SI
SI
discoversearch

We've detected that you're using an ad content blocking browser plug-in or feature. Ads provide a critical source of revenue to the continued operation of Silicon Investor.  We ask that you disable ad blocking while on Silicon Investor in the best interests of our community.  If you are not using an ad blocker but are still receiving this message, make sure your browser's tracking protection is set to the 'standard' level.
Technology Stocks : Intel Corporation (INTC)
INTC 34.10-1.8%10:32 AM EST

 Public ReplyPrvt ReplyMark as Last ReadFilePrevious 10Next 10PreviousNext  
To: Paul Engel who wrote (105053)6/29/2000 11:12:00 AM
From: Rob Young  Read Replies (1) of 186894
 
Paul,

"e elegance and power of this OS can't be underestimated.

However, you must ask yourself - what market forces are responsible for this OS (Galaxy/VMS) NOT garnering
significant market share over the years?

Maybe others have figured out that they can deploy a ton of "industry standard" servers far cheaper than the
ELEGANT solution, and come up with an acceptable performance.

Industry Standard solutions - high volumes - economies of scale - that's where the growth is."

Good questions and points. What was left out in my
description is that Galaxy was actually an outgrowth
of what was coming down the pike. The Alpha 8400
(max 14 CPU) was announced in April 1995 and engineering
in the following months surveyed customers to get a feel
for the next large box... feedback to Unix and VMS
engineering described what they could expect and they
set about modifying their OSes accordingly. By December
1996 VMS engineering was already describing Galaxy. Galaxy
really wasn't/isn't effective until this box came along
(the GS320 aka Wildfire). That was the target hardware.

This was a very productive time in VMS with many
new (patenable) ideas being put into action. Galaxy
is very innovative in that everything is under software
control and that is why a single CPU can leave an
instance and join another while all others move boards
at a time between domains.

But that is only a peak at one aspect. Much of what is
to come is very powerful. Mostly moving OS components
into shared memory gains great scaling. DLM and QIO Server
.. where DLM is the Lock Mananger and QIO Server is the
MSCP follow-on.. I/O is dispatched via or through shared
memory. All the ins and outs of what that means is yet
to be determined (i.e. we will know more when it ships,
also know exactly how it works when the Internals and
Data Structures .. IDS shows up explaining Galaxy internals).

Regarding performance and industry standard servers
comparing to something like this?

No way.

A RamDisk in shared memory in the next box (according to my modest calculations) will be capable of *at least*
20 GByte/second read and write speed... What am I
basing that on? The next box does 100 GByte/sec
aggregate memory bandwidth... but that is always
best case scenario... I'm picking a modest 20% of that.
It may be more it may be less... but you get the
idea.

When it comes to I/O in the near future, you haven't
seen anything yet.

Rob
Report TOU ViolationShare This Post
 Public ReplyPrvt ReplyMark as Last ReadFilePrevious 10Next 10PreviousNext