Joanna Roman schrieb: > > --- Damian Menscher <[EMAIL PROTECTED]> wrote: > > >>On Sun, 5 Jun 2005, Joanna Roman wrote: >> >>>--- Niek <[EMAIL PROTECTED]> wrote: >>> >>>>On 6/5/2005 5:22 AM +0200, Joanna Roman wrote: >>>> >>>>>I am just wondering how feasible it is to do AV >> >>hw acceleration in >> >>>>>general. Besides using faster CPU and faster >> >>memory, ASIC can't >> >>>>>really help. Can anybody shed some light ? >>>> >>>>You answered your own question. >>>> >>> >>>I am just throwing out this question hoping >> >>someone >> >>>might have some novel idea of accelerating AV in >> >>HW. >> >>>Does anybody have any idea how company like >> >>Sensory >> >>>Networks does HW acceleration then. I have real >> >>doubt >> >>>about the result. I wonder how much gain they can >>>achieve. >> >>Disclaimer: I'm making all of this up. >> >>The purpose of the "accelerator" isn't necessarily >>to make it go faster, >>but rather to offload some work from the main CPU. >>(Otherwise, you >>could just buy more CPUs and load-balance the work.) >> That way you save >>the expensive part (the CPU) for stuff it's good at, >>and use the cheaper >>part (the accelerator) for specialty work. >> >>For the case of virus scanning, there is NO >>floating-point arithmetic >>involved. So about half the transistors in your >>typical CPU are going >>to waste. Now imagine if you could simply produce a >>P4 minus the >>floating point units for half the cost, and get the >>same virus-scanning >>speed. Now imagine cutting out the cost of the >>motherboard with all its >>useless goodies like video and sound, and just >>having a raw interface to >>the accelerator (and its local ram). Sounds like a >>good deal to me, >>though obviously it'd require some serious effort to >>get the first >>prototype to work. >> >>Someone tell me if I got any of this anywhere close >>to being right. ;) >> >>Damian Menscher >>-- >>-=#| Physics Grad Student & SysAdmin @ U Illinois >>Urbana-Champaign |#=- >>-=#| 488 LLP, 1110 W. Green St, Urbana, IL 61801 >>Ofc:(217)333-0038 |#=- >>-=#| 4602 Beckman, VMIL/MS, Imaging Technology >>Group:(217)244-3074 |#=- >>-=#| <[EMAIL PROTECTED]> www.uiuc.edu/~menscher/ >>Fax:(217)333-9819 |#=- >>-=#| The above opinions are not necessarily those of >>my employers. |#=- >>_______________________________________________ >>http://lurker.clamav.net/list/clamav-users.html >> > > > With muti core CPU coming out, I just dont feel the > work of building a special AV accelerating chip > justify the need. At least from the price/performance > ratio perspective, it is not worth it. Not to mention > the time to market. If you use off the shelf PC, you > automatically gets 2X performance every 18 to 24 > months. So why bother ?
i) an AV accelerator should be far superior (if designed well) -- you can cast several nice things into this ASIC (or another form) like specialised (un)packing sections supporting main algorithms, etc. ii) furthermore, it can be many times fast computing any amount of stuff to test because one could spend it an incredibly fast memory subsystem. iii) multicores are there for years already (IBM POWER, UltraSPARC IV, HP-PA, MIPS, only to name a few); the x86 etc. stuff is -- compared to them -- badly designed and will give a not-so-good boost; at least worse than most of the people think. iv) as with everything in IT -- one needs a Tezro, another one is satisfied with a PeeCee off the shelf -- YMMV. cheers, -- Timo Schoeler | http://macfinity.net/~tis | [EMAIL PROTECTED] //macfinity -- finest IT services | http://macfinity.net Key fingerprint = F844 51BE C22C F6BD 1196 90B2 EF68 C851 6E12 2D8A There are 10 types of people in the world. Those who understand binary and those who don't. _______________________________________________ http://lurker.clamav.net/list/clamav-users.html
