( ESNUG 576 Item 2 ) ---------------------------------------------- [09/22/17]
Subject: Dean Drako on IC Manage PeerCache P2P caching EDA tool accelerator
DAC'17 Troublemakers Panel in Austin, TX
Cooley: Dean. Last year, IC Manage, you announced this tool called
PeerCache P2P Caching and Shiv Sikand just did an update of it
right before this DAC in ESNUG 571 #5. It was all about EDA tool
speed up and blah blah blah.
How is IC Manage going to compete against rival filer systems?
I mean why buy something from a small guy like you when I can
get it from a big ass NetApp or Isilon or something like that?
Why should I buy it from you?
Dean: IC Manage PeerCache doesn't compete with the filers or the EMCs
or the Dells file systems or filers.
Cooley: How would it not? I don't understand. It's a speed up, it's a
workplace repopulation...
Dean: It's software, first of all -- it's not hardware. It accelerates
the access to files, access to data, access to unmanaged or
managed data by about 10x over what you'd get with the filer. It
does that by creating a peer-to-peer network of all the compute
and nodes in your system, then using those as caches.
We love the filers. The filers are great, they do a great job of
being the storage of record, of giving IT all the tools they
need. But unfortunately, when you're running your tools, your
tools are I/O wait state a little bit too much. I/O wait state
means that they're going slow.
What IC Manage does is eliminate all those bottlenecks using a
peer-to-peer file system. If you think about it, there's a big
change that Intel's pushing with the introduction of NVMe
storage. The distance between the storage system and the compute
is shrinking, because the NVMe is much more like memory.
Cooley: Right.
Dean: So, you get a lot lower latency and you can get a lot higher
performance out of it. Well that doesn't really work so well if
you're using NFS over this tiny little twisted pair of Ethernet
to a filer that's across the room or across the campus.
The latency in the I/O wait state that your EDA tools end up in,
grind them to a halt. You get maybe 100 megabytes of data that
you get from one of those. What we're seeing with our system is
about 600-700 kinds of megabytes of data feeding into the tools,
so the tool can basically not be in that wait state.
The great thing about it is it's a software only solution. It
works with any DM. It works with unmanaged data. So, you can
get significant tool speed up.
Cooley: Okay, Cool.
---- ---- ---- ---- ---- ---- ----
Related Articles:
Anirudh, Sawicki, Hogan go at it over the CDNS Pegasus DRC launch
Dean Drako on IC Manage PeerCache P2P caching EDA tool accelerator
Prakash sees 3X revenue in 2017 by crushing SNPS Atrenta Spyglass
Amit on Solido machine learning, lib characterization, and Siri
Join
Index
Next->Item
|
|
