FAQ
1. Will NewServe speed up the clients who are pulling data from the server?
NewServe will speed up the server, and that will speed up any clients who were waiting for data from the server. Clients who are filling a buffer as fast as they can, for example, will see that buffer fill faster when NewServe is in the server.
But most clients ask for data at a fixed rate, once initial buffers are filled. A client playing a high-def video, for example, will pull data from the server at at an average of, say, 6 mbps. Making the server faster, in this case, won't improve the user experience, other than letting the initial buffers fill faster, which will reduce the initial delay before the video starts. The client will still only pull data at an average rate of 6mbps and they were able to do that before NewServe was added, so once the video starts they won't notice a difference with NewServe.
The main thing that changes when NewServe is added is that more of these 6 mbps clients can attach to the server and play their video before the server drops to its knees and saturates.
For example, consider a server without NewServe. 100 of these clients may be able to get their 6 mbps from the server, even though the disks are running at near 100% utilzation. But if another 50 clients are added, the disks simply can't get around to everyone's request in time and all 150 clients see their their video stop when their local buffers run dry.
With NewServe, the server is able to keep the buffers from 168 clients always filled in time so that no one's video stops. The bottom line for the owner of the server is that they're able to collect more revenue for that server.
2. Does NewServe require any changes at the clients?
No. Nothing needs to be changed at the clients.
3. How can I see the effect that NewServe has on my server?
You can see the effect of NewServe immediately by turning on the operating system's Performance Monitor and watching Disk Queue Length and Disk % Idle Time. For any given load these should improve dramatically when NewServe is added, leaving no doubt that NewServe has given your server more head room (a "bigger throat").
4. What if Microsoft changes their operating system?
Microsoft has published the interface to various levels of their IO and Network stacks to allow developers to innovate on the operating system to improve the user experience. NewServe uses only those published interfaces. By sitting behind these interfaces NewServe is isolated from changes that Microsoft makes to their proprietary code.
5. Is NewServe safe to put in my server?
Yes. NewServe has no write function so it can't change anything on the server disks.
6. Which versions of Windows Server does NewServe run on?
NewServe works with all 32-bit and 64-bit versions of Windows Server 2003 through Windows Server 2008 R2.
7. How much memory does NewServe use?
In x86 servers NewServe uses 800 MBytes of memory. In x64 servers NewServe will use up to 2 GBytes of memory, depending on how much memory is in the system.
Most Windows servers today can be fitted with 64 GBytes of memory, so the memory used by NewServe is usually negligible. If a server has only minimal memory, another 1 GByte DIMM can be added for about $40. NewServe will find the memory at bootup and comandeer what it needs.
8. How much of the server CPU does NewServe use?
NewServe uses a negligible amount of CPU power, on the order of 2-3%.
9. How is NewServe different than a big cache?
NewServe is not a cache at all. It doesn’t cache anything.
Caching keeps accessed data around in the hope that it will be requested again and a new access for the data can be saved. Caching looks backward.
NewServe looks forward. It tries to predict what will be requested and gets it first. Then after the data has been read by clients NewServe gets rid of it as fast as it can to free up that space.
Working from inside the operating system, NewServe does a pretty good job of predicting where the requests will hit next on the disks. It has the intelligence and the memory to keep track of hundreds of separate client processes that are sending requests to the server. It acts like a first class airline stewardess who knows the preferences of every one of her passengers and tries to predict their requests.
