Speakers: Mehdi Daoudi, Co-Founder and CEO, Catchpoint Systems; Bill Kish, CEO, Coyote Point; Peter Melerud, Co-Founder and COO, KEMP Technologies

Moderator: Bojan Simic, President and Principal Analyst, TRAC Research

Bojan Simic: Our topic today is how organizations in the small to medium size market sector manage the performance of their web applications. Before we get into the specifics of how they do that from a technology and processes perspective, one question that I have for all of our guest speakers is how they actually go about defining this sector. As you are probably aware, in some other IT markets companies go about defining their market sectors based on number of users, number of employees or revenue size. Some of these metrics are more specific in web performance management, so if you folks wouldn't mind sharing how you go about defining what falls into that SMB category for your organizations.

Bill Kish: It's interesting. There are a number of different ways of categorizing the size of a company. One of the typical ways of sizing a company is by the number of seats that you have. On the internet, the number of seats has no relationship with the number of chairs in your office. You know, you've got potentially, thousands, tens of thousands or even hundreds of thousands of visitors to your site, and that's really the way we size businesses and try to determine which products are best for them. At Coyote Point, we look at the market that we serve more as the mid-market than the classic SMB. I think of the SMB as smaller shops that maybe have 10 employees or something of the like and that potentially is a market that serves tens of thousands of users. Think about Craigslist, for example, where a very small company was one of the major sites on the internet. But I think that, for our company anyway, it is the amount of clients that you are serving and the amount of bandwith and the amount of page views that you are serving them; that is the critical factor.

Peter Melerud: In our space as well, we also look at a lot of the same criteria that Bill mentioned. From a customer standpoint, sizing wise, we also look at the number of servers they may have, for example. Rather than look at the number of PC's or desktops, we look at the number of servers they have, how many applications, what type of applications that they are using, whether they are web facing or internet based and certainly throughput and bandwidth come into play. Again, from the standpoint of the kind of market we serve, we serve anyone from 2 servers and up from there, which is not necessarily a classic SMB model, but both on the smaller side. We definitely focus on customers that have just added their second server and grow from there from tens, hundreds and sometimes thousands.

Mehdi Daoudi: I agree with Bill and Peter. The definition for SMB for us is that we provide a service, so it's a little bit different than selling hardware/equipment but when we talk to companies in different segments we tend to define a small and medium business based on their needs and budgets. For example, we might have a Fortune 100 that is just getting into the web performance business and they might just want to spend a few hundred dollars a month, so that would "put them in the SMB" category versus a company willing to spend tens of thousands of dollars on Web performance monitoring. So, it's really based on their needs and budget.

Bojan Simic: One thing that we are seeing in our research about SMB companies is many of them understand how important it is to manage their websites performance. We had companies that fall into that category that generate $20 million a year just through business that comes through their websites, but also if you look at the capabilities that a lot of these organizations have, many of them that heavily rely on their websites don't even have basic capabilities for managing web performance. Typically the pushback is "We are a small company and can't afford that". Just to put things into perspective, what type of investment and what capabilities are these organizations looking to if they are trying to optimize the performance of their websites?

Bill Kish: I think that anyone who is generating 20 million dollars through their website and isn't deploying some kind of application delivery enhancement technology and performance monitoring is crazy. It's very cheap insurance for that much revenue. You asked to characterize how much an application delivery control or technology would cost to deploy and we are talking about, in the market that we serve, anywhere between $3,000 to $12,000 for the basic functionality. What you get out of that is reliability. You eliminate single points of failure across the entire application stack. There have been plenty of studies that have been done to show, for an ecommerce site that generates so much per day, not only does how much a minute of downtime cost, but how many visitors. The customers will go away for every additional second it costs to load a page or for every additional second of latency that it takes to process an order. These are really critical things that our users, whether they be businesses deploying internet applications or websites, really need to take into account. Reliability is the first and foremost thing and the types of technology that we are talking about are very cheap insurance for the level of value that these technologies provide to the users that deploy them.

Peter Melerud: Part of the question was about metrics and what can businesses look at when deciding when or if to deploy these types of technologies and certainly the first metric I'd suggest is to look at your downtime, look at the past month, past year, past quarter and see how much downtime the business has experienced as a result of the site being down due to server or application problems. Once you start plugging those metrics and numbers and loss of revenue into the equation, then the cost of buying these technologies becomes almost a non issue. Metrics relative to scalability and performance are easily obtainable these days, and businesses that are let alone doing $20 million in revenue or are not even doing 6 figures yet, it can be quite painful for them not to have a solution like that in place, because that only means a significantly higher loss in revenues to them. 

Mehdi Daoudi: Any company that generates any amount of money online and depends heavily on their online traffic, even if it's a hundred thousand dollars a year, need to buy insurance. Whether it's a technology the ensures reliability or monitoring, I'd look at them like an insurance policy. You don't go buy a million dollar house and not put an alarm on it. Any company needs to put a strategy in place, but the first thing they need to realize is that they need to invest a person. They need to wake up and say performance is important, my reliablility is important, my uptime is important, so they need to at first, make that human investment. After that, it's easy to come up with a plan to dedicate financial resources to implement solutions. There are so many studies where Amazon downtime cost them $30,000 per minute, where Goodge added 400 miliseconds to their site which created a huge drop in terms of search results. One of the things we're noticing is that more and more small companies are coming to that realization. They are waking up and coming to us looking for monitoring services because Google is ranking them lower because their page loads 20 or 30 seconds. The other important factor that we are noticing is that downtime is not as important these days. There is redundancy, people use Amazon, cloud servers, etc., but the biggest challenge that I find with some of the companies that we talk to is performance. They do not understand the concept of how performance is driving away some of their business. We blogged about it a few months ago, but performance is the silent killer, it is the silent uptime or downtime, and people are starting to realize that.

Bojan Simic: Looking back at some of the survey data that we have and looking at the numbers of availability versus response times especially in the area were folks are using their websites to generate revenue, a lot of people in that area have figured out how to deal with availability issues. Looking at the survey data, the average was just over 99% availability per year, but if you look at the issues with response times, even though they have less of a business impact per incident, they are much more likely to happen and they happen 10 times more often than issues with availability. If you do the math and see how much organizations are losing per minute of downtime versus per minute of slowdowns and multiply that over the course of a year, there is much more being lost because of issues with response times. Calculating the ROI with solutions like that, a lot of organizations do not have visibility into what is that business impact. Bill mentioned earlier, investment for basic capabilities, depending on the size, can be anywhere from $3,000-12,000. If that is my investment that I am looking at, am I losing enough to be able to justify that investment, and how do I actually go about calculating that? Thinking about your customers in the SMB space, how often is it that they have some type of internal process in place and say OK, this is approximately how much we can lose if these things don't work well and this is the investment that we are looking to make to ensure that doesn't happen. Do you see your customers in the SMB space actually going through that process?

Bill Kish: Our customers tend to be reactive, rather than proactive. In other words, they've just incurred a situation where either an application was down for an unacceptable period of time, and somebody woke up and said we have to do something about this. As a result, they are out researching solutions. I think that it's not too difficult to make the decision about whether you want to spend a couple of thousands of dollars for a company where your productivity really depends so heavily on these web-based or client-server based applications. At Coyote Point, we have a number of internal web facing applications for doing CRM or for doing our order tracking and management in our production lines and those applications are really mission critical and every time we are able to speed them up, even by a few seconds, by deploying new front-end technologies, or by deploying different enhancements on the servers, you can tell there is a market improvement in efficiency and even moral among the people that have to use those technologies. So, it's not always a hard dollars and cents. A lot of times, it's just about the feeling of whether or not your business is flowing smoothly or not. The same also goes to the outside users who are on the internet, visiting a content site, visiting an ecommerce site, or what have you. If that feels like its not a smooth experience, they have a lower level of confidence in your company. What is that worth? It's worth a lot.

Bojan Simic: Your companies have a strong presence in the SMB market, but you also have a number of large customers across many industries. If you wouldn't mind sharing with us some key differences as they go through the selection process evaluating any solution, how is the selection process of SMB companies different from that of large enterprises?

Peter Melerud: One of the key differences is in the focus on the ease of use on the initial CAPEX and OPEX expenditures. For the SMB, they usually don't have the expertise in house to sometimes handle the sort of brand that the enterprise folks look into. So for them, having a part of this is very simple, easy to deploy, easy to manage and certainly the cost side of it becomes a very important criteria. Lately, we've seen some of the higher-end customers, customers that are traditionally classified in the enterprise space, who are starting to look at the same type of criteria more closely. They will question whether or not they need certain niche functionalities that are built into some other products. For example, do I really need Layer 7 functionality in the ERP system that I don't have today? And do I need to spend 5, 10, 15 times more on that? That idea seems to be increasing, that type of selection criteria is becoming more prevalent, not just in the SMB space, but in the enterprise space as well.

Mehdi Daoudi: What we are seeing, at least on the smaller side of our customers, is that they are looking for simplicity. They want a tool, they want a service that is simple to use because they don't have the same manpower that larger companies have. They want simplicity without losing too much functionality. They want to be able to do what the big guys are able to do, but without having very advanced users on the payroll. The other thing that we see, that is very critical for the small business, is service. They want to have the ability to have someone help them, not only implement, but also drive the implementations. Say this is one way of doing it, but based on our knowledge and history, these are some things that you can do better. So they are very thirsty for knowledge, and actually compared to bigger companies, the small and medium businesses are much more flexible, so you can work with them in a more efficient way to implement your solution and they get more bang for the buck.

Bill Kish: I agree with you as well, particularly on the flexibility end of things. Smaller companies tend to be much more willing to experiment to try different technologies and different types of solutions to solve their problems, whereas larger enterprises typically go after an approach where they analyze the feature set of the market leader, probably the most expensive leader in the market, build a checklist of features that they would like to have based on the most advanced features and then go out for bid with all of those features. The sad part about this is that we all seem to realize that it leads to environments where these larger companies have to throw a large amount of human resources at managing these specific technologies. One of our customers actually had somebody on their staff whose title is load balancing architect and does nothing but worry about how to deploy load balancers within their network. So, that is something that an SMB or a midsize company definitely can't afford to do. They'd like this to be a solution to a problem, not a problem in itself and not a challenge in itself.

Bojan Simic: Going back to some of the things that were mentioned earlier, Bill mentioned some of the pricing points for basic capabilities and Peter mentioned, using the example of Layer 7, that maybe that is a capability that some SMB companies can't afford. Just to clarify these things from a technology perspective, if you are an SMB company and you are trying to gain some of that insurance, what are some of the top 2 or 3 technology features that are "must haves" for them?

Mehdi Daoudi: One thing we build at Catchpoint is a lot of flexibility just to be able to handle their needs and not penalize them from a pricing prospective, though they can chose and pay for what they really need. But at least from a monitoring prospective, the things that we highly recommend to them are the concept of uptime, so you at least you know if you are up or down. Then we try to encourage them as much as possible to be able to understand their performance, which tends to get a little bit more expensive, obviously. But these are the 2 things we try to push them for and then obviously you have all kinds of other solutions, like being able to understand the impact of third parties on your website or your DNS infrastructure on your website. So all of those require different types of modules, but we do not push them, but we highly recommend that at least you know whether you are up or down and how fast you are because those are the 2 key metrics as a company that you need to have.

Peter Melerud: From a standpoint of basic features that an SMB would need, I'd like to say that it varies a bit among customers depending on the applications that they are deploying and running but if you were to bring it to a higher level, higher availability certainly is at the top end of the requirements; making sure that the appliances that they deploy are able to sustain the kind of bandwith and throughput and provide the kind of high availability that the infrastructure is requiring. More specifically, probably a lot of sites these days, especially the ecommerce, are depending on SSL encryption/decryption. Certainly products that can off-load or accelerate applications that are behind SSL would be key. There is a number of other technologies that ADC's deploy that SMB's could certainly leverage. Caching and compression are things that come into play and could certainly help quite a bit in accelerating performance of their sites.

Bill Kish: It's interesting how the combination or the overlap of the types of work that each of our companies do actually impact the way in which you deploy application delivery controllers and load balancers and things like that. Monitoring is what Catchpoint does and it's also very critical to the application delivery controller. If its one important feature set that an application delivery controller should have, it is a good monitoring infrastructure that allows it to determine, not only when a given application incident is available on a server or a given server is available, but also try to figure out when a given server is overloaded so that it can back traffic off from that server. By doing that monitoring and then that proactive adjustment, servers no longer reach the point when they are going to crash and your option is to work around them until they come back up. You can actually prevent the server from reaching an overload condition. So, I think that those sorts of technologies are very important and all of the other types of add-on features that turn a load balancer into an applications delivery controller become important, depending upon the type of business and types of applications that you are delivering. As Peter mentioned, SSL acceleration is very important, it has a high ROI because it can be very expensive to deploy enough servers to handle all of the SSL traffic that you need. Whereas if you have hardware acceleration on your applications delivery controller you can deploy less expensive and fewer servers. Also, the compression technology that Peter mentioned can allow you reduce the amount of bandwith that you need to purchase from your ISP and, as a result, you can reduce the amount of expenditure that you have for that kind of infrastructure and that can pay for itself very quickly. So, all of these sorts of things are things that companies can look at and consult with vendors and try and figure out which feature sets and which products can provide best end results for them, whether their goal is improving reliability, reducing overall capital expenditures or recurring expenditures.

Bojan Simic: One of the things that was mentioned earlier in terms of capabilities and selection criteria that SMBs care more about is flexibility. That is pretty consistent with what we are seeing. When I say flexibility, in many cases it is technology related, but it is also related to the way the organizations do their business and plan their business, in terms of whether their web performance management solutions scale up or down, actually follow the way their business is either growing or going in the opposite direction. Some of the concerns that we are hearing is that "as my company grows, and I bought something that was designed as an SMB solution, do I have to rip and replace that, or is there actually something else that I can do." From a technology perspective and also from a pricing perspective, how can organizations have their solutions kind of grow with them, if you will. How are you all addressing that and how do your customers go about doing these things?

Peter Melerud: One of the key issues is starting from the standpoint of whether or not the solution being deployed is hardware, virtual appliance or whether it is a service, in this case. With respect to hardware platforms, certainly there are some limitations, as far as head room and performance levels, that a particular platform hardware can support. From that respect, there are definitely different ways that both vendors and customers can look at what type of solutions they need. On the virtual appliance side, there is probably a lot more things that can be turned on and off dynamically where customers can dial in the amount of performance they need out of their solutions than out of the hardware platforms, but it really comes down to looking proactively both at the feature set and performance levels. These are two sometimes independent areas to look at. Does the product or solution that I'm looking at have enough feature content that, as our site or application develops in growth, it could support? And the second part of that is, is there performance limitations, and what are the performance limitations, in particular, a hardware solution, and how do we manage those limitations? Do we purchase a model that supports us 1 year, 2 years downstream? So, in that respect, that is where vendors can come into play and really advise the SMBs on various capacities, on both the hardware side and the virtual appliance side, and can help them look forward to that. Different vendors have different stratification models and in terms of how they go about pricing the various elements of their solutions. My best advice would be to engage with that vendor and look at the different options that are on the table and take it from there.

Bill Kish: An important thing for a potential buyer of this type of technology to do before they actually start shopping is to understand what exactly the traffic on their site looks like right now and project it into the future. The hardware devices that we are talking about are really tremendously scalable. Even in our mid range product, we can handle over 50,000 transactions per second of what we call Layer 7 traffic and significantly more at Layer 4. There really aren't that many sites that need that level of performance and you can certainly scale up quite a bit more in the higher end product lines. So, to me, the challenge that we see is when a customer or a potential customer calls us up and says "Help me size a solution" and we say "tell me more about your traffic patterns and characteristics", and if they don't have the answer, it is very hard to help them out. As I said, the products themselves are really scalable and you could probably get into one of the midrange products and be pretty comfortable, unless you are a very busy site, growing orders of magnitude a year, you will be fine for quite a period of time. Virtualization is sort of a solution in a lot of people's minds to this question of flexibility and Coyote Point looks at this primarily from the standpoint of the cloud as it is growing, the cloud service providers and hosting companies that are going to be behind this migration out of the private data center and into the cloud. Those are the people who really need tremendous levels of flexibility. Whether that is handled by allowing you to subdivide hardware and share it among multiple clients, or to be able to spin out virtual instances of an application delivery controller on demand, those are the people who really are going to need that kind of capability. I think that the hardware solutions that we have today, the plug and play boxes, are really the best bet for most of the SMBs and midmarket companies, just because of the simplicity involved. The fewer decisions that you have to make in terms of trying to figure out how you are going to deploy them in a virtual environment, what the interaction then becomes between the ADC instance and the application server instances that are running on the same physical hardware, that adds lots and lots of layers of complexity.

Mehdi Daoudi: On the Catchpoint hand, we provide a service, so it is a little bit different. We have a pricing mechanism and packages that literally allow companies to grow and they start at $99 a month and they can go and spend thousands more a month, so, it is very flexible depending on what they want to do from the monitoring perspective. I am going to agree with Peter, where the biggest challenge, whether they are looking to buy a load balancing technology or something else, the major challenge is not knowing what they actually need and that it is an amazing discovery process for both us and the customer, the fact that there is not much knowledge about what the infrastructure is capable of, what their current needs are, etc.

Bojan Simic: Also, as we are getting closer to the end of the year, it would be interesting to look back and see some of the key trends in the Web Performance Management market, especially in the SMB. If you wouldn't mind spending a few minutes summarizing some of the changes that happened in 2010, in terms of level of interest for management solutions in this market segment or different types of requirements, anything that jumps out at you that happened in the last 12 months or so?

Mehdi Daoudi: One of the things that we have noticed, at least on the web performance management side, I think it will be a big thing in 2011. Already in 2010 you have a lot more democratization of web performance monitoring tools, of folks that are more and more interested. I was at a meet up in Los Angeles, where we had 20-30 people that came to hear about web performance, whether it is optimization or monitoring and these are people that 2 or 3 years ago wouldn't be sitting at the table because the company didn't care or didn't have a budget for these types of solutions. I think Web performance is going to play a very important part in 2011. Performance itself is more and more becoming a differentiator. You can go and shop on site XYZ rather than on site ABC, because with site XYZ you will always be able to get in and get out in 3 seconds without having to wait for pages to load and checkouts, etc. Performance monitoring is going to be important. The other thing that we are noticing is that more and more SMB companies are getting out of their boundaries of their comfort zones. They are looking to buy CDN, they are looking to buy load balancing, they are looking to get into web optimization and performance is getting its place as being the fifth P in the marketing mix at this point because it is a differentiator. So companies of full size are starting to realize that and I think that they are going to spend money in 2011, either improving their service through better data center infrastructure, load balancing or just in monitoring in general.

Bill Kish: You asked about what the trends are that we were seeing this year, and it's interesting, the trend that people are talking about is the cloud. Is anybody actually deploying it, not really right now. The cloud is a great development tool. It is where virtualization was about 2 years ago. People are experimenting with it, they are using it for development platforms and it is really a great tool for start ups, but it is not really something that people are talking to us about right now, in terms of what affects them on a day to day basis. The trends that I am seeing, which is very interesting, is the deployment and the vast expansion in the number of mobile devices that are out there accessing our customer's sites. These devices have very specific applications and network traffic patterns, which is somewhat different than desktop devices just because of the networks that they have to traverse. We are starting to work on some technology which is going to provide some interesting boosts to people trying to deliver applications to mobile devices. AJAX technology is another thing which is finally coming into the forefront. AJAX technology requires that a browser requests lots and lots of small objects very frequently as opposed to a few very large objects. It makes for a very rich user experience, but it really changes the patterns by which servers and the applications that are running on them are accessed. So we have been tweaking the way that we handle traffic for AJAX to enhance that. 2010 was a very interesting year and 2011 should be an even more interesting year.

Peter Melerud: I agree with both Bill and Mehdi about the technology aspects of what's happening in 2010-11, but what we have noticed, not just in 2010, but in previous years also, there has been a slow but sure growth and awareness among even the SMBs about the need for having a solution that addresses web optimization, high availability, and the delivery and enhancements that these types of solutions can provide. We are seeing more conversions between the learning curve of the SMBs and about the fact that they need something like this, and also on the enterprise side as well and some of the larger customers who are seeing this technology more pragmatically. They are looking at it from more of a core aspect of load balancing and ADC, what can it do to really improve the traffic flow, the responsiveness to my users. They look at it in terms specific to the applications that they are running and we are glad to see that there is that trend and we will probably see that increase dramatically in 2011 as well.

To listen to the podcast click here

To listen to individual sections of the podcast:

Section 1: Defining the SMB market
Section 2: ROI of Web performance management solutions
Section 3: Key differences between SMB and large enterprises in managing Web performance
Section 4: Key capabilities for SMB
Section 5: The importance of management flexibility
Section 6: Key trends in 2010
Section 7: Overviews of Catchpoint Systems, Coyote Point and KEMP Technologies