Technology leadership

Steve Santini is Chief Technology Officer within Bank of America Securities. Responsibilities include managing the datacenters that support trading; the installation and maintenance of the equipment; and the proactive support of the development organizations. Typically, the organization has been much more geared around operations and incident management; Santini is trying to enhance that to include support of the development areas.

FST. How is technology enabling you to achieve your aims at Bank of America? What technology strategies are you employing in order to help meet your organizational goals?
SS. Recently we have been implementing the ITIL processes geared around operations, and in doing this we are utilizing various pieces of technology – workflow management, in particular, is a key thing for us right now. The objective that I have is to reduce the variabilities in producing my product or making a service available – it could take one day or it could 100 days, and I’m trying to reduce that variability down by automating the workflow. So, rather than having people looking at e-mails to figure out what they’re doing next, they’ll be following a workflow system – logging in to the system and taking their work orders from the queue. When they finish their part of the job it automatically goes off to the next person.

There are some challenges inherent in this. First, you need to evolve into it; it’s not just something you can come in and say “this is where we should be operating”. We have to start at a very high level, with an understanding of the tasks we want to automate, and then gradually filter that down through the organization and capture the workflows at a more granular level each step of the way.

FST. You were previously a user of the infrastructure organization; now you head that group. This must give you a unique perspective?
SS. I suppose I am something of a poacher-turned-gamekeeper. Previously as a user of the infrastructure I was puzzled by certain aspects that I felt were not helpful or that didn’t seem to make sense; moving to this new role has helped me understand a lot more about why the infrastructure is the way it is, and I’m hoping that I’ve got the willingness, the desire and (as a result of my previous experience) the knowledge to change that in order to provide a better, more intuitive experience for the users.

FST. So how do you align your own division and IT goals with the wider group’s objectives?
SS. I work for an organization called network computing group, which is organized into three core services: IM global trading infrastructure, global enterprise access and, the largest part, desktop support, enterprise and operations. So I’m one of three core service providers.

There is communication between these three division about what we find works, what doesn’t and what we should be doing consistently. It’s something we’ve talked much about recently, for example, with relation to business intelligence. There are certain things you must have in place before you can have realize effective business intelligence – you need a good understanding of what machines you have in place, how they’re configured and what is running on them. Then you need common processes for how we handle these things. We’re looking to achieve this right now.

Ideas and best practices also get seeded in from other areas of the organization. You find best practices as well as elements you don’t like so much. The most important things is to work from a model of success. It’s not enough simply to claim territory without performance; you have to claim a performing model for the developers.

FST. Data center optimization is a big focus right now for many organizations...
SS. When it comes to data centers today, power is the most important thing, rather than space. When machines are packed in densely, it will be the amount of power available that is limiting, not the space it takes to do that. Our intelligence shows that increasing numbers of organizations are now creating mega-datacenters, and there are significant investments being made into datacenters because everyone’s running out of power.

The problem for us at Bank of America is that because we’ve undergone so much growth in recent years (with a significant proportion of that money going into technology), the number of machines has gone up dramatically. This has created challenges on many levels for me, first amongst them being how to administer a data center. A data center is a living organism and there are certain ratios that must be kept in check to ensure it runs properly. The infrastructure servers must be robust enough to support the application servers.

Another challenge in optimizing the data center is latency. There are three kinds of latency that we use – server to user, server to storage and server to exchange. Depending on the type of system, these different latencies will impact on how and where your data centers are.

As chips are becoming more powerful and efficient, it is outpacing the rate at which we can deploy machines. The only solution I can find is to offer options to my management – for example, while I may not need 10MW in my data center now, I want the option to have that at a later stage.

FST. Some companies are seeing virtualization as a means to optimization? What does it mean to you?
SS. It’s an oft-misused term. In one extreme I’ve heard it called ‘hardware separation’, whereby you electronically separate one machine into separate machines so that you have your own little environment. That has its merits but it’s not a virtual environment – it’s just partitioning. The other extreme is where you can truly ‘snap’ an environment, freeze it, save it and bring it back. Those are virtual environments in the true sense of the word. Then of course there are various definitions and interpretations of virtualization in between.

The low-hanging fruit that people commonly refer to as virtualization is shared machinery. We can now partition a machine in a way that, while it is shared, it still appears as a dedicated machine with the users guaranteed a certain amount of resource. This enables a much more consistent and predictable performance by the user and is, ultimately, a cost saver.