Companies can save money on server load balancers (SLBs) if they know a thing or two about their software infrastructure before making the purchase, according to a report from Forrester Research Inc.
That’s one of the suggestions in the Cambridge, Mass.-based firm’s document, Hardware Server Load Balancing Update: Q1 2004, wherein Forrester’s analyst Stan Schatt outlines the ebb and flow of the industry, and advises enterprises on navigating this stormy sector.
In the report, Schatt points out that SLB vendors are at a crossroads. Thanks to advancements in load balancing technology, these devices take on aspects of security and data traffic management, which means companies like F5 Networks Inc., which sells hardware-based SLBs, must compete with the likes of Packeteer Inc. in the traffic management space, and NetScreen Technologies Inc. (recently purchased by Juniper Networks Inc.) in security.
Not to mention that the F5s of the world face an increasing threat from switch makers like Cisco Systems Inc., as companies like that integrate load-balancing functions into their network gear. Then there are software SLB vendors like Microsoft Corp. (the firm offers network load balancing) for hardware SLB vendors to contend with.
There are even different flavours of hardware SLB makers: some focus solely on load balancing, like Radware. Others aim for content acceleration and count load balancing as an added bonus of the prime functionality. NetScaler Inc. falls under this heading in the Forrester report.
Given the myriad technologies and choices available, how does the average enterprise figure out the right SLB for its needs? Schatt provided some pointers.
“The number one question that has to be asked first is: what applications do you really want to load balance, and how can you identify those applications? If you can do it at Layer 4 it’s a lot cheaper and a lot simpler than if you have to go up to Layer 7.”
Schatt explained that it’s relatively simple to figure out just where in the OSI stack an enterprise should aim. If you “can readily identify by well-known port number, by server IP address, by protocol, then you can get by with the very inexpensive Layer 4 load balancers. But if you’re doing things like thin clients — running various applications over Citrix — then all you would know from your Layer 4 information is there’s a thin client server. You wouldn’t know what applications are running on it.” That’s when a load balancer capable of recognizing information higher up in the OSI might work better.
Another suggestion: “If you need more than server load balancing and global load balancing, look for integrated solutions, rather than trying to string together several appliances,” reads the Forrester report, advocating that companies offload the headache of creating higher-function load balancing platforms to the professionals.
As well, “investigate how tightly the management software will integrate with your existing management platform.” And when it comes to application-aware load balancing, “clients will need to develop a comprehensive plan for tightly integrating their mission-critical Web applications with their SLB products. This might mean building an XML development team.”
History is to blame for the upheaval in today’s SLB market, according to the Forrester paper. Once upon a time, SLB vendors targeted hosting companies as startups in this sector tried to keep abreast with the burgeoning e-commerce industry. As dot-com firms increasingly sought server space in data centres, it behooved hosting companies to apply balancing technology, to reap efficient server use from their boxes.
Since the dot-com industry crumbled, however, SLB vendors have turned their attention to the enterprise space. That change led to other amendments as SLB providers tried to make their marks in the sector, which is why today we see F5 touting application intelligence as a prime selling feature of its Big-IP, while Nortel Networks’ Alteon portfolio offers speed and security, and Foundry Network Inc. talks up comprehensive, Layer 4 to 7 load balancing.
Jason Needham, product manager for F5 in Seattle, seems to recall well how his firm moved from server virtualization to encryption, deep packet inspection and finally application awareness with iControl, an API that links into offerings from firms like BEA Systems Inc. and Microsoft.
“It’s really about us going down this road of not only doing special things by looking at the application traffic, but also getting those vendors to integrate intelligence into their applications to inform our product about how things are running,” Needham said.
But there’s no question that one of the biggest challenges before F5 is keeping ahead of the competition. Schatt pointed out that the networking juggernaut Cisco offers load-balancing features in its Catalyst switches.
Schatt said the F5s of the world must continue their research and development efforts if they’re to stay in the game.
“They always have to be the first ones to offer new features. It’s tough for companies like that to keep pouring enough money into R&D to stay ahead of the Ciscos of the world, which every six months like clockwork come out with new, upgraded features.”
“There’s certainly pressure to innovate,” Needham said.
Alan Freedman, research manager, infrastructure hardware at IDC Canada Ltd. in Toronto, had one more suggestion for the enterprise trying to suss out the SLB space: keep Canadian support in mind. Make sure the SLB vendor has sufficient support in this country, lest the customer feels somewhat abandoned after the sales people have left the building.
Needham said F5 has that aspect covered.
“Our Canada coverage is solid and our teams have good results. We have an established direct sales force, channel partners, and system integrators and which sell into Canada.”
For more information about Forrester’s Hardware Server Load Balancing Update: Q1 2004, visit www.forrester.com.