Panel: William Lam, Duncan Epping, Chad Sakac, Scott Lowe
Moderator: Rick Scherer
This was a pure question driven session, and I tried to capture the questions and response as best I could. Please forgive any grammatical errors or typos.
Q: Is there any capability to address any external storage?
A: Duncan – EVO RAIL will allow you to scale out and up to 16 nodes, so the capacity will go up. VSAN is included in the box. The RAIL appliance can connect to existing external storage, just as you would with any ESXi host.
Q: Can you add a non-EVO RAIL host to the cluster?
A: Duncan – It just uses VSAN, so technically you could. However, the big selling point is ease of management and you can’t use the EVO GUI. So keep cluster members homogeneous. Chad – It’s totally technically possible, but the EVO manager may “freak out” a bit and you will also have OEM support agreement issues. Support system would break.
Q: As a former vBlock architect, what is the messaging regarding EVO RACK regarding vBLOCK.
A: Chad – SAP, Oracle and replicate with lots of features, then use VBLOCK. If you are deploying IT apps without expecting those vBLOCK services, then use EVO RACK. Duncan – EVO RACK is basically a reference architecture. VCE might be interested in taking the EVO logic and using their hardware.
Q: A customer changed a VM to use 30 ‘sockets’ vice changing the cores setting. In the monster sizing preso VMware said to change the socket count vice the core number to fit in the NUMA node.
A: Duncan – The recommendations in that session are generic. There are many combinations of settings for VMs. There’s a lot of licensing aspects to consider as well. Monitor ESX top to see what the RDY time is. The generic recommendations don’t always apply.
Q: From a VM configuration perspective is there a scheduling difference between specifying cores and sockets?
A: Chad/Duncan – There are scheduler differences between cores and sockets. From a performance point of view if you cross NUMA nodes then you may get a memory performance impact. So best to size VMs to fit within a NUMA node. Chad – Crossing the NUMA node could reduce memory performance a lot.
Q: Hyperconverged looks like a big threat to infrastructure vendors like switches and storage arrays.
A: Chad – You can never win playing a defensive game. People who view fundamental technology disruption, like SDS, need to realize that architecture changes and some sales will cannibalize. Duncan – It opens up a lot of opportunities like networking. Some solutions don’t include networking today or even have NSX, so it opens more opportunities than it closes. EMC doesn’t worry about it, except for maybe the sales people. Scott – There are plenty of opportunities for EVO RAIL vendors to add value add, so it’s not really that big of a threat to most vendors. Rick – From a customer perspective this drives innovation and pushes other hyperconvered companies to innovate even more. It’s a win win for the customer. It’s the OEM’s battle to lose.
Q: What do you guys use for your home labs?
A: Chad – Describes a huge home lab with like 300TB of storage, and he had to put in a 30A circuit to power his equipment. Instead of a raise EMC pays Chad’s power bill. Now his lab, based on SSDs, only draws 10A. Describe his Ivy Bridge lab specs. Chad now also uses vClould AIR a lot, to learn about OpenStack and other technologies. William – Use a Mac Mini, which is light on power. Or use hands-on-labs, or use vCloud AIR. There are also good providers that let you stand up VMs or use bare metal. Duncan – Duncan acquired a bunch of EVO RAIL servers. He also uses VMware Workstation, which is great to grab screenshots or just learn. Scott – Describes an OpenStack suite that he uses. Storage is provided by Synology. Rick – He feels that work should stay at work. He also says mini ITX cases can be really great for a home lab. Healthy work life balance is key. Chad – The hosted cloud labs are awesome. If you are an EMC customer then talk to your SE about getting. Portal.demoemc.com.
Q: As VMware Cloud Air tries to build out, how do you talk to partners that use VSPP. What happens to other service providers?
A: Scott – The vCloud Air team is bringing in partners, so that could expand the network. Chad – Says if he was a provider then he would join the vCloud Air network. VSPP’s usually do something special that adds additional value like an industry vertical. Scott – vCloud Air is built to address a general, broad market. There’s always a space for VSPP partners to innovate, such as higher SLAs, or offer unique software. Chad – Once saw a service provider that specialized in clinical research. They are going like crazy, and is an example of going crazy. This is a use case that vCloud air is not going to address
Q: Seems like EVO RAIL is making virtualizing easy. But firmware management can be a pain.
A: Duncan – We do realize this is a problem, but the current version does not address this. All EVO partners handle firmware update differently. So they are working on an abstraction layer that will enable customers to update firmware across the EVO line. EVO RACK is also aiming to solve this problem too. Chad – Chimes in that EMC and VMware are working on the problem. William – Some vendors do have VIBs and other drivers to help automate the firmware deployment process. Chad – IPMI is great, but every vendor does it differently. An abstraction layer to across IPMI, drivers, firmware, etc. is in the works.
Q: We use EMC storage with auto-tiered and the SAP team complains about latency. IOPS goes up and latency is up. Is the root cause the auto tiering itself or something else?
A: Chad – No technology is a panacea. All ILMs will have a delayed response to load. vCOPS has great metrics that can help you find the root cause. This is where all flash arrays have an advantage.
Q: NSX question. Nexus core/edge with Cisco UCS. The value prop of NSX is compelling. How quickly will load balancing and firewall appliances like ASA and F5 be replaced by NSX?
A: Scott – This week we announced the availability of NSX 6.1, for pure vSphere environments. One included feature is integration with F5 load balancers. There is no work currently with Cisco and the ASAs to integrate. NSX doesn’t attempt to address also use cases for load balancers and firewalls, but won’t address all features. So displacing hardware appliances will depend on what features you are using. Chad – Often hardware appliance are used in generic applications and don’t use proprietary features that NSX won’t have. NSX will continue to innovate and add new features.
Q: I’m a storage admin, for about 10 months. Now he has 1000 servers, has NetApp, EMC, Dell Equalogic. Can you give me any ideas for a new storage administrator? University environment. He describes an environment that is in a very unhealthy state.
A: Duncan – Go VSAN. Chad – There’s this thing called LinkedIn to find a new job. Just kidding. Keep firmware up to date. Automate. When you find yourself logging into the UIs, you are doing it wrong. Standardize. Standardize. Standardize. Stick with 2-3 core vendors. William – Use vCO to help automate the storage tasks. You can create workflows for different operational requirements. You can create runbooks. Chad – You could also use EMC VIPR, which is free with community support. VIPR is a harmonizer for all storage “Crap”. Rick – Seriously, VIPR controller is a good option.
Q: If you are looking to write to APIs, and don’t want to use VIPR controller, how would you do it? What kind of SDK platform should I use?
A: William – Depends on where you want to standardize. You could choose OpenStack or vSphere APIs. There’s no magic answer to bridge everything. Scott – There’s no magic rosetta stone. Chad – VIPR controller makes EMC no money. It was written to help bridge the storage gaps. There is no answer to this problem that exists. William – This is a hard problem to solve. You also need to understand what it means, and maintaining the interface is hard. Chad – 200 engineering folks worked on creating the VIPR controller.
Q: Can you please name one product or feature announced at VMworld that makes the biggest impact in the industry? Name some new product or vendor that you also like.
A: Scott – VMware Integrated OpenStack is huge. Pay close attention with Pivotal and Docker around containers. Duncan – EVO RAIL is the clear winner. Look at Thousand Eyes for networking monitoring. Duncan things this is a really cool solution. William – EVO RAIL, since this is a different consumption model. The Google partnership is also good. Chad – Cloud Air is a big one for Chad. Company to watch is Node Prime. Rick – Project Fargo is amazing.
Q: I’m looking for some clarification and need some help in setting the IOPS for round robin (VMAX).
A: Chad – Asks what array you should be using. Use IOPS set to 1 across the board for VMAX.
Q: Question around vVols.
A: Chad – VMware has submitted the vVol spec to the T10 standard’s body. So other vendors may support it. Duncan – VMware is actively trying to help the open source community, and drive industry standards. But it will take time. Scott – VMware is pushing for enablements across the board.