22,188 reputation
12450
bio website ripplelabs.com
location Oakland, California
age 45
visits member for 3 years, 5 months
seen 4 hours ago

I am Chief Cryptographer at Ripple Labs and one of the architects of the Ripple payment system.


Oct
14
comment Load Balancing over multiple ISP connections
What do you mean by "a connection"? Do you mean a TCP connection?
Oct
9
comment How can I host multiple websites from the same Windows 2008R2 server?
Note that "upstream" can be a VM hosted by the Windows machine.
Oct
9
comment How can I host multiple websites from the same Windows 2008R2 server?
The rule you state in your last paragraph is the wrong rule. For one thing, how would it work if a packet is not a reply? What you mean is this: For an outbound packet whose source address is bound to interface A and whose destination address is not covered by a more specific route, send the packet using the default route whose next hop is reached through interface A. (And the same for interface B.)
Oct
6
comment Failed to ping through gateway which is not default gateway
You do understand that ping -I and traceroute -i do completely different things, right?
Sep
30
comment Static routing on CentOS through managed switch
You're trying to solve the wrong problem. The problem is not routing these packets to the Internet, that's easy. The problem is getting the replies back.
Sep
29
comment “Lock” a certain user to a certain user in SSH (Cygwin)?
What's the purpose of the user having SSH access? What is he supposed to be doing?
Sep
27
comment Service port accessible only by one network card
What error does the service return?
Sep
24
comment How can I have ntpdate ignore skew?
You have to bring it up to date. Modify the $cmd line to what is appropriate for your platform.
Sep
24
comment How can I have ntpdate ignore skew?
Yes. Start ntp and then wait for it to report that it is synchronized. There used to be a tool called ntp-wait for this. It was a fairly trivial perl script.
Sep
24
comment How can I have ntpdate ignore skew?
There hasn't been a good reason to do this for many years now. The ntp service handles this case already. (You may want to specify appropriate options, such as -g.)
Sep
22
comment How do we find out what is causing high cache usage on our server?
You need to understand why the machine is dying, not why it is using memory. Using memory is normal, dying is not. Can you describe the problem a bit better? When it "dies", what exactly happens? Does it crash? Reboot? Respond to pings? Log something?
Sep
17
comment High memory usage, where is it coming from?
"Why is my server wasting so little memory?" (There are only two things you can do with RAM, use it or waste it. You can't save it for later. So "lots of use" means "little waste".)
Sep
11
comment tcp ack delay almost 600ms, is this all right?
You seem to be expecting some unreasonable level of precision from TCP timers. It's not intended to be a real-time, low-latency service.
Sep
11
comment tcp ack delay almost 600ms, is this all right?
It looks like a packet was lost and retransmitted. The recipient didn't know the packet was lost and retransmitted, and so started a 200ms timer when it received the retransmit. A 500ms timeout over TCP is idiotic.
Sep
11
comment tcp ack delay almost 600ms, is this all right?
What client timeout are you talking about?
Sep
9
comment RAID6 generates quite a IO wait at relatively low TPS
For a given level of performance, the wait time is purely determined by load. You seem to be getting reasonable performance -- 200 TPS. (The heads have to seek.)
Sep
1
comment HA router with keepalived showing unexpected traceroute
Yes, it's normal. When the device originates packets itself, it won't use the floating address.
Sep
1
comment HA router with keepalived showing unexpected traceroute
Nothing unusual about the traceroute. The machine wisely uses its native address as the source address for packets it generates (in this case, the ICMP time exceeded ICMP error) so you can tell what physical machine it came from.
Sep
1
comment How to convince my Administrator that Java ON A SERVER is not insecure per se?
@MichaelHampton I don't see why you think that claim appears bizarre. It's quite straightforward. Unless a process locks something in memory, the OS backs a mapping with RAM only if it thinks that's the best use of the RAM.
Sep
1
comment How to convince my Administrator that Java ON A SERVER is not insecure per se?
The data is not clobbered. It is opportunistically saved to swap before the system is under memory pressure. That allows the physical page to be considered discardable until it's modified. This also helps to minimize the I/O demands should the system later come under memory pressure since fewer writes are needed. The disk cache won't be reduced by the outstanding memory allocations unless the OS thinks that's the best use of physical memory. It has choices. (Hence all the "why is my system using swap when there's plenty of free RAM" questions -- it's to make the disk cache bigger.)