and bandwidth management can play a significant role in mitigating the WAN contention issues. Screen refresh, for example, is highly interactive and very sensitive to congestion. Video traffic is also very sensitive to congestion. QoS and bandwidth management can ensure that these applications perform well. While file transfer and print jobs are not very sensitive to congestion, they can induce congestion on the WAN and hence impact the other types of applications. QoS and bandwidth management can ensure that these applications do not interfere with applications that are sensitive to congestion.


VDI will help IT departments consolidate and optimize remote desktop management, but they need to spend time focusing on optimizing and controlling the WAN connections between the VDI client and the VDI server between branches. Any bumps in the WAN will translate to a bad user experience for the remote VDI user and a support call to IT.


Major WAN challenges are:



  • Hidden choke points that only become apparent when stressed

  • Spikes in network traffic that are hard-to-predict before full roll out

  • Intermittent network even when average bandwidth is high

WAN optimization in VDI


VDI clients will be mobile. Laptops, Netbooks, Tablets, & even smart phones - will be the typical clients for VDI. This implies that the user’s Internet connection standards cannot be gauranteed. In such cases, users on a low bandwidth connection will experience severe performance degradation.

LAN Issues


VDI could also increase LAN traffic subtle ways. As part of VDI, a virtual machine (VM) on a data center server hosts a complete user desktop including all its applications, configurations, and privileges. The client then accesses the applications via the network with the desktop and application objects delivered on demand over the network from the virtual desktop servers via a remote display protocol, such as Microsoft Remote Desktop Protocol (RDP) and/or Citrix’s ICA protocol. The RDP/ICA traffic could spike at times creating choke points within the corporate LAN networks. In general with VDI, the RDP/ICA traffic will be much higher than the average.


LAN traffic in a traditional IT deployments.


In addition to the LAN RDP/ICA traffic, user’s systems could also have other applications (data/music/video/photos that could be running over the LAN network.
So in a nutshell with VDI deployments, one would see rapid increase of LAN traffic and create multiple choke points in the LAN network. The problem will get worse as more seats are added into the system.


Storage Issues


VDI is essentially a hybrid approach, where each end user has a thin client and connects to a private Windows XP or Vista image—a virtual machine hosted on VMware Virtual Infrastructure. This approach allows IT administrators the greater control over the user environment usually provided by Terminal Services or Citrix environments by consolidating the Windows images on server class hardware. It also allows the images to be stored and managed in the datacenter, while giving each user a full personal copy, which requires no introduction or explanation to a normal user.


VDI relies on central data storage for both block & file type data. VDI must handle both structured and unstructured data.


The following table, adapted from the VMware VDI Server Sizing and Scaling white paper, compares the disk usage of light and heavy users for a large number of VMware VDI virtual machines (approximately 20) on a single VMware SEX host. It suggests that over 90% of the average information worker’s disk I/O consists of read operations.




Before intelligent storage subsystem choices can be made, these throughput values need to be converted to Input/output operations Per Second (IOPS) values used by the SAN/NAS storage industry. A throughput rate can be converted to IOPS by the following formula:
.Throughput (MBtyes / sec)×1024(kbytes/MByte)
IOPS = -----------------------------------------------------------
.Blocksize (kbytes/ IO)


Even though the standard NTFS file system allocation size is 4k, Windows XP uses a 64-Kbyte block size, and Windows Vista uses a 1-MByte block size, for disk I/O. Using the worst case (heavy user) scenario of 7.0 MBytes/sec throughput and the smaller block size of 64kbytes, of a full Windows XP group of machines, the generated IOPS for approximately 20 virtual machines is 112 IOPS.

ESX server supports Fiber Channel Protocol, FCoE, iSCSI, NFS & 10GbE.
As the number of VDI users increase, the storage system will face both capacity and performance issues. This types of VDI workload senario can bring a traditional storage system to its knees.
VDI tends to create a I/O spike which will require complete redesign of storage systems. Storage tiering - ‘tier 0’ along with Flash drives or Solid state drives(SSD) can solve the storage I/O spikes and also improve the performance problems by having the most frequently used data into SSD. VDI deployments are typically read intensive (90% read & 10% write).
In short, VDI creates additional overheads on storage management and administration.


VDI Fault & Performance Management


One of the most common faults that occur in a VDI environment is VMWare losing connection to storage. When a VM losses connection to its datastore, the VM becomes unresponsive. This problem becomes more acute in VDI deployments with uses vMotion.



In case of VDI, as such a case will result in an unresponsive desktops and the subsequent increase to IT tickets.


WAN Management
The basic requirements of WAN management are:
1. Discover the entire WAN gateway network components. Including all WAN optimizing devices (Blue coat, F5, Cisco WAAS, Silver Peak, Exinda, etc.)
2. Fault Management of WAN gateway network devices
3. Performance Monitoring of WAN gateway network devices. Monitor QoS & SLA parameters through SLA & QoS MIBs.
4. Performance Management via vCenter Operations Enterprise for the entire WAN gateway network
5. Remote Configuration management of WAN network for bandwidth/performance optimization.



LAN Management



1. Discover & monitor the entire enterprise LAN - wire & Wi-Fi network for availability & performance.
2. Performance monitoring of Wi-Fi networks as people are increasingly using Wi-Fi for connectivity to corporate network
3. VPN tunnel connection monitoring. Monitor VPN tunnels & VPN gateway for any faults that cause VPN tunnels to go down.
4. Security & Authentication management to detect any unauthorized log-in & intruder detection on LAN.


Storage Mangement
1. Discover & Monitor SAN switches & SAN network
2. Discover Storage array, LUNs & WWN
3. Discover all VM’s for VDI and the WWN associated with the VDI
4. Corrleate unresponsive VM/VDI events to respective storage or network failures


VDI Needs better Storage Managment


VDI creates humungous volumes of data under managment. Since all the user data will be centrally stored, the volume of data that needs to be managed will be HUGE. To understand this, consider this:


Average user today has 100GB of data in their desktops. So for 2000 users, the total volume of data will be 200TB. This translates to a disk space of 1000TB - accounting for RAID-5, and active backups. Note that this is additional data under management, which resides today unmanaged in physical desktops.



Naturally, it makes sense to use data de-duplication technologies and implement rule based data management system to minimize the total volume of data under management and disk requirements.



Closing Thoughts


VDI deployments will create a big need for an automated fault, performance, & configuration management solution that can span across the virtual domain into the physical domain of servers, network and storage.

Successful VDI deployment will rely on automated IT infrastructure management solution that can provide provisioning, Automated root cause analysis and alerts to potential provisioning issues, automatic fault identification and active performance management systems.

In an ideal world, the IT infrastructure management solution should be able to correlate faults in the virtual domain (VDI) to the underlying hardware, network & storage problems and alert users and IT administrators before the faults are detected by the end users.


References:
http://kb.vmware.com/selfservice/microsites/search.do?language=en_US&cmd=displayKC&externalId=1009553

3 comments:

durga said...

Wonderful blog & good post.Its really helpful for me, awaiting for more new post. Keep Blogging!

Software Product Development

Shane Castane said...


Publish links to other blog posts and have bloggers feature your content on their website. This can work should you return the favor and when bloggers see your blog as being a great way to get publicity. Ensure there's no conflict of curiosity in between you and the blogger who's going to feature you. Discover More Here

Performance Improvement Plan

Unknown said...

I think it is very useful device. Thanks for sharing this information through your article.
Best Network Management Solution | Best Application Management Solution