Celebrating its tenth year, VMworld was a busy event this year—as it has been every year for the past nine. Among the key takeaways I picked up this year are:
1. Virtualization and Cloud Computing Have Come a Long Way
There were some great statistics to give virtualization and cloud computing historical context and let us know how far we’ve come and how far we have to go. To put things broadly, there have been three computing eras:
- The Mainframe Era: where there were thousands of users and hundreds of apps
- The Client-Server Era: where there were millions of users and thousands of apps
- The Mobile Cloud Era: where there are billions of users and millions of apps
Each of these three phases is still around, but as you can see, each era brought three orders of magnitude increase (that’s 1,000 times) in both the numbers of users and the numbers of apps. That is a very dramatic shift in half a century!
Looking just at the progress around virtualization, there have also been three periods:
- Basic Virtual Machines (VMs): where test and dev and non-production critical VMs are virtualized. Most companies today have at least reached this stage.
- Mission-Critical VMs: where the critical servers the company runs on (SQL, Exchange, Oracle, SAP, etc.) are virtualized. According to VMware’s research, about 54% of companies are in this stage today. They also noted that half of all Oracle and SAP servers and three quarters of Microsoft servers in this category are virtualized today. VMware announced vSphere 5.5 (more on that below) and how it is optimized for Hadoop. Hadoop is used for big data analytics and the associated required data storage, which may be in the petabytes, and spreads the compute and storage load across many nodes.
- IT-as-a-Service (IaaS): the phase where we use cloud computing (private, public, or a hybrid of the two). IT provides the resources and controls, but the VMs may be anywhere. Again, according to VMware, 21% of companies are in this stage today.
It is interesting to note how, in the roughly ten years that VMware has been pushing server virtualization, the number of VMs per admin has changed. In the first phase, where there are typically more and smaller physical servers, the number of VMs per admin averages around 120 (across companies of all sizes). In the Mission-Critical VMs phase, the number is 170, and in the IaaS phase, the current average is 363! To manage that number of VMs well, IT has to have better tools to control what is deployed where, the governance to control data security and access, and the ability to manage and monitor such a diverse set of assets.
2. vSphere 5.5 and vCloud Suite 5.5 are Here
At about this time every year, VMware announces a new version of vSphere and the associated vCloud Suite. This year was no different. Some of the new features include:
- Double the amount of physical host resources: For example: the number of CPUs has increased from 160 to 320, the maximum RAM from 2 TB to 4 TB, and the number of virtual CPUs on an ESXi host has doubled from 2048 to 4096.
- Hot pluggable PCIe SSD drives: SSD drives can now be hot added and removed to increase performance and minimize down time.
- Hardware version 10: This hardware version adds support for a virtual SATA interface that supports up to 120 virtual disks and/or CDs.
- VM latency sensitivity: A new group of settings has been added for those virtual machines that are latency sensitive, guaranteeing physical memory and CPU to the VM (among other resources), instead of sharing those resources with other VMs to minimize the latency inherent in that sharing.
- Graphics acceleration for Linux VMs: Out of the box, most modern Linux distributions include VMware’s drivers, providing network and video optimizations among others.
- vCenter Single Sign-On (SSO): SSO has been greatly improved in terms of installation and support and by adding support for multiple Active Directory forests. It was redesigned from the ground up to be simpler to install and maintain as well as to configure for high availability by incorporating a new multi-master architecture.
- The vSphere Web Client: The web client’s performance has been increased and new features added, including support for drag-and-drop and better filtering.
- vCenter Virtual Appliance (VCVA): The VCVA now supports many more hosts and VMs when using the embedded Postgres database.
- 62 TB VMDK: The maximum disk size for a VMDK or an RDM has increased to 62 TB when using VMFS-5 volumes.
- 16 Gb Fibre Channel (FC) support: 16 Gb FC support was added in 5.1, with 16 Gb cards supported to connect the server with the FC switches, but only 8 Gb was supported from the array to the switches. vSphere 5.5 extends that to 16 Gb end-to-end.
- Link Aggregation Control Protocol (LACP) enhancements: LACP has been upgraded to support 22 new load-balancing (hashing) algorithms and support for up to 64 Link Aggregation Groups (LAGs).
- 40 GbE support: vSphere 5.5 supports 40 GbE, up from the previous maximum of 10 GbE.
There are many more enhancements. Refer to the documentation and release notes for more details.
3. VSAN is Here
VSAN is part of VMware’s drive towards Software Defined Storage (SDS). It takes local drives (both magnetic and SSD) and turns them into centrally managed, replicated, fault tolerant storage. There are three layers in VSAN:
- The policy-driven control plane: Policies define how many copies and how the copies are managed and replicated. As such, even if a drive fails, the policy will automatically replicate the data to another location. The policy can specify the available capacity (or reserve space for things such as snapshots) as well as availability (tolerating n failures).
- VM-centric data services: Services such as deduplication, encryption, and thin provisioning can all be implemented as needed and controlled by policy.
- Virtualized data plane: The virtualized data plane is built into the VMkernel and allows the hypervisor to pool and manage local storage resources natively. By default, VSAN will use all local disks that do not have a partition on them.
This feature works seamlessly with the features that utilize shared storage, including HA, DRS, and vMotion. It is managed with vCenter and is enabled when creating a DRS or HA cluster with at least three nodes; it can also be enabled on an existing cluster if desired. VSAN allows for dynamic performance and capacity scaling.
VSANs have their own datastore format, though you can use sVmotion to move VMs between them. It will automatically load balance across the disks and servers based on capacity and performance. VSAN will automatically throttle the rebuild I/Os to minimize the impact on production traffic.
Suggested use cases include VDI, Tier two and three production workloads, test and dev, and as a DR target (via vSphere Replication [VR] and/or Site Recovery Manager [SRM]).
It is currently in public beta with general availability expected with the vSphere 5.5 Update 1 release expected in the first half of 2014. You can access the VSAN Beta at http://www.vsanbeta.com.
4. NSX is Here
NSX extends network virtualization across both hypervisors and network vendors. You create virtual networks programmatically and then clone and deploy them via templates automatically as needed.
NSX provides Layer 2 (L2) switching and Layer 3 (L3) routing, firewall, and load-balancing services. NSX allows for switching on top of existing physical networking infrastructure without changing anything, much like creating a VM does not require any changes to the ESXi server. NSX can route between two VMs on the same server in different subnets internally (i.e., without accessing the network). You can vMotion a VM from a standard or distributed switch to the NSX platform with no downtime.
This technology is expected to be available in the fourth quarter of 2013.
5. Log Insight is a Great Tool to Analyze and Filter Large Quantities of Unstructured Data
Log Insight was released earlier in 2013 and is designed to import large quantities (potentially in the hundreds of gigabytes or terabytes) of unstructured log data, analyze it, and automatically highlight issues. It natively manages vSphere logs and is extensible to include operating system, application, storage, and networking logs as well. It is deployed as a Virtual Appliance (VA). It can scale out across VAs easily, scaling to additional ESXi servers and/or storage locations for better performance.
Log Insight can integrate with vCenter Operations Manager (vC OPS) for richer data analysis and reporting. It can help find anomalies and predict where things will be in the future. It is extensible via content packs from VMware, third-party vendors, and partners, as well as the community at large. Currently View and NSX have content packs available from VMware; other content packs are available from EMC, NetApp, Cisco, Brocade, VCE, and others. These and others will be available on the Solution Exchange (http://solutionexchange.vmware.com/store/loginsight).
Among the use cases are troubleshooting and root cause analysis, security, and compliance. It is needed as larger and larger distributed architectures with more nodes, more devices, and more complexity are integrated with increasing numbers of layers in the application and infrastructure stacks. It is designed to remove complexity and find hidden value in these often-ignored logs. It is simple and easy to use. Just click and highlight to select words, click Extract field, and query like a field in a database. These queries can even be graphed, including how often it has been triggered historically to help set proper thresholds. You can configure e-mail alerts for when these thresholds are exceeded.
Tip: Anyone who follows @vmLogInsight on Twitter will get five free Log Insight licenses.
6. vCloud Hybrid Service (VCHS) is Here
VCHS is designed to allow for common networking (including keeping the same IP address on-site as well as in the cloud), management, security, and support. The early access program began in June, and VCHS has a release cycle of every six weeks. It also enables DR-as-a-Service (DRaaS), integrating with SRM (with availability expected in the fourth quarter), as well as Desktop-as-a-Service, integrating with Horizon seamlessly. This product is available from VMware directly and is based on their data centers in Dallas, Santa Clara, Las Vegas, and Sterling, Virginia (greater Washington DC), as well as through a partnership with Savvis with their data centers in Chicago and New York City.
As always, there was so much more that was discussed, including VDI, DR, and the future developments in many products (including vMotion, Fault Tolerance, vSphere Data Protection [VDP], and the advanced version of VDP). There were broad sessions for executives, deep-dive technical sessions for administrators, planning sessions for designers, round-table discussions, and more. In addition to all of the sessions, there was a huge expo hall with a lot of great vendors showing their integration with and support for various VMware products, from VDI to cloud to network to storage to servers. I could have spent my entire time in the expo hall alone just to learn about the depth and breadth of products that integrate with VMware. All of this still doesn’t begin to cover the sessions for partners or those in government or other vertical markets, such as healthcare or finance. In fact, VMworld has grown so large that the Moscone Center wasn’t large enough. They had sessions in several off-site hotels.
If you are involved with virtualization or cloud computing in any way, I suggest you make plans to attend as often as time and budget will allow and to catch some of the sessions that they publish (some for free and some as a paid subscription) at www.vmworld.com. If you couldn’t make San Francisco this year, there is always VMworld Barcelona in October, but that is a day shorter than the US version.