What size do you need: Small, medium or large?

I had lunch with my work buddies and good friends Susan and Wendy yesterday. It was great to see them. You may think that three women got together and chatted about shopping or something like that. No; in fact we talked about IBM PureSystems.

I work in the IBM Partner Enablement organization, and my role is to help partners to enable their solutions on IBM PureFlex System. Susan is a team lead in another organization, and she is responsible for the production of a PureFlex component. And Wendy works in the Innovation and Technology team. As you can imagine, IBM has a huge task force for the production and support of PureFlex. We put the experts together from various fields to take care of all aspects including computing capabilities, networking, storage, thermal management, deployment and so forth, so that our customers and partners don’t have to worry about the infrastructure and can focus their effort on growing their businesses. In addition, the Flex System Manager (FSM) is a unified management module that can be put in the PureFlex System to manage all components in that chassis or up to four chassis at the current moment.

We often offer help to our partners to size their solutions. I’m sure the same solution can be configured for a small, medium or large setup depending on the customer’s needs. Take the digital video surveillance solution as an example. Video surveillance systems are typically installed with the intent of preventing crime and other mischief. Small convenience stores need surveillance, as do department stores and large casinos. A small convenience store may have a few of cameras, while department stores may have no less than 50 cameras and typical casinos in Vegas have hundreds.

How do we size these configurations using PureFlex? How do we scale the solution when data grows?

IBM’s Smart Vision Suite (SVS) solution overview

62692 200380440-001 shadow of figure on wall by cctv camerasIBM’s Smart Vision Suite enhances digital video surveillance to make our cities smarter and safer. In a nutshell, IBM’s SVS solution consists of three layers of components. Layer one is video capture. Layer two is smart analytics. Layer three is database. Video capture is the first step and is done by IBM’s partners who provide cameras and video management systems (VMS). Video is first captured either by IP-based cameras in digital format or by analog cameras and then converted into digital format using an encoder, before it is distributed to the backend IBM Smart Analytics System for video processing. Video is streamed and analyzed by the objects, motions or events. Metadata is created for each picture frame and stored in a database. In case of law enforcement inquiries, both real-time and historical event data from the system can be searched from anywhere at anytime.

Sizing

Surveillance solutions are usually sized according to the number of cameras. Other factors also vary the sizing of the solution such as frame rate per second, number of moving objects, number of alerts desired, number of users interacting with the system and so on.

In this example, let’s just focus on the number of cameras for a simple illustration.

Case 1: Small solution – less than 10 cameras

A convenience store or train station usually has less than 10 cameras on the premises. This can be considered a small setup. Most partners’ VMS can be contained within one physical server for a small setup. The IBM Smart Surveillance Engine (SSE) processes each video frame and captures real-time events. One SSE server can support multiple cameras (30 or more). For a small setup, one SSE should be sufficient. Last, the SSE sends video metadata in Extensible Markup Language (XML) to a Middleware for Large Scale Surveillance (MILS) centralized data repository and instructs it how to index video. One MILS server, which is capable of more than 100 cameras, is sufficient. A storage adapter is needed to connect from the VMS server to the video storage, and from MILS to the metadata storage. With PureFlex, three physical compute nodes should be sufficient for a small surveillance solution:

1 PureFlex chassis

1 Flex System Manger (FSM)

1 VMS (deployed on a physical compute node)

1 SSE (deployed on a physical compute node)

1 MILS (deployed on a physical compute node)

Storage (either internal or external Storwize V7000 controller unit)

Case 2: Medium solution – 10 to 100 cameras

A school or a small department store that has 10 to 100 cameras can use a medium surveillance solution. Depending on the partner’s VMS scalability, we may need more than one compute node for VMS. Some partners’ licenses may support 60 cameras while others support up to 100. Each IBM SSE is capable of video streaming from multiple cameras. Three or four SSEs should be sufficient to support feeds from 100 cameras. Finally, a single node MILS is more than sufficient as it can handle more than 100 cameras. With PureFlex, five to seven compute nodes should be sufficient for a medium-size solution:

1 PureFlex chassis

1 FSM

1 or 2 VMS (each deployed on a physical compute node)

3 or 4 SSE (each deployed on a physical compute node)

1 MILS (deployed on a physical compute node)

Storage (either internal or external Storwize V7000 controller unit)

Case 3: Large solution – 100 to 400 cameras

Midsize casinos in Las Vegas usually have hundreds of cameras. More compute nodes are needed for each layer of infrastructure. For the video capture layer, some partners may support 60 to 100 cameras with one license per server; therefore, three or more compute nodes would be needed. Assuming each SSE is capable of handling 30 or 40 camera feeds, 10 or more compute nodes would be used. For the metadata layer, a three-node clustered MILS is recommended. Therefore, a large surveillance solution would need two PureFlex systems:

PureFlex chassis A:

1 FSM

3 or more VMS (each deployed on a physical compute node)

3-node clustered MILS (deployed the cluster with three compute nodes)

Storage (either internal or external Storwize V7000 controller unit)

 

PureFlex chassis B:

10 or more SSE (deploy each on a physical compute node)
94863 Dome Security Camera iStock_000015722816Large

Note that we only need one Flex System Manager (FSM) in chassis A, as a single FSM can manage components across multiple chassis. Also, chassis B that contains just the SSE would not need any storage components because SSE processes video streams to generate metadata of each video frame; it does not store anything. The metadata is passed to MILS for storage. For storage scalability, V7000 can be easily expanded to accommodate more data. If the capacity of the main V7000 controller unit is maxed out, up to four expansion units can be added to the same controller unit for more capacity.

This is an example to illustrate how to size an industry solution with PureFlex hardware. So far we have only calculated how much workload can be handled by each solution component; we didn’t need to calculate how many networking switches, fiber channel switches, power supply units and so forth are needed, or how they should be connected. The core infrastructure is already built into the PureFlex System and you don’t have to worry about whether the system will run or not when the chassis slots are maxed out. This makes the sizing a lot easier.

Each PureFlex System can hold two network switches, two storage switches, plenty of fan units and power supplies that are sufficient to power up all the components in the chassis. There are 14 slots in a PureFlex System. If you need an FSM in the chassis, it takes one slot and you have 13 slots left for compute nodes. You can choose half-width compute nodes (that occupy one slot each) or full-width compute nodes (that occupy two slots each). You can choose Intel-based or IBM Power-based compute nodes. For storage, you have a choice of an external V7000 or internal V7000 compute node that takes up four slots.

If your solution maxes out one chassis, you can use a second chassis and put as many compute nodes as you need. Then put a top-of-the-rack switch to connect between the two chassis. You can really treat each PureFlex System as a building block when it comes to sizing and scalability.

For more information about IBM Smart Vision Suite (SVS), please refer to the IBM Redbook “Smarter Cities Series: Understanding the IBM Approach to Public Safety.”

For more information about industry solutions with IBM PureFlex System, please visit our PureSystems  Center and select “Industries & solutions.”


Comments Off
Jenny Li

About Jenny Li

Jenny Li is a Senior Certified IT Architect and an IBM Master Inventor. Jenny held various technical positions in IBM STG, CHQ, and Research. She has a diverse technical background. Her area of expertise includes application architecture, enterprise architecture and strategy, and enterprise solution research. In her current role, Jenny is the Partner Enablement Lead on IBM PureFlex Systems at IBM STG ISV organization. Being the PureFlex technology advocate, Jenny promotes PureFlex technology to business partners, and works closely with cross teams including IBM Innovation Centers for partners’ engagement projects. Jenny has experience in complex solution architecture and integration, and software development. Jenny holds 29 US issued patents. In her spare time, she enjoys golf and creates patentable ideas